@misc{RoserMeinikheimMendeletal., author = {Roser, David and Meinikheim, Michael and Mendel, Robert and Palm, Christoph and Probst, Andreas and Muzalyova, Anna and Scheppach, Markus W. and Nagl, Sandra and Schnoy, Elisabeth and R{\"o}mmele, Christoph and Schulz, Dominik Andreas Helmut Otto and Schlottmann, Jakob and Prinz, Friederike and Rauber, David and R{\"u}ckert, Tobias and Matsumura, Tomoaki and Fernandez-Esparrach, G. and Parsa, Nasim and Byrne, Michael F. and Messmann, Helmut and Ebigbo, Alanna}, title = {Human-Computer Interaction: Impact of Artificial Intelligence on the diagnostic confidence of endoscopists assessing videos of Barrett's esophagus}, series = {Endoscopy}, volume = {56}, journal = {Endoscopy}, number = {S 02}, publisher = {Georg Thieme Verlag}, issn = {1438-8812}, doi = {10.1055/s-0044-1782859}, pages = {79}, abstract = {Aims Human-computer interactions (HCI) may have a relevant impact on the performance of Artificial Intelligence (AI). Studies show that although endoscopists assessing Barrett's esophagus (BE) with AI improve their performance significantly, they do not achieve the level of the stand-alone performance of AI. One aspect of HCI is the impact of AI on the degree of certainty and confidence displayed by the endoscopist. Indirectly, diagnostic confidence when using AI may be linked to trust and acceptance of AI. In a BE video study, we aimed to understand the impact of AI on the diagnostic confidence of endoscopists and the possible correlation with diagnostic performance. Methods 22 endoscopists from 12 centers with varying levels of BE experience reviewed ninety-six standardized endoscopy videos. Endoscopists were categorized into experts and non-experts and randomly assigned to assess the videos with and without AI. Participants were randomized in two arms: Arm A assessed videos first without AI and then with AI, while Arm B assessed videos in the opposite order. Evaluators were tasked with identifying BE-related neoplasia and rating their confidence with and without AI on a scale from 0 to 9. Results The utilization of AI in Arm A (without AI first, with AI second) significantly elevated confidence levels for experts and non-experts (7.1 to 8.0 and 6.1 to 6.6, respectively). Only non-experts benefitted from AI with a significant increase in accuracy (68.6\% to 75.5\%). Interestingly, while the confidence levels of experts without AI were higher than those of non-experts with AI, there was no significant difference in accuracy between these two groups (71.3\% vs. 75.5\%). In Arm B (with AI first, without AI second), experts and non-experts experienced a significant reduction in confidence (7.6 to 7.1 and 6.4 to 6.2, respectively), while maintaining consistent accuracy levels (71.8\% to 71.8\% and 67.5\% to 67.1\%, respectively). Conclusions AI significantly enhanced confidence levels for both expert and non-expert endoscopists. Endoscopists felt significantly more uncertain in their assessments without AI. Furthermore, experts with or without AI consistently displayed higher confidence levels than non-experts with AI, irrespective of comparable outcomes. These findings underscore the possible role of AI in improving diagnostic confidence during endoscopic assessment.}, language = {en} } @misc{RoserMeinikheimMendeletal., author = {Roser, David and Meinikheim, Michael and Mendel, Robert and Palm, Christoph and Muzalyova, Anna and Rauber, David and R{\"u}ckert, Tobias and Parsa, Nasim and Byrne, Michael F. and Messmann, Helmut and Ebigbo, Alanna}, title = {Mensch-Maschine-Interaktion: Einfluss k{\"u}nstlicher Intelligenz auf das diagnostische Vertrauen von Endoskopikern bei der Beurteilung des Barrett-{\"O}sophagus}, series = {Zeitschrift f{\"u}r Gastroenterologie}, volume = {62}, journal = {Zeitschrift f{\"u}r Gastroenterologie}, number = {09}, publisher = {Georg Thieme Verlag KG}, doi = {10.1055/s-0044-1789656}, pages = {e575 -- e576}, abstract = {Ziele: Das Ziel der Studie war es, den Einfluss von KI auf die diagnostische Sicherheit (Konfidenzniveau) von Endoskopikern anhand von B{\"O}-Videos zu untersuchen und m{\"o}gliche Korrelationen mit der Untersuchungsqualit{\"a}t zu erforschen. Methodik: 22 Endoskopiker aus zw{\"o}lf Zentren mit unterschiedlicher Barrett-Erfahrung untersuchten 96 standardisierte Endoskopievideos. Die Untersucher wurden in Experten und Nicht-Experten eingeteilt und nach dem Zufallsprinzip f{\"u}r die Bewertung der Videos mit oder ohne KI eingeteilt. Die Teilnehmer wurden in zwei Gruppen aufgeteilt: Arm A bewertete zun{\"a}chst Videos ohne KI und dann mit KI, w{\"a}hrend Arm B die umgekehrte Reihenfolge einhielt. Die Untersucher hatten die Aufgabe, B{\"O}-assoziierte Neoplasien zu erkennen und ihr Konfidenzniveau sowohl mit als auch ohne KI auf einer Skala von 0 bis 9 anzugeben. Ergebnis: In Arm A erh{\"o}hte der Einsatz von KI das Konfidenzniveau bei beiden signifikant (p<0.001). Bemerkenswert ist, dass jedoch nur Nicht-Experten durch die KI eine signifikante Verbesserung der Sensitivit{\"a}t und Spezifit{\"a}t (p<0.001 bzw. p<0.05) erfuhren. W{\"a}hrend Experten ohne KI im Vergleich zu Nicht-Experten mit KI ein h{\"o}heres Konfidenzniveau aufwiesen, gab es keinen signifikanten Unterschied in der Genauigkeit. In Arm B zeigten beide Gruppen eine signifikante Abnahme des Konfidenzniveaus (p<0.001) bei gleichbleibender Genauigkeit. Dar{\"u}ber hinaus wurden in 9\% der Entscheidungen trotz korrekter KI eine falsche Wahl getroffen. Schlussfolgerung: Der Einsatz k{\"u}nstlicher Intelligenz steigerte das Konfidenzniveau sowohl bei Experten als auch bei Nicht-Experten signifikant - ein Effekt, der im Studienmodell reversibel war. Dar{\"u}ber hinaus wiesen Experten mit oder ohne KI durchweg h{\"o}here Konfidenzniveaus auf als Nicht-Experten mit KI, trotz vergleichbarer Ergebnisse. Zudem konnte beobachtet werden, dass die Untersucher in 9\% der F{\"a}lle die KI zuungunsten des Patienten ignorierten.}, language = {de} } @article{WallnerGutbrodRauberetal., author = {Wallner, M. and Gutbrod, Max and Rauber, David and Ebigbo, Alanna and Probst, Andreas and Palm, Christoph and Messmann, Helmut and Roser, David}, title = {KI-gest{\"u}tzte Detektion und Segmentierung von Magenkarzinomen in westlichen endoskopischen Bilddaten anhand eines fine-tuned Deep-Learning Ansatzes}, series = {Zeitschrift f{\"u}r Gastroenterologie}, volume = {64}, journal = {Zeitschrift f{\"u}r Gastroenterologie}, number = {03}, publisher = {Thieme}, doi = {10.1055/s-0046-1817751}, pages = {e64 -- e65}, abstract = {Diese vorl{\"a}ufige monozentrische Studie zeigt, dass ein aus einem Barrett-{\"O}sophagus-KI-System feinjustiertes Deep-Learning-Modell Magenkarzinome in westlichen multimodalen endoskopischen Bilddaten zuverl{\"a}ssig detektieren und pr{\"a}zise segmentieren kann. Die hohe Segmentierungsgenauigkeit und Detektionssensitivit{\"a}t {\"u}ber verschiedene Bildmodalit{\"a}ten hinweg unterstreichen die Machbarkeit eines pathologiegest{\"u}tzten KI-Ansatzes auch in einer westlichen Niedriginzidenzpopulation. Aufgrund der ausschließlichen Verwendung von Bildern mit sichtbaren Tumoren lassen sich keine Aussagen zur Spezifit{\"a}t treffen; eine {\"U}bertragbarkeit auf Screening- oder Mischkollektive ist daher limitiert. Weitere Studien mit a) gr{\"o}ßerem Datensatz inklusive Videodaten, b) externer Validierung an einer multizentrischen westlichen Kohorte, sowie c) Anwendung und Pr{\"u}fung an nicht-neoplastischen Vergleichsbildern oder anderen Pathologien sind erforderlich. Nach unserem Kenntnisstand z{\"a}hlt dieses System zu den ersten in einer westlichen Population entwickelten endoskopischen KI-Ans{\"a}tzen zur Magenkarzinomdetektion, und zu wenigen, die vollst{\"a}ndige ESD-pr{\"a}paratbasierte Referenzdaten f{\"u}r Training und Validierung nutzen.}, language = {de} } @article{EbigboMendelScheppachetal., author = {Ebigbo, Alanna and Mendel, Robert and Scheppach, Markus W. and Probst, Andreas and Shahidi, Neal and Prinz, Friederike and Fleischmann, Carola and R{\"o}mmele, Christoph and G{\"o}lder, Stefan Karl and Braun, Georg and Rauber, David and R{\"u}ckert, Tobias and Souza Jr., Luis Antonio de and Papa, Jo{\~a}o Paulo and Byrne, Michael F. and Palm, Christoph and Messmann, Helmut}, title = {Vessel and tissue recognition during third-space endoscopy using a deep learning algorithm}, series = {Gut}, volume = {71}, journal = {Gut}, number = {12}, publisher = {BMJ}, address = {London}, doi = {10.1136/gutjnl-2021-326470}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:898-opus4-54293}, pages = {2388 -- 2390}, abstract = {In this study, we aimed to develop an artificial intelligence clinical decision support solution to mitigate operator-dependent limitations during complex endoscopic procedures such as endoscopic submucosal dissection and peroral endoscopic myotomy, for example, bleeding and perforation. A DeepLabv3-based model was trained to delineate vessels, tissue structures and instruments on endoscopic still images from such procedures. The mean cross-validated Intersection over Union and Dice Score were 63\% and 76\%, respectively. Applied to standardised video clips from third-space endoscopic procedures, the algorithm showed a mean vessel detection rate of 85\% with a false-positive rate of 0.75/min. These performance statistics suggest a potential clinical benefit for procedure safety, time and also training.}, language = {en} } @article{MeinikheimMendelPalmetal., author = {Meinikheim, Michael and Mendel, Robert and Palm, Christoph and Probst, Andreas and Muzalyova, Anna and Scheppach, Markus W. and Nagl, Sandra and Schnoy, Elisabeth and R{\"o}mmele, Christoph and Schulz, Dominik Andreas Helmut Otto and Schlottmann, Jakob and Prinz, Friederike and Rauber, David and R{\"u}ckert, Tobias and Matsumura, Tomoaki and Fern{\´a}ndez-Esparrach, Gl{\`o}ria and Parsa, Nasim and Byrne, Michael F. and Messmann, Helmut and Ebigbo, Alanna}, title = {Influence of artificial intelligence on the diagnostic performance of endoscopists in the assessment of Barrett's esophagus: a tandem randomized and video trial}, series = {Endoscopy}, volume = {56}, journal = {Endoscopy}, publisher = {Georg Thieme Verlag}, address = {Stuttgart}, doi = {10.1055/a-2296-5696}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:898-opus4-72818}, pages = {641 -- 649}, abstract = {Background This study evaluated the effect of an artificial intelligence (AI)-based clinical decision support system on the performance and diagnostic confidence of endoscopists in their assessment of Barrett's esophagus (BE). Methods 96 standardized endoscopy videos were assessed by 22 endoscopists with varying degrees of BE experience from 12 centers. Assessment was randomized into two video sets: group A (review first without AI and second with AI) and group B (review first with AI and second without AI). Endoscopists were required to evaluate each video for the presence of Barrett's esophagus-related neoplasia (BERN) and then decide on a spot for a targeted biopsy. After the second assessment, they were allowed to change their clinical decision and confidence level. Results AI had a stand-alone sensitivity, specificity, and accuracy of 92.2\%, 68.9\%, and 81.3\%, respectively. Without AI, BE experts had an overall sensitivity, specificity, and accuracy of 83.3\%, 58.1\%, and 71.5\%, respectively. With AI, BE nonexperts showed a significant improvement in sensitivity and specificity when videos were assessed a second time with AI (sensitivity 69.8\% [95\%CI 65.2\%-74.2\%] to 78.0\% [95\%CI 74.0\%-82.0\%]; specificity 67.3\% [95\%CI 62.5\%-72.2\%] to 72.7\% [95\%CI 68.2\%-77.3\%]). In addition, the diagnostic confidence of BE nonexperts improved significantly with AI. Conclusion BE nonexperts benefitted significantly from additional AI. BE experts and nonexperts remained significantly below the stand-alone performance of AI, suggesting that there may be other factors influencing endoscopists' decisions to follow or discard AI advice.}, language = {en} } @misc{EbigboRauberAyoubetal., author = {Ebigbo, Alanna and Rauber, David and Ayoub, Mousa and Birzle, Lisa and Matsumura, Tomoaki and Probst, Andreas and Steinbr{\"u}ck, Ingo and Nagl, Sandra and R{\"o}mmele, Christoph and Meinikheim, Michael and Scheppach, Markus W. and Palm, Christoph and Messmann, Helmut}, title = {Early Esophageal Cancer and the Generalizability of Artificial Intelligence}, series = {Endoscopy}, volume = {56}, journal = {Endoscopy}, number = {S 02}, publisher = {Thieme}, address = {Stuttgart}, doi = {10.1055/s-0044-1783775}, pages = {S428}, abstract = {Aims Artificial Intelligence (AI) systems in gastrointestinal endoscopy are narrow because they are trained to solve only one specific task. Unlike Narrow-AI, general AI systems may be able to solve multiple and unrelated tasks. We aimed to understand whether an AI system trained to detect, characterize, and segment early Barrett's neoplasia (Barrett's AI) is only capable of detecting this pathology or can also detect and segment other diseases like early squamous cell cancer (SCC). Methods 120 white light (WL) and narrow-band endoscopic images (NBI) from 60 patients (1 WL and 1 NBI image per patient) were extracted from the endoscopic database of the University Hospital Augsburg. Images were annotated by three expert endoscopists with extensive experience in the diagnosis and endoscopic resection of early esophageal neoplasias. An AI system based on DeepLabV3+architecture dedicated to early Barrett's neoplasia was tested on these images. The AI system was neither trained with SCC images nor had it seen the test images prior to evaluation. The overlap between the three expert annotations („expert-agreement") was the ground truth for evaluating AI performance. Results Barrett's AI detected early SCC with a mean intersection over reference (IoR) of 92\% when at least 1 pixel of the AI prediction overlapped with the expert-agreement. When the threshold was increased to 5\%, 10\%, and 20\% overlap with the expert-agreement, the IoR was 88\%, 85\% and 82\%, respectively. The mean Intersection Over Union (IoU) - a metric according to segmentation quality between the AI prediction and the expert-agreement - was 0.45. The mean expert IoU as a measure of agreement between the three experts was 0.60. Conclusions In the context of this pilot study, the predictions of SCC by a Barrett's dedicated AI showed some overlap to the expert-agreement. Therefore, features learned from Barrett's cancer-related training might be helpful also for SCC prediction. Our results allow different possible explanations. On the one hand, some Barrett's cancer features generalize toward the related task of assessing early SCC. On the other hand, the Barrett's AI is less specific to Barrett's cancer than a general predictor of pathological tissue. However, we expect to enhance the detection quality significantly by extending the training to SCC-specific data. The insight of this study opens the way towards a transfer learning approach for more efficient training of AI to solve tasks in other domains.}, language = {en} } @misc{ScheppachMendelRauberetal., author = {Scheppach, Markus W. and Mendel, Robert and Rauber, David and Probst, Andreas and Nagl, Sandra and R{\"o}mmele, Christoph and Meinikheim, Michael and Palm, Christoph and Messmann, Helmut and Ebigbo, Alanna}, title = {Artificial Intelligence (AI) improves endoscopists' vessel detection during endoscopic submucosal dissection (ESD)}, series = {Endoscopy}, volume = {56}, journal = {Endoscopy}, number = {S 02}, publisher = {Thieme}, address = {Stuttgart}, doi = {10.1055/s-0044-1782891}, pages = {S93}, abstract = {Aims While AI has been successfully implemented in detecting and characterizing colonic polyps, its role in therapeutic endoscopy remains to be elucidated. Especially third space endoscopy procedures like ESD and peroral endoscopic myotomy (POEM) pose a technical challenge and the risk of operator-dependent complications like intraprocedural bleeding and perforation. Therefore, we aimed at developing an AI-algorithm for intraprocedural real time vessel detection during ESD and POEM. Methods A training dataset consisting of 5470 annotated still images from 59 full-length videos (47 ESD, 12 POEM) and 179681 unlabeled images was used to train a DeepLabV3+neural network with the ECMT semi-supervised learning method. Evaluation for vessel detection rate (VDR) and time (VDT) of 19 endoscopists with and without AI-support was performed using a testing dataset of 101 standardized video clips with 200 predefined blood vessels. Endoscopists were stratified into trainees and experts in third space endoscopy. Results The AI algorithm had a mean VDR of 93.5\% and a median VDT of 0.32 seconds. AI support was associated with a statistically significant increase in VDR from 54.9\% to 73.0\% and from 59.0\% to 74.1\% for trainees and experts, respectively. VDT significantly decreased from 7.21 sec to 5.09 sec for trainees and from 6.10 sec to 5.38 sec for experts in the AI-support group. False positive (FP) readings occurred in 4.5\% of frames. FP structures were detected significantly shorter than true positives (0.71 sec vs. 5.99 sec). Conclusions AI improved VDR and VDT of trainees and experts in third space endoscopy and may reduce performance variability during training. Further research is needed to evaluate the clinical impact of this new technology.}, language = {en} } @misc{ZellmerRauberProbstetal., author = {Zellmer, Stephan and Rauber, David and Probst, Andreas and Weber, Tobias and Braun, Georg and R{\"o}mmele, Christoph and Nagl, Sandra and Schnoy, Elisabeth and Messmann, Helmut and Ebigbo, Alanna and Palm, Christoph}, title = {Artificial intelligence as a tool in the detection of the papillary ostium during ERCP}, series = {Endoscopy}, volume = {56}, journal = {Endoscopy}, number = {S 02}, publisher = {Thieme}, address = {Stuttgart}, doi = {10.1055/s-0044-1783138}, pages = {S198}, abstract = {Aims Endoscopic retrograde cholangiopancreaticography (ERCP) is the gold standard in the diagnosis as well as treatment of diseases of the pancreatobiliary tract. However, it is technically complex and has a relatively high complication rate. In particular, cannulation of the papillary ostium remains challenging. The aim of this study is to examine whether a deep-learning algorithm can be used to detect the major duodenal papilla and in particular the papillary ostium reliably and could therefore be a valuable tool for inexperienced endoscopists, particularly in training situation. Methods We analyzed a total of 654 retrospectively collected images of 85 patients. Both the major duodenal papilla and the ostium were then segmented. Afterwards, a neural network was trained using a deep-learning algorithm. A 5-fold cross-validation was performed. Subsequently, we ran the algorithm on 5 prospectively collected videos of ERCPs. Results 5-fold cross-validation on the 654 labeled data resulted in an F1 value of 0.8007, a sensitivity of 0.8409 and a specificity of 0.9757 for the class papilla, and an F1 value of 0.5724, a sensitivity of 0.5456 and a specificity of 0.9966 for the class ostium. Regardless of the class, the average F1 value (class papilla and class ostium) was 0.6866, the sensitivity 0.6933 and the specificity 0.9861. In 100\% of cases the AI-detected localization of the papillary ostium in the prospectively collected videos corresponded to the localization of the cannulation performed by the endoscopist. Conclusions In the present study, the neural network was able to identify the major duodenal papilla with a high sensitivity and high specificity. In detecting the papillary ostium, the sensitivity was notably lower. However, when used on videos, the AI was able to identify the location of the subsequent cannulation with 100\% accuracy. In the future, the neural network will be trained with more data. Thus, a suitable tool for ERCP could be established, especially in the training situation.}, language = {en} } @misc{ScheppachNunesArizietal., author = {Scheppach, Markus W. and Nunes, Danilo Weber and Arizi, X. and Rauber, David and Probst, Andreas and Nagl, Sandra and R{\"o}mmele, Christoph and Meinikheim, Michael and Palm, Christoph and Messmann, Helmut and Ebigbo, Alanna}, title = {Procedural phase recognition in endoscopic submucosal dissection (ESD) using artificial intelligence (AI)}, series = {Endoscopy}, volume = {56}, journal = {Endoscopy}, number = {S 02}, publisher = {Thieme}, address = {Stuttgart}, doi = {10.1055/s-0044-1783804}, pages = {S439}, abstract = {Aims Recent evidence suggests the possibility of intraprocedural phase recognition in surgical operations as well as endoscopic interventions such as peroral endoscopic myotomy and endoscopic submucosal dissection (ESD) by AI-algorithms. The intricate measurement of intraprocedural phase distribution may deepen the understanding of the procedure. Furthermore, real-time quality assessment as well as automation of reporting may become possible. Therefore, we aimed to develop an AI-algorithm for intraprocedural phase recognition during ESD. Methods A training dataset of 364385 single images from 9 full-length ESD videos was compiled. Each frame was classified into one procedural phase. Phases included scope manipulation, marking, injection, application of electrical current and bleeding. Allocation of each frame was only possible to one category. This training dataset was used to train a Video Swin transformer to recognize the phases. Temporal information was included via logarithmic frame sampling. Validation was performed using two separate ESD videos with 29801 single frames. Results The validation yielded sensitivities of 97.81\%, 97.83\%, 95.53\%, 85.01\% and 87.55\% for scope manipulation, marking, injection, electric application and bleeding, respectively. Specificities of 77.78\%, 90.91\%, 95.91\%, 93.65\% and 84.76\% were measured for the same parameters. Conclusions The developed algorithm was able to classify full-length ESD videos on a frame-by-frame basis into the predefined classes with high sensitivities and specificities. Future research will aim at the development of quality metrics based on single-operator phase distribution.}, language = {en} } @misc{ScheppachRauberStallhoferetal., author = {Scheppach, Markus W. and Rauber, David and Stallhofer, Johannes and Muzalyova, Anna and Otten, Vera and Manzeneder, Carolin and Schwamberger, Tanja and Wanzl, Julia and Schlottmann, Jakob and Tadic, Vidan and Probst, Andreas and Schnoy, Elisabeth and R{\"o}mmele, Christoph and Fleischmann, Carola and Meinikheim, Michael and Miller, Silvia and M{\"a}rkl, Bruno and Palm, Christoph and Messmann, Helmut and Ebigbo, Alanna}, title = {Performance comparison of a deep learning algorithm with endoscopists in the detection of duodenal villous atrophy (VA)}, series = {Endoscopy}, volume = {55}, journal = {Endoscopy}, number = {S02}, publisher = {Thieme}, doi = {10.1055/s-0043-1765421}, pages = {S165}, abstract = {Aims VA is an endoscopic finding of celiac disease (CD), which can easily be missed if pretest probability is low. In this study, we aimed to develop an artificial intelligence (AI) algorithm for the detection of villous atrophy on endoscopic images. Methods 858 images from 182 patients with VA and 846 images from 323 patients with normal duodenal mucosa were used for training and internal validation of an AI algorithm (ResNet18). A separate dataset was used for external validation, as well as determination of detection performance of experts, trainees and trainees with AI support. According to the AI consultation distribution, images were stratified into "easy" and "difficult". Results Internal validation showed 82\%, 85\% and 84\% for sensitivity, specificity and accuracy. External validation showed 90\%, 76\% and 84\%. The algorithm was significantly more sensitive and accurate than trainees, trainees with AI support and experts in endoscopy. AI support in trainees was associated with significantly improved performance. While all endoscopists showed significantly lower detection for "difficult" images, AI performance remained stable. Conclusions The algorithm outperformed trainees and experts in sensitivity and accuracy for VA detection. The significant improvement with AI support suggests a potential clinical benefit. Stable performance of the algorithm in "easy" and "difficult" test images may indicate an advantage in macroscopically challenging cases.}, language = {en} } @misc{ScheppachWeberNunesArizietal., author = {Scheppach, Markus W. and Weber Nunes, Danilo and Arizi, X. and Rauber, David and Probst, Andreas and Nagl, Sandra and R{\"o}mmele, Christoph and Palm, Christoph and Messmann, Helmut and Ebigbo, Alanna}, title = {Single frame workflow recognition during endoscopic submucosal dissection (ESD) using artificial intelligence (AI)}, series = {Endoscopy}, volume = {57}, journal = {Endoscopy}, number = {S 02}, publisher = {Thieme}, address = {Stuttgart}, doi = {10.1055/s-0045-1806324}, pages = {S511}, abstract = {Aims Precise surgical phase recognition and evaluation may improve our understanding of complex endoscopic procedures. Furthermore, quality control measurements and endoscopy training could benefit from objective descriptions of surgical phase distributions. Therefore, we aimed to develop an artificial intelligence algorithm for frame-by-frame operational phase recognition during endoscopic submucosal dissection (ESD). Methods Full length ESD-videos from 31 patients comprising 6.297.782 single images were collected retrospectively. Videos were annotated on a frame-by-frame basis for the operational macro-phases diagnostics, marking, injection, dissection and bleeding. Further subphases were the application of electrical current, visible injection of fluid into the submucosal space and scope manipulation, leading to 11 phases in total. 4.975.699 frames (21 patients) were used for training of a video swin transformer using uniform frame sampling for temporal information. Hyperparameter tuning was performed with 897.325 further frames (6 patients), while 424.758 frames (4 patients) were used for validation. Results The overall F1 scores on the test dataset for the macro-phases and all 11 phases were 0.96 and 0.90, respectively. The recall values for diagnostics, marking, injection, dissection and bleeding were 1.00, 1.00, 0.95, 0.96 and 0.93, respectively. Conclusions The algorithm classified operational phases during ESD with high accuracy. A precise evaluation of phase distribution may allow for the development of objective quality metrics for quality control and training.}, language = {en} } @misc{ZellmerRauberProbstetal., author = {Zellmer, Stephan and Rauber, David and Probst, Andreas and Weber, Tobias and Braun, Georg and Nagl, Sandra and R{\"o}mmele, Christoph and Schnoy, Elisabeth and Birzle, Lisa and Aehling, Niklas and Schulz, Dominik Andreas Helmut Otto and Palm, Christoph and Messmann, Helmut and Ebigbo, Alanna}, title = {K{\"u}nstliche Intelligenz als Hilfsmittel zur Detektion der Papilla duodeni major und des papill{\"a}ren Ostiums w{\"a}hrend der ERCP}, series = {Zeitschrift f{\"u}r Gastroenterologie}, volume = {63}, journal = {Zeitschrift f{\"u}r Gastroenterologie}, number = {5}, publisher = {Thieme}, address = {Stuttgart}, doi = {10.1055/s-0045-1806882}, pages = {e295}, abstract = {Einleitung Die Endoskopische Retrograde Cholangiopankreatikographie (ERCP) ist der Goldstandard in der endoskopischen Therapie von Erkrankungen des pankreatobili{\"a}ren Trakts. Allerdings ist sie technisch anspruchsvoll, schwer zu erlernen und mit einer relativ hohen Komplikationsrate assoziiert. Daher soll in der vorliegenden Machbarkeitsstudie gepr{\"u}ft werden, ob mithilfe eines Deeplearning- Algorithmus die Papille und das Ostium zuverl{\"a}ssig detektiert werden k{\"o}nnen und dieser f{\"u}r Endoskopiker, insbesondere in der Ausbildungssituation, ein geeignetes Hilfsmittel darstellen k{\"o}nnte. Material und Methodik Insgesamt wurden 1534 ERCP-Bilder von 134 Patienten analysiert, wobei sowohl die Papilla duodeni major als auch das Ostium segmentiert wurden. Anschließend erfolgte das Training eines neuronalen Netzes unter Verwendung eines Deep-Learning-Algorithmus. F{\"u}r den Test des Algorithmus erfolgte eine f{\"u}nffache Kreuzvalidierung. Ergebnisse Auf den 1534 gelabelten Bildern wurden f{\"u}r die Klasse Papille ein F1-Wert von 0,7996, eine Sensitivit{\"a}t von 0,8488 und eine Spezifit{\"a}t von 0,9822 erzielt. F{\"u}r die Klasse Ostium ergaben sich ein F1-Wert von 0,5198, eine Sensitivit{\"a}t von 0,5945 und eine Spezifit{\"a}t von 0,9974. Klassen{\"u}bergreifend (Klasse Papille und Klasse Ostium) betrug der F1-Wert 0,6593, die Sensitivit{\"a}t 0,7216 und f{\"u}r die Spezifit{\"a}t 0,9898. Zusammenfassung In der vorliegenden Machbarkeitsstudie zeigte das neuronale Netz eine hohe Sensitivit{\"a}t und eine sehr hohe Spezifit{\"a}t bei der Identifikation der Papilla duodeni major. Die Detektion des Ostiums erfolgte hingegen mit einer deutlich geringeren Sensitivit{\"a}t. Zuk{\"u}nftig ist eine Erweiterung des Trainingsdatensatzes um Videos und klinische Daten vorgesehen, um die Leistungsf{\"a}higkeit des Netzwerks zu verbessern. Hierdurch k{\"o}nnte langfristig ein geeignetes Assistenzsystem f{\"u}r die ERCP, insbesondere in der Ausbildungssituation etabliert werden.}, language = {de} } @article{EbigboMendelRueckertetal., author = {Ebigbo, Alanna and Mendel, Robert and R{\"u}ckert, Tobias and Schuster, Laurin and Probst, Andreas and Manzeneder, Johannes and Prinz, Friederike and Mende, Matthias and Steinbr{\"u}ck, Ingo and Faiss, Siegbert and Rauber, David and Souza Jr., Luis Antonio de and Papa, Jo{\~a}o Paulo and Deprez, Pierre and Oyama, Tsuneo and Takahashi, Akiko and Seewald, Stefan and Sharma, Prateek and Byrne, Michael F. and Palm, Christoph and Messmann, Helmut}, title = {Endoscopic prediction of submucosal invasion in Barrett's cancer with the use of Artificial Intelligence: A pilot Study}, series = {Endoscopy}, volume = {53}, journal = {Endoscopy}, number = {09}, publisher = {Thieme}, address = {Stuttgart}, doi = {10.1055/a-1311-8570}, pages = {878 -- 883}, abstract = {Background and aims: The accurate differentiation between T1a and T1b Barrett's cancer has both therapeutic and prognostic implications but is challenging even for experienced physicians. We trained an Artificial Intelligence (AI) system on the basis of deep artificial neural networks (deep learning) to differentiate between T1a and T1b Barrett's cancer white-light images. Methods: Endoscopic images from three tertiary care centres in Germany were collected retrospectively. A deep learning system was trained and tested using the principles of cross-validation. A total of 230 white-light endoscopic images (108 T1a and 122 T1b) was evaluated with the AI-system. For comparison, the images were also classified by experts specialized in endoscopic diagnosis and treatment of Barrett's cancer. Results: The sensitivity, specificity, F1 and accuracy of the AI-system in the differentiation between T1a and T1b cancer lesions was 0.77, 0.64, 0.73 and 0.71, respectively. There was no statistically significant difference between the performance of the AI-system and that of human experts with sensitivity, specificity, F1 and accuracy of 0.63, 0.78, 0.67 and 0.70 respectively. Conclusion: This pilot study demonstrates the first multicenter application of an AI-based system in the prediction of submucosal invasion in endoscopic images of Barrett's cancer. AI scored equal to international experts in the field, but more work is necessary to improve the system and apply it to video sequences and in a real-life setting. Nevertheless, the correct prediction of submucosal invasion in Barret´s cancer remains challenging for both experts and AI.}, subject = {Maschinelles Lernen}, language = {en} } @article{MendelRauberSouzaJretal., author = {Mendel, Robert and Rauber, David and Souza Jr., Luis Antonio de and Papa, Jo{\~a}o Paulo and Palm, Christoph}, title = {Error-Correcting Mean-Teacher: Corrections instead of consistency-targets applied to semi-supervised medical image segmentation}, series = {Computers in Biology and Medicine}, volume = {154}, journal = {Computers in Biology and Medicine}, number = {March}, publisher = {Elsevier}, issn = {0010-4825}, doi = {10.1016/j.compbiomed.2023.106585}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:898-opus4-57790}, pages = {13}, abstract = {Semantic segmentation is an essential task in medical imaging research. Many powerful deep-learning-based approaches can be employed for this problem, but they are dependent on the availability of an expansive labeled dataset. In this work, we augment such supervised segmentation models to be suitable for learning from unlabeled data. Our semi-supervised approach, termed Error-Correcting Mean-Teacher, uses an exponential moving average model like the original Mean Teacher but introduces our new paradigm of error correction. The original segmentation network is augmented to handle this secondary correction task. Both tasks build upon the core feature extraction layers of the model. For the correction task, features detected in the input image are fused with features detected in the predicted segmentation and further processed with task-specific decoder layers. The combination of image and segmentation features allows the model to correct present mistakes in the given input pair. The correction task is trained jointly on the labeled data. On unlabeled data, the exponential moving average of the original network corrects the student's prediction. The combined outputs of the students' prediction with the teachers' correction form the basis for the semi-supervised update. We evaluate our method with the 2017 and 2018 Robotic Scene Segmentation data, the ISIC 2017 and the BraTS 2020 Challenges, a proprietary Endoscopic Submucosal Dissection dataset, Cityscapes, and Pascal VOC 2012. Additionally, we analyze the impact of the individual components and examine the behavior when the amount of labeled data varies, with experiments performed on two distinct segmentation architectures. Our method shows improvements in terms of the mean Intersection over Union over the supervised baseline and competing methods. Code is available at https://github.com/CloneRob/ECMT.}, language = {en} } @inproceedings{MendelSouzaJrRauberetal., author = {Mendel, Robert and Souza Jr., Luis Antonio de and Rauber, David and Papa, Jo{\~a}o Paulo and Palm, Christoph}, title = {Semi-supervised Segmentation Based on Error-Correcting Supervision}, series = {Computer vision - ECCV 2020: 16th European conference, Glasgow, UK, August 23-28, 2020, Proceedings, Part XXIX}, booktitle = {Computer vision - ECCV 2020: 16th European conference, Glasgow, UK, August 23-28, 2020, Proceedings, Part XXIX}, publisher = {Springer}, address = {Cham}, isbn = {978-3-030-58525-9}, doi = {10.1007/978-3-030-58526-6_9}, pages = {141 -- 157}, abstract = {Pixel-level classification is an essential part of computer vision. For learning from labeled data, many powerful deep learning models have been developed recently. In this work, we augment such supervised segmentation models by allowing them to learn from unlabeled data. Our semi-supervised approach, termed Error-Correcting Supervision, leverages a collaborative strategy. Apart from the supervised training on the labeled data, the segmentation network is judged by an additional network. The secondary correction network learns on the labeled data to optimally spot correct predictions, as well as to amend incorrect ones. As auxiliary regularization term, the corrector directly influences the supervised training of the segmentation network. On unlabeled data, the output of the correction network is essential to create a proxy for the unknown truth. The corrector's output is combined with the segmentation network's prediction to form the new target. We propose a loss function that incorporates both the pseudo-labels as well as the predictive certainty of the correction network. Our approach can easily be added to supervised segmentation models. We show consistent improvements over a supervised baseline on experiments on both the Pascal VOC 2012 and the Cityscapes datasets with varying amounts of labeled data.}, subject = {Semi-Supervised Learning}, language = {en} } @article{RoemmeleMendelBarrettetal., author = {R{\"o}mmele, Christoph and Mendel, Robert and Barrett, Caroline and Kiesl, Hans and Rauber, David and R{\"u}ckert, Tobias and Kraus, Lisa and Heinkele, Jakob and Dhillon, Christine and Grosser, Bianca and Prinz, Friederike and Wanzl, Julia and Fleischmann, Carola and Nagl, Sandra and Schnoy, Elisabeth and Schlottmann, Jakob and Dellon, Evan S. and Messmann, Helmut and Palm, Christoph and Ebigbo, Alanna}, title = {An artificial intelligence algorithm is highly accurate for detecting endoscopic features of eosinophilic esophagitis}, series = {Scientific Reports}, volume = {12}, journal = {Scientific Reports}, publisher = {Nature Portfolio}, address = {London}, doi = {10.1038/s41598-022-14605-z}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:898-opus4-46928}, pages = {10}, abstract = {The endoscopic features associated with eosinophilic esophagitis (EoE) may be missed during routine endoscopy. We aimed to develop and evaluate an Artificial Intelligence (AI) algorithm for detecting and quantifying the endoscopic features of EoE in white light images, supplemented by the EoE Endoscopic Reference Score (EREFS). An AI algorithm (AI-EoE) was constructed and trained to differentiate between EoE and normal esophagus using endoscopic white light images extracted from the database of the University Hospital Augsburg. In addition to binary classification, a second algorithm was trained with specific auxiliary branches for each EREFS feature (AI-EoE-EREFS). The AI algorithms were evaluated on an external data set from the University of North Carolina, Chapel Hill (UNC), and compared with the performance of human endoscopists with varying levels of experience. The overall sensitivity, specificity, and accuracy of AI-EoE were 0.93 for all measures, while the AUC was 0.986. With additional auxiliary branches for the EREFS categories, the AI algorithm (AI-EoEEREFS) performance improved to 0.96, 0.94, 0.95, and 0.992 for sensitivity, specificity, accuracy, and AUC, respectively. AI-EoE and AI-EoE-EREFS performed significantly better than endoscopy beginners and senior fellows on the same set of images. An AI algorithm can be trained to detect and quantify endoscopic features of EoE with excellent performance scores. The addition of the EREFS criteria improved the performance of the AI algorithm, which performed significantly better than endoscopists with a lower or medium experience level.}, language = {en} } @misc{RoemmeleMendelRauberetal., author = {R{\"o}mmele, Christoph and Mendel, Robert and Rauber, David and R{\"u}ckert, Tobias and Byrne, Michael F. and Palm, Christoph and Messmann, Helmut and Ebigbo, Alanna}, title = {Endoscopic Diagnosis of Eosinophilic Esophagitis Using a deep Learning Algorithm}, series = {Endoscopy}, volume = {53}, journal = {Endoscopy}, number = {S 01}, publisher = {Georg Thieme Verlag}, address = {Stuttgart}, doi = {10.1055/s-0041-1724274}, abstract = {Aims Eosinophilic esophagitis (EoE) is easily missed during endoscopy, either because physicians are not familiar with its endoscopic features or the morphologic changes are too subtle. In this preliminary paper, we present the first attempt to detect EoE in endoscopic white light (WL) images using a deep learning network (EoE-AI). Methods 401 WL images of eosinophilic esophagitis and 871 WL images of normal esophageal mucosa were evaluated. All images were assessed for the Endoscopic Reference score (EREFS) (edema, rings, exudates, furrows, strictures). Images with strictures were excluded. EoE was defined as the presence of at least 15 eosinophils per high power field on biopsy. A convolutional neural network based on the ResNet architecture with several five-fold cross-validation runs was used. Adding auxiliary EREFS-classification branches to the neural network allowed the inclusion of the scores as optimization criteria during training. EoE-AI was evaluated for sensitivity, specificity, and F1-score. In addition, two human endoscopists evaluated the images. Results EoE-AI showed a mean sensitivity, specificity, and F1 of 0.759, 0.976, and 0.834 respectively, averaged over the five distinct cross-validation runs. With the EREFS-augmented architecture, a mean sensitivity, specificity, and F1-score of 0.848, 0.945, and 0.861 could be demonstrated respectively. In comparison, the two human endoscopists had an average sensitivity, specificity, and F1-score of 0.718, 0.958, and 0.793. Conclusions To the best of our knowledge, this is the first application of deep learning to endoscopic images of EoE which were also assessed after augmentation with the EREFS-score. The next step is the evaluation of EoE-AI using an external dataset. We then plan to assess the EoE-AI tool on endoscopic videos, and also in real-time. This preliminary work is encouraging regarding the ability for AI to enhance physician detection of EoE, and potentially to do a true "optical biopsy" but more work is needed.}, language = {en} } @article{ScheppachRauberStallhoferetal., author = {Scheppach, Markus W. and Rauber, David and Stallhofer, Johannes and Muzalyova, Anna and Otten, Vera and Manzeneder, Carolin and Schwamberger, Tanja and Wanzl, Julia and Schlottmann, Jakob and Tadic, Vidan and Probst, Andreas and Schnoy, Elisabeth and R{\"o}mmele, Christoph and Fleischmann, Carola and Meinikheim, Michael and Miller, Silvia and M{\"a}rkl, Bruno and Stallmach, Andreas and Palm, Christoph and Messmann, Helmut and Ebigbo, Alanna}, title = {Detection of duodenal villous atrophy on endoscopic images using a deep learning algorithm}, series = {Gastrointestinal Endoscopy}, journal = {Gastrointestinal Endoscopy}, publisher = {Elsevier}, doi = {10.1016/j.gie.2023.01.006}, abstract = {Background and aims Celiac disease with its endoscopic manifestation of villous atrophy is underdiagnosed worldwide. The application of artificial intelligence (AI) for the macroscopic detection of villous atrophy at routine esophagogastroduodenoscopy may improve diagnostic performance. Methods A dataset of 858 endoscopic images of 182 patients with villous atrophy and 846 images from 323 patients with normal duodenal mucosa was collected and used to train a ResNet 18 deep learning model to detect villous atrophy. An external data set was used to test the algorithm, in addition to six fellows and four board certified gastroenterologists. Fellows could consult the AI algorithm's result during the test. From their consultation distribution, a stratification of test images into "easy" and "difficult" was performed and used for classified performance measurement. Results External validation of the AI algorithm yielded values of 90 \%, 76 \%, and 84 \% for sensitivity, specificity, and accuracy, respectively. Fellows scored values of 63 \%, 72 \% and 67 \%, while the corresponding values in experts were 72 \%, 69 \% and 71 \%, respectively. AI consultation significantly improved all trainee performance statistics. While fellows and experts showed significantly lower performance for "difficult" images, the performance of the AI algorithm was stable. Conclusion In this study, an AI algorithm outperformed endoscopy fellows and experts in the detection of villous atrophy on endoscopic still images. AI decision support significantly improved the performance of non-expert endoscopists. The stable performance on "difficult" images suggests a further positive add-on effect in challenging cases.}, language = {en} } @article{ScheppachMendelMuzalyovaetal., author = {Scheppach, Markus W. and Mendel, Robert and Muzalyova, Anna and Rauber, David and Probst, Andreas and Nagl, Sandra and R{\"o}mmele, Christoph and Yip, Hon Chi and Lau, Louis Ho Shing and G{\"o}lder, Stefan Karl and Schmidt, Arthur and Kouladouros, Konstantinos and Abdelhafez, Mohamed and Walter, Benjamin M. and Meinikheim, Michael and Chiu, Philip Wai Yan and Palm, Christoph and Messmann, Helmut and Ebigbo, Alanna}, title = {Artificial intelligence improves submucosal vessel detection during third space endoscopy}, series = {Endoscopy}, journal = {Endoscopy}, publisher = {Thieme}, address = {Stuttgart}, doi = {10.1055/a-2534-1164}, abstract = {Background and study aims: While artificial intelligence (AI) shows high potential in decision support for diagnostic gastrointestinal endoscopy, its role in therapeutic endoscopy remains unclear. Third space endoscopic procedures pose the risk of intraprocedural bleeding. Therefore, we aimed to develop an AI algorithm for intraprocedural blood vessel detection. Patients and Methods: Using a test dataset with 101 standardized video clips containing 200 predefined submucosal blood vessels, 19 endoscopists were evaluated for the vessel detection rate (VDR) and time (VDT) with and without support of an AI algorithm. Test subjects were grouped according to experience in ESD. Results: With AI support, endoscopists VDR increased from 56.4\% [CI 54.1-58.6] to 72.4\% [CI 70.3-74.4]. Endoscopists' VDT dropped from 6.7sec [CI 6.2-7.1] to 5.2sec [CI 4.8-5.7]. False positive (FP) readings appeared in 4.5\% of frames and were marked significantly shorter than true positives (6.0sec [CI 5.28-6.70] vs. 0.7sec [CI 0.55-0.87]). Conclusions: AI improved the vessel detection rate and time of endoscopists during third space endoscopy. While these data need to be corroborated by clinical trials, AI may prove to be an invaluable tool for the improvement of endoscopic interventions.}, language = {en} } @unpublished{AllanKondoBodenstedtetal., author = {Allan, Max and Kondo, Satoshi and Bodenstedt, Sebastian and Leger, Stefan and Kadkhodamohammadi, Rahim and Luengo, Imanol and Fuentes, Felix and Flouty, Evangello and Mohammed, Ahmed and Pedersen, Marius and Kori, Avinash and Alex, Varghese and Krishnamurthi, Ganapathy and Rauber, David and Mendel, Robert and Palm, Christoph and Bano, Sophia and Saibro, Guinther and Shih, Chi-Sheng and Chiang, Hsun-An and Zhuang, Juntang and Yang, Junlin and Iglovikov, Vladimir and Dobrenkii, Anton and Reddiboina, Madhu and Reddy, Anubhav and Liu, Xingtong and Gao, Cong and Unberath, Mathias and Kim, Myeonghyeon and Kim, Chanho and Kim, Chaewon and Kim, Hyejin and Lee, Gyeongmin and Ullah, Ihsan and Luna, Miguel and Park, Sang Hyun and Azizian, Mahdi and Stoyanov, Danail and Maier-Hein, Lena and Speidel, Stefanie}, title = {2018 Robotic Scene Segmentation Challenge}, doi = {10.48550/arXiv.2001.11190}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:898-opus4-50049}, pages = {11}, abstract = {In 2015 we began a sub-challenge at the EndoVis workshop at MICCAI in Munich using endoscope images of exvivo tissue with automatically generated annotations from robot forward kinematics and instrument CAD models. However, the limited background variation and simple motion rendered the dataset uninformative in learning about which techniques would be suitable for segmentation in real surgery. In 2017, at the same workshop in Quebec we introduced the robotic instrument segmentation dataset with 10 teams participating in the challenge to perform binary, articulating parts and type segmentation of da Vinci instruments. This challenge included realistic instrument motion and more complex porcine tissue as background and was widely addressed with modfications on U-Nets and other popular CNN architectures [1]. In 2018 we added to the complexity by introducing a set of anatomical objects and medical devices to the segmented classes. To avoid over-complicating the challenge, we continued with porcine data which is dramatically simpler than human tissue due to the lack of fatty tissue occluding many organs.}, subject = {Minimal-invasive Chirurgie}, language = {en} } @misc{ScheppachRauberMendeletal., author = {Scheppach, Markus W. and Rauber, David and Mendel, Robert and Palm, Christoph and Byrne, Michael F. and Messmann, Helmut and Ebigbo, Alanna}, title = {Detection Of Celiac Disease Using A Deep Learning Algorithm}, series = {Endoscopy}, volume = {53}, journal = {Endoscopy}, number = {S 01}, publisher = {Georg Thieme Verlag}, address = {Stuttgart}, doi = {10.1055/s-0041-1724970}, abstract = {Aims Celiac disease (CD) is a complex condition caused by an autoimmune reaction to ingested gluten. Due to its polymorphic manifestation and subtle endoscopic presentation, the diagnosis is difficult and thus the disorder is underreported. We aimed to use deep learning to identify celiac disease on endoscopic images of the small bowel. Methods Patients with small intestinal histology compatible with CD (MARSH classification I-III) were extracted retrospectively from the database of Augsburg University hospital. They were compared to patients with no clinical signs of CD and histologically normal small intestinal mucosa. In a first step MARSH III and normal small intestinal mucosa were differentiated with the help of a deep learning algorithm. For this, the endoscopic white light images were divided into five equal-sized subsets. We avoided splitting the images of one patient into several subsets. A ResNet-50 model was trained with the images from four subsets and then validated with the remaining subset. This process was repeated for each subset, such that each subset was validated once. Sensitivity, specificity, and harmonic mean (F1) of the algorithm were determined. Results The algorithm showed values of 0.83, 0.88, and 0.84 for sensitivity, specificity, and F1, respectively. Further data showing a comparison between the detection rate of the AI model and that of experienced endoscopists will be available at the time of the upcoming conference. Conclusions We present the first clinical report on the use of a deep learning algorithm for the detection of celiac disease using endoscopic images. Further evaluation on an external data set, as well as in the detection of CD in real-time, will follow. However, this work at least suggests that AI can assist endoscopists in the endoscopic diagnosis of CD, and ultimately may be able to do a true optical biopsy in live-time.}, language = {en} } @inproceedings{RauberMendelScheppachetal., author = {Rauber, David and Mendel, Robert and Scheppach, Markus W. and Ebigbo, Alanna and Messmann, Helmut and Palm, Christoph}, title = {Analysis of Celiac Disease with Multimodal Deep Learning}, series = {Bildverarbeitung f{\"u}r die Medizin 2022: Proceedings, German Workshop on Medical Image Computing, Heidelberg, June 26-28, 2022}, booktitle = {Bildverarbeitung f{\"u}r die Medizin 2022: Proceedings, German Workshop on Medical Image Computing, Heidelberg, June 26-28, 2022}, publisher = {Springer Vieweg}, address = {Wiesbaden}, doi = {10.1007/978-3-658-36932-3_25}, pages = {115 -- 120}, abstract = {Celiac disease is an autoimmune disorder caused by gluten that results in an inflammatory response of the small intestine.We investigated whether celiac disease can be detected using endoscopic images through a deep learning approach. The results show that additional clinical parameters can improve the classification accuracy. In this work, we distinguished between healthy tissue and Marsh III, according to the Marsh score system. We first trained a baseline network to classify endoscopic images of the small bowel into these two classes and then augmented the approach with a multimodality component that took the antibody status into account.}, language = {en} } @inproceedings{MendelRauberPalm, author = {Mendel, Robert and Rauber, David and Palm, Christoph}, title = {Exploring the Effects of Contrastive Learning on Homogeneous Medical Image Data}, series = {Bildverarbeitung f{\"u}r die Medizin 2023: Proceedings, German Workshop on Medical Image Computing, July 2- 4, 2023, Braunschweig}, booktitle = {Bildverarbeitung f{\"u}r die Medizin 2023: Proceedings, German Workshop on Medical Image Computing, July 2- 4, 2023, Braunschweig}, publisher = {Springer Vieweg}, address = {Wiesbaden}, doi = {10.1007/978-3-658-41657-7}, pages = {128 -- 13}, abstract = {We investigate contrastive learning in a multi-task learning setting classifying and segmenting early Barrett's cancer. How can contrastive learning be applied in a domain with few classes and low inter-class and inter-sample variance, potentially enabling image retrieval or image attribution? We introduce a data sampling strategy that mines per-lesion data for positive samples and keeps a queue of the recent projections as negative samples. We propose a masking strategy for the NT-Xent loss that keeps the negative set pure and removes samples from the same lesion. We show cohesion and uniqueness improvements of the proposed method in feature space. The introduction of the auxiliary objective does not affect the performance but adds the ability to indicate similarity between lesions. Therefore, the approach could enable downstream auto-documentation tasks on homogeneous medical image data.}, language = {en} } @misc{MendelSouzaJrRauberetal., author = {Mendel, Robert and Souza Jr., Luis Antonio de and Rauber, David and Papa, Jo{\~a}o Paulo and Palm, Christoph}, title = {Abstract: Semi-supervised Segmentation Based on Error-correcting Supervision}, series = {Bildverarbeitung f{\"u}r die Medizin 2021. Proceedings, German Workshop on Medical Image Computing, Regensburg, March 7-9, 2021}, journal = {Bildverarbeitung f{\"u}r die Medizin 2021. Proceedings, German Workshop on Medical Image Computing, Regensburg, March 7-9, 2021}, publisher = {Springer Vieweg}, address = {Wiesbaden}, isbn = {978-3-658-33197-9}, doi = {10.1007/978-3-658-33198-6_43}, pages = {178}, abstract = {Pixel-level classification is an essential part of computer vision. For learning from labeled data, many powerful deep learning models have been developed recently. In this work, we augment such supervised segmentation models by allowing them to learn from unlabeled data. Our semi-supervised approach, termed Error-Correcting Supervision, leverages a collaborative strategy. Apart from the supervised training on the labeled data, the segmentation network is judged by an additional network.}, subject = {Deep Learning}, language = {en} } @misc{RueckertRiederRauberetal., author = {R{\"u}ckert, Tobias and Rieder, Maximilian and Rauber, David and Xiao, Michel and Humolli, Eg and Feussner, Hubertus and Wilhelm, Dirk and Palm, Christoph}, title = {Augmenting instrument segmentation in video sequences of minimally invasive surgery by synthetic smoky frames}, series = {International Journal of Computer Assisted Radiology and Surgery}, volume = {18}, journal = {International Journal of Computer Assisted Radiology and Surgery}, number = {Suppl 1}, publisher = {Springer Nature}, doi = {10.1007/s11548-023-02878-2}, pages = {S54 -- S56}, language = {en} } @article{MaerklRueckertRauberetal., author = {Maerkl, Raphaela and Rueckert, Tobias and Rauber, David and Gutbrod, Max and Weber Nunes, Danilo and Palm, Christoph}, title = {Enhancing generalization in zero-shot multi-label endoscopic instrument classification}, series = {International Journal of Computer Assisted Radiology and Surgery}, volume = {20}, journal = {International Journal of Computer Assisted Radiology and Surgery}, publisher = {Springer Nature}, doi = {10.1007/s11548-025-03439-5}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:898-opus4-85674}, pages = {1577 -- 1587}, abstract = {Purpose Recognizing previously unseen classes with neural networks is a significant challenge due to their limited generalization capabilities. This issue is particularly critical in safety-critical domains such as medical applications, where accurate classification is essential for reliability and patient safety. Zero-shot learning methods address this challenge by utilizing additional semantic data, with their performance relying heavily on the quality of the generated embeddings. Methods This work investigates the use of full descriptive sentences, generated by a Sentence-BERT model, as class representations, compared to simpler category-based word embeddings derived from a BERT model. Additionally, the impact of z-score normalization as a post-processing step on these embeddings is explored. The proposed approach is evaluated on a multi-label generalized zero-shot learning task, focusing on the recognition of surgical instruments in endoscopic images from minimally invasive cholecystectomies. Results The results demonstrate that combining sentence embeddings and z-score normalization significantly improves model performance. For unseen classes, the AUROC improves from 43.9\% to 64.9\%, and the multi-label accuracy from 26.1\% to 79.5\%. Overall performance measured across both seen and unseen classes improves from 49.3\% to 64.9\% in AUROC and from 37.3\% to 65.1\% in multi-label accuracy, highlighting the effectiveness of our approach. Conclusion These findings demonstrate that sentence embeddings and z-score normalization can substantially enhance the generalization performance of zero-shot learning models. However, as the study is based on a single dataset, future work should validate the method across diverse datasets and application domains to establish its robustness and broader applicability.}, language = {en} } @misc{ScheppachWeberNunesRauberetal., author = {Scheppach, Markus W. and Weber Nunes, Danilo and Rauber, David and Arizi, X. and Probst, Andreas and Nagl, Sandra and R{\"o}mmele, Christoph and Ebigbo, Alanna and Palm, Christoph and Messmann, Helmut}, title = {K{\"u}nstliche Intelligenz-basierte Erkennung von interventionellen Phasen bei der endoskopischen Submukosadissektion}, series = {Zeitschrift f{\"u}r Gastroenterologie}, volume = {63}, journal = {Zeitschrift f{\"u}r Gastroenterologie}, number = {08}, publisher = {Thieme}, address = {Stuttgart}, doi = {10.1055/s-0045-1811093}, pages = {e612 -- e613}, abstract = {Einleitung: Die endoskopische Submukosadissektion (ESD) ist ein komplexes endoskopisches Verfahren, das technische Expertise erfordert. Objektive Methoden zur Analyse von interventionellen Abl{\"a}ufen bei ESD k{\"o}nnten f{\"u}r Qualit{\"a}tssicherung und Ausbildung, wie auch eine automatische Befunderstellung von Nutzen sein. Ziele: In dieser Studie wurde ein KI-Algorithmus f{\"u}r die Erkennung und Klassifizierung der interventionellen Phasen der ESD entwickelt, um die technische Basis f{\"u}r eine standardisierte Leistungsbewertung und automatische Befunderstellung zu schaffen. Methodik: Vollst{\"a}ndige ESD-Videoaufnahmen von 49 Patienten wurden retrospektiv zusammengestellt. Der Datensatz umfasste 6.390.151 Einzelbilder, die alle f{\"u}r die folgenden interventionellen Phasen annotiert wurden: Diagnostik, Markierung, Injektion, Dissektion und H{\"a}mostase. 3.973.712 Bilder (28 Patienten) wurden f{\"u}r das Training eines Video-Swin-Transformers genutzt. Dabei wurde temporale Information durch standardisierte BIldextraktion in festgelegten zeitlichen Abst{\"a}nden zum analysierten Bild inkorporiert. 2.416.439 separate Bilder (21 Patienten) wurden f{\"u}r eine interne Validierung genutzt. Ergebnis: Bei der internen Evaluation erreichte das System insgesamt einen F1-Wert von 0,88. Es wurden F1-Werte von 0,99, 0,89, 0,89, 0,91 und 0,52 f{\"u}r Diagnostik, Markierung, Injektion, Dissektion bzw. Blutungsmanagement gemessen. Die Sensitivit{\"a}ten f{\"u}r dieselben Parameter betrugen 1,00, 0,80, 0,94, 0,89 und 0,67, die Spezifit{\"a}ten lagen bei 1,00, 1,00, 0,98, 0,88 und 0,93. Positive pr{\"a}diktive Werte wurden mit 0,98, 1,00, 0,85, 0,94 und 0,43 gemessen. Schlussfolgerung: In dieser vorl{\"a}ufigen Studie zeigte ein KI-Algorithmus eine hohe Leistungsf{\"a}higkeit f{\"u}r die Einzelbild-Erkennung von Verfahrensphasen w{\"a}hrend der ESD. Die vergleichsweise niedrige Leistung f{\"u}r die Blutungsphase wurde auf das seltene Auftreten von Blutungsepisoden im Trainingsdatensatz zur{\"u}ckgef{\"u}hrt, der zu diesem Zeitpunkt nur Videos in voller L{\"a}nge umfasste. Die zuk{\"u}nftige Entwicklung des Algorithmus wird sich auf die Reduzierung von Klassenungleichgewichten durch selektive Annotationsprotokolle konzentrieren.}, language = {de} } @misc{ScheppachRauberZingleretal., author = {Scheppach, Markus W. and Rauber, David and Zingler, C. and Weber Nunes, Danilo and Probst, Andreas and R{\"o}mmele, Christoph and Nagl, Sandra and Ebigbo, Alanna and Palm, Christoph and Messmann, Helmut}, title = {Instrumentenerkennung w{\"a}hrend der endoskopischen Submukosadissektion mittels k{\"u}nstlicher Intelligenz}, series = {Zeitschrift f{\"u}r Gastroenterologie}, volume = {63}, journal = {Zeitschrift f{\"u}r Gastroenterologie}, number = {8}, publisher = {Thieme}, doi = {10.1055/s-0045-1811092}, abstract = {Einleitung: Die endoskopische Submukosadissektion (ESD) ist eine komplexe Technik zur Resektion gastrointestinaler Fr{\"u}hneoplasien. Dabei werden f{\"u}r die verschiedenen Schritte der Intervention spezifische endoskopische Instrumente verwendet. Die pr{\"a}zise und automatische Erkennung und Abgrenzung der verwendeten Instrumente (Injektionsnadeln, elektrochirurgische Messer mit unterschiedlichen Konfigurationen, h{\"a}mostatische Zangen) k{\"o}nnte wertvolle Informationen {\"u}ber den Fortschritt und die Verfahrensmerkmale der ESD liefern und eine automatische standardisierte Berichterstattung erm{\"o}glichen. Ziele: Ziel dieser Studie war die Entwicklung eines KI-Algorithmus zur Erkennung und Delineation von endoskopischen Instrumenten bei der ESD. Methodik: 17 ESD-Videos (9×rektal, 5×{\"o}sophageal, 3×gastrisch) wurden retrospektiv zusammengestellt. Auf 8530 Einzelbilder dieser Videos wurden durch 2 Studienmitarbeiter die folgenden Klassen eingezeichnet: Hakenmesser - Spitze, Hakenmesser - Katheter, Nadelmesser - Spitze und - Katheter, Injektionsnadel -Spitze und - Katheter sowie h{\"a}mostatische Zange - Spitze und - Katheter. Der annotierte Datensatz wurde zum Training eines DeepLabV3+-Deep-Learning-Algorithmus mit ConvNeXt-Backbone zur Erkennung und Abgrenzung der genannten Klassen verwendet. Die Evaluation erfolgte durch 5-fache interne Kreuzvalidierung. Ergebnis: Die Validierung auf Einzelpixelbasis ergab insgesamt einen F1-Score von 0,80, eine Sensitivit{\"a}t von 0,81 und eine Spezifit{\"a}t von 1,00. Es wurden F1-Scores von 1,00, 0,97, 0,80, 0,98, 0,85, 0,97, 0,80, 0,51 bzw. 0,85 f{\"u}r die Klassen Hakenmesser - Katheter und - Spitze, Nadelmesser - Katheter und - Spitze, Injektionsnadel - Katheter und - Spitze, h{\"a}mostatische Zange - Katheter und - Spitze gemessen. Schlussfolgerung: In dieser Studie wurden die wichtigsten endoskopischen Instrumente, die w{\"a}hrend der ESD verwendet werden, mit hoher Genauigkeit erkannt. Die geringere Leistung bei der h{\"a}mostatische Zange - Katheter kann auf die Unterrepr{\"a}sentation dieser Klassen in den Trainingsdaten zur{\"u}ckgef{\"u}hrt werden. Zuk{\"u}nftige Studien werden sich auf die Erweiterung der Instrumentenklassen sowie auf die Ausbalancierung der Trainingsdaten konzentrieren.}, language = {de} } @inproceedings{KlausmannRueckertRauberetal., author = {Klausmann, Leonard and Rueckert, Tobias and Rauber, David and Maerkl, Raphaela and Yildiran, Suemeyye R. and Gutbrod, Max and Palm, Christoph}, title = {DIY challenge blueprint: from organization to technical realization in biomedical image analysis}, series = {Medical Image Computing and Computer Assisted Intervention - MICCAI 2025 ; Proceedings Part XI}, booktitle = {Medical Image Computing and Computer Assisted Intervention - MICCAI 2025 ; Proceedings Part XI}, publisher = {Springer}, address = {Cham}, isbn = {978-3-032-05141-7}, doi = {10.1007/978-3-032-05141-7_9}, pages = {85 -- 95}, abstract = {Biomedical image analysis challenges have become the de facto standard for publishing new datasets and benchmarking different state-of-the-art algorithms. Most challenges use commercial cloud-based platforms, which can limit custom options and involve disadvantages such as reduced data control and increased costs for extended functionalities. In contrast, Do-It-Yourself (DIY) approaches have the capability to emphasize reliability, compliance, and custom features, providing a solid basis for low-cost, custom designs in self-hosted systems. Our approach emphasizes cost efficiency, improved data sovereignty, and strong compliance with regulatory frameworks, such as the GDPR. This paper presents a blueprint for DIY biomedical imaging challenges, designed to provide institutions with greater autonomy over their challenge infrastructure. Our approach comprehensively addresses both organizational and technical dimensions, including key user roles, data management strategies, and secure, efficient workflows. Key technical contributions include a modular, containerized infrastructure based on Docker, integration of open-source identity management, and automated solution evaluation workflows. Practical deployment guidelines are provided to facilitate implementation and operational stability. The feasibility and adaptability of the proposed framework are demonstrated through the MICCAI 2024 PhaKIR challenge with multiple international teams submitting and validating their solutions through our self-hosted platform. This work can be used as a baseline for future self-hosted DIY implementations and our results encourage further studies in the area of biomedical image analysis challenges.}, language = {en} } @unpublished{GutbrodRauberWeberNunesetal., author = {Gutbrod, Max and Rauber, David and Weber Nunes, Danilo and Palm, Christoph}, title = {OpenMIBOOD: Open Medical Imaging Benchmarks for Out-Of-Distribution Detection}, doi = {10.48550/arXiv.2503.16247}, pages = {18}, abstract = {The growing reliance on Artificial Intelligence (AI) in critical domains such as healthcare demands robust mechanisms to ensure the trustworthiness of these systems, especially when faced with unexpected or anomalous inputs. This paper introduces the Open Medical Imaging Benchmarks for Out-Of-Distribution Detection (OpenMIBOOD), a comprehensive framework for evaluating out-of-distribution (OOD) detection methods specifically in medical imaging contexts. OpenMIBOOD includes three benchmarks from diverse medical domains, encompassing 14 datasets divided into covariate-shifted in-distribution, near-OOD, and far-OOD categories. We evaluate 24 post-hoc methods across these benchmarks, providing a standardized reference to advance the development and fair comparison of OOD detection methods. Results reveal that findings from broad-scale OOD benchmarks in natural image domains do not translate to medical applications, underscoring the critical need for such benchmarks in the medical field. By mitigating the risk of exposing AI models to inputs outside their training distribution, OpenMIBOOD aims to support the advancement of reliable and trustworthy AI systems in healthcare. The repository is available at this https URL.}, language = {en} } @misc{ScheppachMendelProbstetal., author = {Scheppach, Markus W. and Mendel, Robert and Probst, Andreas and Rauber, David and R{\"u}ckert, Tobias and Meinikheim, Michael and Palm, Christoph and Messmann, Helmut and Ebigbo, Alanna}, title = {Real-time detection and delineation of tissue during third-space endoscopy using artificial intelligence (AI)}, series = {Endoscopy}, volume = {55}, journal = {Endoscopy}, number = {S02}, publisher = {Thieme}, doi = {10.1055/s-0043-1765128}, pages = {S53 -- S54}, abstract = {Aims AI has proven great potential in assisting endoscopists in diagnostics, however its role in therapeutic endoscopy remains unclear. Endoscopic submucosal dissection (ESD) is a technically demanding intervention with a slow learning curve and relevant risks like bleeding and perforation. Therefore, we aimed to develop an algorithm for the real-time detection and delineation of relevant structures during third-space endoscopy. Methods 5470 still images from 59 full length videos (47 ESD, 12 POEM) were annotated. 179681 additional unlabeled images were added to the training dataset. Consequently, a DeepLabv3+ neural network architecture was trained with the ECMT semi-supervised algorithm (under review elsewhere). Evaluation of vessel detection was performed on a dataset of 101 standardized video clips from 15 separate third-space endoscopy videos with 200 predefined blood vessels. Results Internal validation yielded an overall mean Dice score of 85\% (68\% for blood vessels, 86\% for submucosal layer, 88\% for muscle layer). On the video test data, the overall vessel detection rate (VDR) was 94\% (96\% for ESD, 74\% for POEM). The median overall vessel detection time (VDT) was 0.32 sec (0.3 sec for ESD, 0.62 sec for POEM). Conclusions Evaluation of the developed algorithm on a video test dataset showed high VDR and quick VDT, especially for ESD. Further research will focus on a possible clinical benefit of the AI application for VDR and VDT during third-space endoscopy.}, subject = {Speiser{\"o}hrenkrankheit}, language = {en} } @article{SouzaJrPassosSantanaetal., author = {Souza Jr., Luis Antonio de and Passos, Leandro A. and Santana, Marcos Cleison S. and Mendel, Robert and Rauber, David and Ebigbo, Alanna and Probst, Andreas and Messmann, Helmut and Papa, Jo{\~a}o Paulo and Palm, Christoph}, title = {Layer-selective deep representation to improve esophageal cancer classification}, series = {Medical \& Biological Engineering \& Computing}, volume = {62}, journal = {Medical \& Biological Engineering \& Computing}, publisher = {Springer Nature}, address = {Heidelberg}, doi = {10.1007/s11517-024-03142-8}, pages = {3355 -- 3372}, abstract = {Even though artificial intelligence and machine learning have demonstrated remarkable performances in medical image computing, their accountability and transparency level must be improved to transfer this success into clinical practice. The reliability of machine learning decisions must be explained and interpreted, especially for supporting the medical diagnosis.For this task, the deep learning techniques' black-box nature must somehow be lightened up to clarify its promising results. Hence, we aim to investigate the impact of the ResNet-50 deep convolutional design for Barrett's esophagus and adenocarcinoma classification. For such a task, and aiming at proposing a two-step learning technique, the output of each convolutional layer that composes the ResNet-50 architecture was trained and classified for further definition of layers that would provide more impact in the architecture. We showed that local information and high-dimensional features are essential to improve the classification for our task. Besides, we observed a significant improvement when the most discriminative layers expressed more impact in the training and classification of ResNet-50 for Barrett's esophagus and adenocarcinoma classification, demonstrating that both human knowledge and computational processing may influence the correct learning of such a problem.}, language = {en} } @inproceedings{WeberNunesRauberPalm, author = {Weber Nunes, Danilo and Rauber, David and Palm, Christoph}, title = {Self-supervised 3D Vision Transformer Pre-training for Robust Brain Tumor Classification}, series = {Bildverarbeitung f{\"u}r die Medizin 2025: Proceedings, German Conference on Medical Image Computing, Regensburg March 09-11, 2025}, booktitle = {Bildverarbeitung f{\"u}r die Medizin 2025: Proceedings, German Conference on Medical Image Computing, Regensburg March 09-11, 2025}, editor = {Palm, Christoph and Breininger, Katharina and Deserno, Thomas M. and Handels, Heinz and Maier, Andreas and Maier-Hein, Klaus H. and Tolxdorff, Thomas}, publisher = {Springer Vieweg}, address = {Wiesbaden}, doi = {10.1007/978-3-658-47422-5_69}, pages = {298 -- 303}, abstract = {Brain tumors pose significant challenges in neurology, making precise classification crucial for prognosis and treatment planning. This work investigates the effectiveness of a self-supervised learning approach-masked autoencoding (MAE)-to pre-train a vision transformer (ViT) model for brain tumor classification. Our method uses non-domain specific data, leveraging the ADNI and OASIS-3 MRI datasets, which primarily focus on degenerative diseases, for pretraining. The model is subsequently fine-tuned and evaluated on the BraTS glioma and meningioma datasets, representing a novel use of these datasets for tumor classification. The pre-trained MAE ViT model achieves an average F1 score of 0.91 in a 5-fold cross-validation setting, outperforming the nnU-Net encoder trained from scratch, particularly under limited data conditions. These findings highlight the potential of self-supervised MAE in enhancing brain tumor classification accuracy, even with restricted labeled data.}, language = {en} } @inproceedings{GutbrodGeislerRauberetal., author = {Gutbrod, Max and Geisler, Benedikt and Rauber, David and Palm, Christoph}, title = {Data Augmentation for Images of Chronic Foot Wounds}, series = {Bildverarbeitung f{\"u}r die Medizin 2024: Proceedings, German Workshop on Medical Image Computing, March 10-12, 2024, Erlangen}, booktitle = {Bildverarbeitung f{\"u}r die Medizin 2024: Proceedings, German Workshop on Medical Image Computing, March 10-12, 2024, Erlangen}, editor = {Maier, Andreas and Deserno, Thomas M. and Handels, Heinz and Maier-Hein, Klaus H. and Palm, Christoph and Tolxdorff, Thomas}, publisher = {Springer}, address = {Wiesbaden}, doi = {10.1007/978-3-658-44037-4_71}, pages = {261 -- 266}, abstract = {Training data for Neural Networks is often scarce in the medical domain, which often results in models that struggle to generalize and consequently showpoor performance on unseen datasets. Generally, adding augmentation methods to the training pipeline considerably enhances a model's performance. Using the dataset of the Foot Ulcer Segmentation Challenge, we analyze two additional augmentation methods in the domain of chronic foot wounds - local warping of wound edges along with projection and blurring of shapes inside wounds. Our experiments show that improvements in the Dice similarity coefficient and Normalized Surface Distance metrics depend on a sensible selection of those augmentation methods.}, language = {en} } @inproceedings{GutbrodRauberWeberNunesetal., author = {Gutbrod, Max and Rauber, David and Weber Nunes, Danilo and Palm, Christoph}, title = {OpenMIBOOD: Open Medical Imaging Benchmarks for Out-Of-Distribution Detection}, series = {2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 10.-17. June 2025, Nashville}, booktitle = {2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 10.-17. June 2025, Nashville}, publisher = {IEEE}, isbn = {979-8-3315-4364-8}, doi = {10.1109/CVPR52734.2025.02410}, pages = {25874 -- 25886}, abstract = {The growing reliance on Artificial Intelligence (AI) in critical domains such as healthcare demands robust mechanisms to ensure the trustworthiness of these systems, especially when faced with unexpected or anomalous inputs. This paper introduces the Open Medical Imaging Benchmarks for Out-Of-Distribution Detection (OpenMIBOOD), a comprehensive framework for evaluating out-of-distribution (OOD) detection methods specifically in medical imaging contexts. OpenMIBOOD includes three benchmarks from diverse medical domains, encompassing 14 datasets divided into covariate-shifted in-distribution, nearOOD, and far-OOD categories. We evaluate 24 post-hoc methods across these benchmarks, providing a standardized reference to advance the development and fair comparison of OODdetection methods. Results reveal that findings from broad-scale OOD benchmarks in natural image domains do not translate to medical applications, underscoring the critical need for such benchmarks in the medical field. By mitigating the risk of exposing AI models to inputs outside their training distribution, OpenMIBOOD aims to support the advancement of reliable and trustworthy AI systems in healthcare. The repository is available at https://github.com/remic-othr/OpenMIBOOD.}, language = {en} } @unpublished{RueckertRauberMaerkletal., author = {R{\"u}ckert, Tobias and Rauber, David and Maerkl, Raphaela and Klausmann, Leonard and Yildiran, Suemeyye R. and Gutbrod, Max and Nunes, Danilo Weber and Moreno, Alvaro Fernandez and Luengo, Imanol and Stoyanov, Danail and Toussaint, Nicolas and Cho, Enki and Kim, Hyeon Bae and Choo, Oh Sung and Kim, Ka Young and Kim, Seong Tae and Arantes, Gon{\c{c}}alo and Song, Kehan and Zhu, Jianjun and Xiong, Junchen and Lin, Tingyi and Kikuchi, Shunsuke and Matsuzaki, Hiroki and Kouno, Atsushi and Manesco, Jo{\~a}o Renato Ribeiro and Papa, Jo{\~a}o Paulo and Choi, Tae-Min and Jeong, Tae Kyeong and Park, Juyoun and Alabi, Oluwatosin and Wei, Meng and Vercauteren, Tom and Wu, Runzhi and Xu, Mengya and an Wang, and Bai, Long and Ren, Hongliang and Yamlahi, Amine and Hennighausen, Jakob and Maier-Hein, Lena and Kondo, Satoshi and Kasai, Satoshi and Hirasawa, Kousuke and Yang, Shu and Wang, Yihui and Chen, Hao and Rodr{\´i}guez, Santiago and Aparicio, Nicol{\´a}s and Manrique, Leonardo and Lyons, Juan Camilo and Hosie, Olivia and Ayobi, Nicol{\´a}s and Arbel{\´a}ez, Pablo and Li, Yiping and Khalil, Yasmina Al and Nasirihaghighi, Sahar and Speidel, Stefanie and R{\"u}ckert, Daniel and Feussner, Hubertus and Wilhelm, Dirk and Palm, Christoph}, title = {Comparative validation of surgical phase recognition, instrument keypoint estimation, and instrument instance segmentation in endoscopy: Results of the PhaKIR 2024 challenge}, pages = {36}, abstract = {Reliable recognition and localization of surgical instruments in endoscopic video recordings are foundational for a wide range of applications in computer- and robot-assisted minimally invasive surgery (RAMIS), including surgical training, skill assessment, and autonomous assistance. However, robust performance under real-world conditions remains a significant challenge. Incorporating surgical context - such as the current procedural phase - has emerged as a promising strategy to improve robustness and interpretability. To address these challenges, we organized the Surgical Procedure Phase, Keypoint, and Instrument Recognition (PhaKIR) sub-challenge as part of the Endoscopic Vision (EndoVis) challenge at MICCAI 2024. We introduced a novel, multi-center dataset comprising thirteen full-length laparoscopic cholecystectomy videos collected from three distinct medical institutions, with unified annotations for three interrelated tasks: surgical phase recognition, instrument keypoint estimation, and instrument instance segmentation. Unlike existing datasets, ours enables joint investigation of instrument localization and procedural context within the same data while supporting the integration of temporal information across entire procedures. We report results and findings in accordance with the BIAS guidelines for biomedical image analysis challenges. The PhaKIR sub-challenge advances the field by providing a unique benchmark for developing temporally aware, context-driven methods in RAMIS and offers a high-quality resource to support future research in surgical scene understanding.}, language = {en} } @misc{ScheppachMendelMuzalyovaetal., author = {Scheppach, Markus W. and Mendel, Robert and Muzalyova, Anna and Rauber, David and Probst, Andreas and Nagl, Sandra and R{\"o}mmele, Christoph and Yip, Hon Chi and Lau, Louis Ho Shing and G{\"o}lder, Stefan Karl and Schmidt, Arthur and Kouladouros, Konstantinos and Abdelhafez, Mohamed and Walter, B. and Meinikheim, Michael and Chiu, Philip Wai Yan and Palm, Christoph and Messmann, Helmut and Ebigbo, Alanna}, title = {K{\"u}nstliche Intelligenz erh{\"o}ht die Gef{\"a}ßerkennung von Endoskopikern bei third space Endoskopie}, series = {Zeitschrift f{\"u}r Gastroenterologie}, volume = {62}, journal = {Zeitschrift f{\"u}r Gastroenterologie}, number = {09}, publisher = {Georg Thieme Verlag KG}, doi = {10.1055/s-0044-1790087}, pages = {e830}, abstract = {Einleitung: K{\"u}nstliche Intelligenz (KI)-Algorithmen unterst{\"u}tzen Endoskopiker bei der Erkennung und Charakterisierung von Kolonpolypen in der klinischen Praxis und f{\"u}hren zu einer Erh{\"o}hung der Adenomdetektionsrate. Auch bei therapeutischen Maßnahmen wie der endoskopischen Submukosadissektion (ESD) k{\"o}nne relevante anatomische Strukturen durch KI mit hoher Genauigkeit erkannt und im endoskopischen Bild in Echtzeit markiert werden. Der Effekt einer solchen Applikation auf die Gef{\"a}ßdetektion von Endoskopikern ist bislang nicht erforscht. Ziele: In dieser Studie wurde der Effekt eines KI-Algorithmus zur Echtzeit-Gef{\"a}ßmarkierung bei ESD auf die Gef{\"a}ßdetektionsrate von Endoskopikern untersucht. Methodik: 59 third space Endoskopievideos wurde aus der Datenbank des Universit{\"a}tsklinikums Augsburg extrahiert. Auf 5470 Einzelbildern dieser Untersuchungen wurde submukosale Blutgef{\"a}ße annotiert. Zusammen mit weiteren 179681 unmarkierten Bildern wurde ein DeepLabV3+ neuronales Netzwerk mit einer semi-supervised learning Methode darin trainiert, submukosale Blutgef{\"a}ße auf dem endoskopischen Bild zu erkennen und in Echtzeit einzuzeichnen. Anhand eines Videotests mit 101 Videoclips und 200 vordefinierten Blutgef{\"a}ßen wurden 19 Endoskopiker mit und ohne KI Unterst{\"u}tzung getestet. Ergebnis: Der Algorithmus erkannte in dem Videotest 93.5\% der Gef{\"a}ße in einer Detektionszeit von im Median 0,3 Sekunden. Die Gef{\"a}ßdetektionsrate von Endoskopikern erh{\"o}hte sich durch KI Unterst{\"u}tzung von 56,4\% auf 72,4\% (p<0.001). Die Gef{\"a}ßdetektionszeit reduzierte sich durch KI-Unterst{\"u}tzung von 6,7 auf 5.2 Sekunden (p<0.001). Der Algorithmus zeigte eine Rate an falsch positiven Detektionen in 4.5\% der Einzelbilder. Falsch positiv erkannte Strukturen wurde k{\"u}rzer detektiert, als richtig positive (0.7 und 6.0 Sekunden, p<0.001). Schlussfolgerung: KI Unterst{\"u}tzung f{\"u}hrte zu einer erh{\"o}hten Gef{\"a}ßdetektionsrate und schnelleren Gef{\"a}ßdetektionszeit von Endoskopikern. Ein m{\"o}glicher klinischer Effekt auf die intraprozedurale Komplikationsrate oder Operationszeit k{\"o}nnte in prospektiven Studien ermittelt werden.}, language = {de} } @misc{ScheppachNunesArizietal., author = {Scheppach, Markus W. and Nunes, Danilo Weber and Arizi, X. and Rauber, David and Probst, Andreas and Nagl, Sandra and R{\"o}mmele, Christoph and Palm, Christoph and Messmann, Helmut and Ebigbo, Alanna}, title = {Intraoperative Phasenerkennung bei endoskopischer Submukosadissektion mit Hilfe von k{\"u}nstlicher Intelligenz}, series = {Zeitschrift f{\"u}r Gastroenterologie}, volume = {62}, journal = {Zeitschrift f{\"u}r Gastroenterologie}, number = {09}, publisher = {Georg Thieme Verlag KG}, doi = {10.1055/s-0044-1790084}, pages = {e828}, abstract = {Einleitung: K{\"u}nstliche Intelligenz (KI) wird in der Endoskopie des Gastrointestinaltraktes zur Erkennung und Charakterisierung von Kolonpolypen eingesetzt. Die Rolle von KI bei therapeutischen Maßnahmen wurde noch nicht eingehend untersucht. Eine intraprozedurale Phasenerkennung bei endoskopischer Submukoasdissektion (ESD) k{\"o}nnte die Erhebung von Qualit{\"a}tsindikatoren erm{\"o}glichen. Weiterhin k{\"o}nnte diese Technologie zu einem tieferen Verst{\"a}ndnis {\"u}ber die Eigenschaften der Prozedur f{\"u}hren und weiterf{\"u}hrende Applikationen zur automatischen Dokumentation oder standardisiertem Training vorbereiten. Ziele: Ziel dieser Studie war die Entwicklung eines KI Algorithmus zur intraprozeduralen Phasenerkennung bei endoskopischer Submukosadissektion. Methodik: 2071546 Einzelbilder aus 27 ESD Videos in voller L{\"a}nge wurden f{\"u}r die {\"u}bergeordneten Klassen Diagnostik, Markierung, Nadelinjektion, Dissektion und Blutung, sowie die untergeordneten Klassen Endoskop-Manipulation, Injektion und Applikation von elektrischem Strom annotiert. Mit einem Trainingsdatensatz (898440 Einzelbilder, 17 ESDs) wurde ein Video Swin Transformer mit uniformer Stichprobenentnahme trainiert und intern validiert (769523 Einzelbilder, 6 ESDs). Neben der internen Validierung wurde der Algorithmus anhand von einem separaten Testdatensatz (403583 Einzelbilder, 4 ESDs) evaluiert. Ergebnis: Der F1 Score des Algorithmus f{\"u}r alle Klassen lag in der internen Validierung bei 83\%, in dem separaten Test bei 90\%. Anhand des separaten Tests wurden true positive (TP)-Raten f{\"u}r Diagnostik, Markierung, Nadelinjektion, Dissektion und Blutung von 100\%, 100\%, 96\%, 97\% und 93\% ermittelt. F{\"u}r Endoskopmanipulation, Injektion und Applikation von Elektrizit{\"a}t lagen die TP-Raten bei 92\%, 98\% und 91\%. Schlussfolgerung: Der entwickelte Algorithmus klassifizierte ESD Videos in voller L{\"a}nge und anhand jedes einzelnen Bildes mit hoher Genauigkeit. Zuk{\"u}nftige Forschungsvorhaben k{\"o}nnten intraoperative Qualit{\"a}tsindikatioren auf Basis dieser Informationen entwickeln und eine automatisierte Dokumentation erm{\"o}glichen.}, language = {de} } @misc{ZellmerRauberProbstetal., author = {Zellmer, Stephan and Rauber, David and Probst, Andreas and Weber, Tobias and Nagl, Sandra and R{\"o}mmele, Christoph and Schnoy, Elisabeth and Palm, Christoph and Messmann, Helmut and Ebigbo, Alanna}, title = {Verwendung k{\"u}nstlicher Intelligenz bei der Detektion der Papilla duodeni major}, series = {Zeitschrift f{\"u}r Gastroenterologie}, volume = {61}, journal = {Zeitschrift f{\"u}r Gastroenterologie}, number = {08}, publisher = {Thieme}, address = {Stuttgart}, doi = {10.1055/s-0043-1772000}, pages = {e539 -- e540}, abstract = {Einleitung Die Endoskopische Retrograde Cholangiopankreatikographie (ERCP) ist der Goldstandard in der Diagnostik und Therapie von Erkrankungen des pankreatobili{\"a}ren Trakts. Jedoch ist sie technisch sehr anspruchsvoll und weist eine vergleichsweise hohe Komplikationsrate auf. Ziele In der vorliegenden Machbarkeitsstudie soll gepr{\"u}ft werden, ob mithilfe eines Deep-learning-Algorithmus die Papille und das Ostium zuverl{\"a}ssig detektiert werden k{\"o}nnen und somit f{\"u}r Endoskopiker mit geringer Erfahrung ein geeignetes Hilfsmittel, insbesondere f{\"u}r die Ausbildungssituation, darstellen k{\"o}nnten. Methodik Wir betrachteten insgesamt 606 Bilddatens{\"a}tze von 65 Patienten. In diesen wurde sowohl die Papilla duodeni major als auch das Ostium segmentiert. Anschließend wurde eine neuronales Netz mittels eines Deep-learning-Algorithmus trainiert. Außerdem erfolgte eine 5-fache Kreuzvaldierung. Ergebnisse Bei einer 5-fachen Kreuzvaldierung auf den 606 gelabelten Daten konnte f{\"u}r die Klasse Papille eine F1-Wert von 0,7908, eine Sensitivit{\"a}t von 0,7943 und eine Spezifit{\"a}t von 0,9785 erreicht werden, f{\"u}r die Klasse Ostium eine F1-Wert von 0,5538, eine Sensitivit{\"a}t von 0,5094 und eine Spezifit{\"a}t von 0,9970 (vgl. [Tab. 1]). Unabh{\"a}ngig von der Klasse zeigte sich gemittelt (Klasse Papille und Klasse Ostium) ein F1-Wert von 0,6673, eine Sensitivit{\"a}t von 0,6519 und eine Spezifit{\"a}t von 0,9877 (vgl. [Tab. 2]). Schlussfolgerung In vorliegende Machbarkeitsstudie konnte das neuronale Netz die Papilla duodeni major mit einer hohen Sensitivit{\"a}t und sehr hohen Spezifit{\"a}t identifizieren. Bei der Detektion des Ostiums war die Sensitivit{\"a}t deutlich geringer. Zuk{\"u}nftig soll das das neuronale Netz mit mehr Daten trainiert werden. Außerdem ist geplant, den Algorithmus auch auf Videos anzuwenden. Somit k{\"o}nnte langfristig ein geeignetes Hilfsmittel f{\"u}r die ERCP etabliert werden.}, language = {de} } @article{RueckertRauberMaerkletal., author = {Rueckert, Tobias and Rauber, David and Maerkl, Raphaela and Klausmann, Leonard and Yildiran, Suemeyye R. and Gutbrod, Max and Nunes, Danilo Weber and Moreno, Alvaro Fernandez and Luengo, Imanol and Stoyanov, Danail and Toussaint, Nicolas and Cho, Enki and Kim, Hyeon Bae and Choo, Oh Sung and Kim, Ka Young and Kim, Seong Tae and Arantes, Gon{\c{c}}alo and Song, Kehan and Zhu, Jianjun and Xiong, Junchen and Lin, Tingyi and Kikuchi, Shunsuke and Matsuzaki, Hiroki and Kouno, Atsushi and Manesco, Jo{\~a}o Renato Ribeiro and Papa, Jo{\~a}o Paulo and Choi, Tae-Min and Jeong, Tae Kyeong and Park, Juyoun and Alabi, Oluwatosin and Wei, Meng and Vercauteren, Tom and Wu, Runzhi and Xu, Mengya and Wang, An and Bai, Long and Ren, Hongliang and Yamlahi, Amine and Hennighausen, Jakob and Maier-Hein, Lena and Kondo, Satoshi and Kasai, Satoshi and Hirasawa, Kousuke and Yang, Shu and Wang, Yihui and Chen, Hao and Rodr{\´i}guez, Santiago and Aparicio, Nicol{\´a}s and Manrique, Leonardo and Palm, Christoph and Wilhelm, Dirk and Feussner, Hubertus and Rueckert, Daniel and Speidel, Stefanie and Nasirihaghighi, Sahar and Al Khalil, Yasmina and Li, Yiping and Arbel{\´a}ez, Pablo and Ayobi, Nicol{\´a}s and Hosie, Olivia and Lyons, Juan Camilo}, title = {Comparative validation of surgical phase recognition, instrument keypoint estimation, and instrument instance segmentation in endoscopy: Results of the PhaKIR 2024 challenge}, series = {Medical Image Analysis}, volume = {109}, journal = {Medical Image Analysis}, publisher = {Elsevier}, issn = {1361-8415}, doi = {10.1016/j.media.2026.103945}, pages = {31}, abstract = {Reliable recognition and localization of surgical instruments in endoscopic video recordings are foundational for a wide range of applications in computer- and robot-assisted minimally invasive surgery (RAMIS), including surgical training, skill assessment, and autonomous assistance. However, robust performance under real-world conditions remains a significant challenge. Incorporating surgical context - such as the current procedural phase - has emerged as a promising strategy to improve robustness and interpretability. To address these challenges, we organized the Surgical Procedure Phase, Keypoint, and Instrument Recognition (PhaKIR) sub-challenge as part of the Endoscopic Vision (EndoVis) challenge at MICCAI 2024. We introduced a novel, multi-center dataset comprising thirteen full-length laparoscopic cholecystectomy videos collected from three distinct medical institutions, with unified annotations for three interrelated tasks: surgical phase recognition, instrument keypoint estimation, and instrument instance segmentation. Unlike existing datasets, ours enables joint investigation of instrument localization and procedural context within the same data while supporting the integration of temporal information across entire procedures. We report results and findings in accordance with the BIAS guidelines for biomedical image analysis challenges. The PhaKIR sub-challenge advances the field by providing a unique benchmark for developing temporally aware, context-driven methods in RAMIS and offers a high-quality resource to support future research in surgical scene understanding.}, language = {en} } @misc{RueckertRauberKlausmannetal., author = {Rueckert, Tobias and Rauber, David and Klausmann, Leonard and Gutbrod, Max and Rueckert, Daniel and Feussner, Hubertus and Wilhelm, Dirk and Palm, Christoph}, title = {PhaKIR Dataset - Surgical Procedure Phase, Keypoint, and Instrument Recognition [Data set]}, doi = {10.5281/zenodo.15740620}, abstract = {Note: A script for extracting the individual frames from the video files while preserving the challenge-compliant directory structure and frame-to-mask naming conventions is available on GitHub and can be accessed here: https://github.com/remic-othr/PhaKIR_Dataset. The dataset is described in the following publications: Rueckert, Tobias et al.: Comparative validation of surgical phase recognition, instrument keypoint estimation, and instrument instance segmentation in endoscopy: Results of the PhaKIR 2024 challenge. arXiv preprint, https://arxiv.org/abs/2507.16559. 2025. Rueckert, Tobias et al.: Video Dataset for Surgical Phase, Keypoint, and Instrument Recognition in Laparoscopic Surgery (PhaKIR). arXiv preprint, https://arxiv.org/abs/2511.06549. 2025. The proposed dataset was used as the training dataset in the PhaKIR challenge (https://phakir.re-mic.de/) as part of EndoVis-2024 at MICCAI 2024 and consists of eight real-world videos of human cholecystectomies ranging from 23 to 60 minutes in duration. The procedures were performed by experienced physicians, and the videos were recorded in three hospitals. In addition to existing datasets, our annotations provide pixel-wise instance segmentation masks of surgical instruments for a total of 19 categories, coordinates of relevant instrument keypoints (instrument tip(s), shaft-tip transition, shaft), both at an interval of one frame per second, and specifications regarding the intervention phases for a total of eight different phase categories for each individual frame in one dataset and thus comprehensively cover instrument localization and the context of the operation. Furthermore, the provision of the complete video sequences offers the opportunity to include the temporal information regarding the respective tasks and thus further optimize the resulting methods and outcomes.}, language = {en} } @misc{GutbrodRauberWeberNunesetal., author = {Gutbrod, Max and Rauber, David and Weber Nunes, Danilo and Palm, Christoph}, title = {A cleaned subset of the first five CATARACTS test videos [Data set]}, doi = {10.5281/zenodo.14924735}, abstract = {This dataset is a subset of the original CATARACTS test dataset and is used by the OpenMIBOOD framework to evaluate a specific out-of-distribution setting. When using this dataset, it is mandatory to cite the corresponding publication (OpenMIBOOD (10.1109/CVPR52734.2025.02410)) and follow the acknowledgement and citation requirements of the original dataset (CATARACTS). The original CATARACTS dataset (associated publication,Homepage) consists of 50 videos of cataract surgeries, split into 25 train and 25 test videos. This subset contains the frames of the first 5 test videos. Further, black frames at the beginning of each video were removed.}, language = {en} } @misc{GutbrodRauberWeberNunesetal., author = {Gutbrod, Max and Rauber, David and Weber Nunes, Danilo and Palm, Christoph}, title = {Cropped single instrument frames subset from Cholec80 [Data set]}, doi = {10.5281/zenodo.14921670}, abstract = {This dataset is a subset of the original Cholec80 dataset and is used by the OpenMIBOOD framework to evaluate a specific out-of-distribution setting. When using this dataset, it is mandatory to cite the corresponding publication (OpenMIBOOD) and to follow the acknowledgement and citation requirements of the original dataset (Cholec80). The original Cholec80 dataset (associated paper,Homepage) consists of 80 cholecystectomy surgery videos recorded at 25 fps, performed by 13 surgeons. It includes phase annotations (25 fps) and tool presence labels (1 fps), with phase definitions provided by a senior surgeon. A tool is considered present if at least half of its tip is visible. The dataset categorizes tools into seven types: Grasper, Bipolar, Hook, Scissors, Clipper, Irrigator, and Specimen bag. Multiple tools may be present in each frame. Additionally, 76 of the 80 videos exhibit a strong black vignette. For this dataset subset, frames were extracted based on tool presence labels, selecting only those containing Grasper, Bipolar, Hook, or Clipper while ensuring that only a single tool appears per frame. To enhance visual consistency, the black vignette was removed by extracting an inner rectangular region, where applicable.}, language = {en} } @misc{GutbrodRauberWeberNunesetal., author = {Gutbrod, Max and Rauber, David and Weber Nunes, Danilo and Palm, Christoph}, title = {OpenMIBOOD's classification models for the MIDOG, PhaKIR, and OASIS-3 benchmarks [Data set]}, doi = {10.5281/zenodo.14982267}, abstract = {These models are provided for evaluating post-hoc out-of-distribution methods on the three OpenMIBOOD benchmarks: MIDOG, PhaKIR, and OASIS-3. When using these models, make sure to give appropriate credit and cite the OpenMIBOOD publication.}, language = {en} } @misc{KlausmannRueckertRauberetal., author = {Klausmann, Leonard and Rueckert, Tobias and Rauber, David and Maerkl, Raphaela and Yildiran, Suemeyye R. and Gutbrod, Max and Palm, Christoph}, title = {Abstract: DIY Challenge Blueprint}, series = {Bildverarbeitung f{\"u}r die Medizin 2025: Proceedings, German Conference on Medical Image Computing, L{\"u}beck March 15-17, 2026}, journal = {Bildverarbeitung f{\"u}r die Medizin 2025: Proceedings, German Conference on Medical Image Computing, L{\"u}beck March 15-17, 2026}, editor = {Handels, Heinz and Breininger, Katharina and Deserno, Thomas M. and Maier, Andreas and Maier-Hein, Klaus H. and Palm, Christoph and Tolxdorff, Thomas}, publisher = {Springer Vieweg}, address = {Wiesbaden}, doi = {10.1007/978-3-658-51100-5_27}, pages = {131 -- 131}, abstract = {The high cost of challenge platforms prevents many people from organizing their own competitions. The do-it-yourself (DIY) challenge blueprint [1] allows you to host your own biomedical AI benchmark challenge. Our DIY approach circumvents the current constraints of commercial challenge platforms. A sovereign, extensible and cost-efficient deployment is provided via containerised, identity-managed and reproducible pipelines. Focus lies on GDPR-compliant hosting via infrastructure-as-code, automated evaluation, modular orchestration, and role-based identity and access management. The framework integrates Docker-based execution and standardised interfaces for task definitions, dataset curation and evaluation. All in all it is designed to be flexible and modular, as demonstrated in the MICCAI 2024 PhaKIR challenge [2, 3]. In this case study, different medical tasks on a multicentre laparoscopic dataset with framewise labels for phases and spatial annotations for instruments across fulllength videos were supported. This case study empirically validates the DIY challenge blueprint as a reproducible and customizable challenge-hosting infrastructure. The full code can be found at https://github.com/remic-othr/PhaKIR_DIY.}, subject = {Bildverarbeitung}, language = {en} } @inproceedings{GutbrodRauberPalm, author = {Gutbrod, Max and Rauber, David and Palm, Christoph}, title = {Improving Generalization in Mitotic Cell Detection via Domain Transformations}, series = {Bildverarbeitung f{\"u}r die Medizin 2025: Proceedings, German Conference on Medical Image Computing, L{\"u}beck March 15-17, 2026}, booktitle = {Bildverarbeitung f{\"u}r die Medizin 2025: Proceedings, German Conference on Medical Image Computing, L{\"u}beck March 15-17, 2026}, editor = {Handels, Heinz and Breininger, Katharina and Deserno, Thomas M. and Maier, Andreas and Maier-Hein, Klaus H. and Palm, Christoph and Tolxdorff, Thomas}, publisher = {Springer Vieweg}, address = {Wiesbaden}, doi = {10.1007/978-3-658-51100-5_71}, pages = {362 -- 367}, abstract = {We address domain generalization (DG) in mitotic-cell (MC) detection by combining a β-variational autoencoder (VAE) for domain transformations with feature-space alignment together with an object detector. The β-VAE synthesizes domain-transformed images, and the detector is trained to map originals and their transformed counterparts to equal representations. On the MIDOG++ dataset, this approach improves out-of-domain detection F1 scores by 7 and 3 percentage points compared to the color-variation augmentation and stain-normalization baselines. Results further suggest that morphology shifts hinder generalization more than stain shifts.}, subject = {K{\"u}nstliche Intelligenz}, language = {en} }