Refine
Year of publication
Document Type
- Article (1828)
- conference proceeding (article) (1681)
- Part of a Book (567)
- conference proceeding (presentation, abstract) (267)
- Book (161)
- conference talk (137)
- Report (122)
- Review (100)
- Preprint (49)
- Working Paper (49)
- conference proceeding (volume) (34)
- Doctoral Thesis (25)
- Edited collection (21)
- Part of Periodical (20)
- Master's Thesis (15)
- Lecture (12)
- Bachelor-/Diplom Thesis (11)
- Patent (10)
- Other (9)
- Moving Images (7)
- Study Thesis (4)
- Sound (3)
- Course Material (2)
- Habilitation (2)
Language
Is part of the Bibliography
- no (5136)
Keywords
- Simulation (37)
- Digitalisierung (30)
- Erneuerbare Energien (28)
- Energiewende (27)
- field emission (24)
- Bildgebendes Verfahren (22)
- Deep Learning (22)
- Deutschland (22)
- Klimaneutralität (22)
- Power-to-Gas (21)
Institute
- Fakultät Angewandte Sozial- und Gesundheitswissenschaften (1295)
- Fakultät Informatik und Mathematik (1137)
- Fakultät Maschinenbau (899)
- Fakultät Elektro- und Informationstechnik (800)
- Institut für Sozialforschung und Technikfolgenabschätzung (IST) (553)
- Labor Empirische Sozialforschung (458)
- Fakultät Angewandte Natur- und Kulturwissenschaften (403)
- Fakultät Bauingenieurwesen (354)
- Labor für Technikfolgenabschätzung und Angewandte Ethik (LaTe) (311)
- Laboratory for Safe and Secure Systems (LAS3) (277)
Begutachtungsstatus
- peer-reviewed (1566)
- begutachtet (93)
As global demand for green hydrogen rises, potential hydrogen exporters move into the spotlight. However, the large-scale installation of on-grid hydrogen electrolysis for export can have profound impacts on domestic energy prices and energy-related emissions. Our investigation explores the interplay of hydrogen exports, domestic energy transition and temporal hydrogen regulation, employing a sector-coupled energy model in Morocco. We find substantial co-benets of domestic climate change mitigation and hydrogen exports, whereby exports can reduce domestic electricity prices while mitigation reduces hydrogen export prices. However, increasing hydrogen exports quickly in a system that is still dominated by fossil fuels can substantially raise domestic electricity prices, if green hydrogen production is not regulated. Surprisingly, temporal matching of hydrogen production lowers domestic electricity cost by up to 31% while the effect on exporters is minimal. This policy instrument can steer the welfare (re-)distribution between hydrogen exporting firms, hydrogen importers, and domestic electricity consumers and hereby increases acceptance among actors.
A novel method for controlling the rebound behavior of small balls made of Al2O3 with a radius of 2.381 mm is presented. It uses different types of micro-structured surfaces of soft magnetoactive elastomers. These surfaces were fabricated via laser micromachining and include fully ablated surfaces as well as micrometer-sized lamellas with a fixed width of 90 µm, height of 250 µm and three different gap sizes (15, 60 and 105 µm). The lamellas can change their orientation from edge-on to face-on configuration according to the direction of the external magnetic field from a permanent magnet. The orientation of the external magnetic field significantly influences the rebound behavior of the balls, from a coefficient of restitution e of to < 0.1. The highest relative change in the coefficient of restitution between zero field and face-on configuration of is observed for lamellas with a gap of 60 µm. Other characteristics of the ball rebound such as the penetration depth into an Magnetoactive elastomer and the maximum deceleration are investigated as well. The proposed method does not require a constant power supply due to the use of permanent magnets. It may find novel applications in the field of impact engineering.
With the ongoing miniaturization of wireless devices, the importance of wearable textiles in the antenna segment has increased significantly in recent years. Due to the widespread utilization of wireless body sensor networks for healthcare and ubiquitous applications, the design of wearable antennas offers the possibility of comprehensive monitoring, communication, and energy harvesting and storage. This article reviews a number of properties and benefits to realize comprehensive background information and application ideas for the development of lightweight, compact and low-cost wearable patch antennas. Furthermore, problems and challenges that arise are addressed. Since both electromagnetic and mechanical specifications must be fulfilled, textile and flexible antennas require an appropriate trade-off between materials, antenna topologies, and fabrication methods—depending on the intended application and environmental factors. This overview covers each of the above issues, highlighting research to date while correlating antenna topology, feeding techniques, textile materials, and contacting options for the defined application of wearable planar patch antennas.
ntervention with motivational emails can have a positive effect on course retention in e-learning. It is, however, not yet clear whether different forms of emails affect course retention and how students make progress during the sending of emails. We therefore used a voluntary asynchronous online course with 206 students. Students were randomly divided into four groups: text–picture personalised email vs. text personalised email vs. generalised email vs. no email. Emails were sent weekly for 3 months. Results yield that more students made progress in the text–picture personalised email group than in the control group. Students in all email groups progressed by more units than students in the control group. Only students in email groups completed the course and only students in personalised email groups reacted to the emails. Emails were accepted by most students enrolled. The findings suggest that cost-effective and easily implemented emails can encourage students to progress from unit to unit.
Im Betrieb von Förderbändern auftretende Phänomene sind das Bandwandern in Achsrichtung der Walzen und die Bandverformung in der Bandebene [1, S. 142]. Das Bandwandern ist bisher detailliert untersucht [2–4], während die Bandverformung Gegenstand weniger Veröffentlichungen ist. Deshalb wird in diesem Artikel eine Methode zur Berechnung der sich im dynamischen Betrieb einstellenden Verformung eines elastischen Förderbandes vorgestellt. Dazu wird ein Mehrkörpersimulations-Modell basierend auf der Lumped-Mass-Modellierung verwendet. Als Untersuchungsbeispiel dient ein Zwei-Walzensystem mit flexiblen und zylindrischen Walzen, sowie ein flexibles Förderband. Es zeigt sich, dass mit der Lumped-Mass-Modellierung eine dynamische Bandverformung, die von der gewählten Diskretisierung abhängt, berechnet werden kann. Aufgrund dieser Abhängigkeit ist es notwendig eine Konvergenzanalyse durchzuführen. Zusätzlich ist darauf zu achten, dass die künstliche Anregung aufgrund der Lumped-Mass-Modellierung nicht mit einer Eigenfrequenz des Modells zusammenfällt.
Der Einbau mineralischer, granularer Lasttransferschichten (LTS) zwischen einem Bau-werk und einer Tiefgründung z. B. aus pfahlartigen Traggliedern (PTG) beeinflusst erheb-lich die Lastaufteilung auf die PTG und den zumeist weichen Boden dazwischen. Eine genaue Kenntnis über die Lastaufteilung ist für die Bemessung der LTS und PTG, aber auch des darauf gegründeten Bauwerks erforderlich. Für die Quantifizierung der Lastauf-teilung wurden eine Vielzahl an Berechnungsansätzen entwickelt, welche zumeist auf Modell- oder Großversuchen basieren, aus deren Ergebnissen analytische Ansätze entwi-ckelt wurden. Zunächst werden ausgewählte Berechnungsansätze zur Lastaufteilung vor-gestellt, dann mit Hilfe von Finite-Element-Berechnungen der Einfluss verschiedener Parameter auf die Lastaufteilung veranschaulicht und diskutiert. Für geringe Dicken der LTS ergibt sich dabei eine hohe Übereinstimmung der FE-Berechnungen mit einem Durchstanzmechanismus. Ab einer bestimmten Dicke der LTS, abhängig von ver-schiedensten Randbedingungen, bildet sich in den FE-Berechnungen ein Gewölbe aus. Das Einstanzen der PTG in die LTS kann dann durch einen Grundbruchmechanismus beschrieben und als obere Begrenzung der Lastumlagerung auf die PTG angesehen wer-den. Anhand zahlreicher Vergleichsberechnungen konnte gezeigt werden, dass die beiden Mechanismen Durchstanzen und Grundbruch in sehr vielen Fällen ausreichen, um die Lastaufteilung auch ohne FE-Berechnungen abzuschätzen.
Bei der Untersuchung und Optimierung der Getriebedynamik, im Sinne der Akustik und Lastverteilung, wird oftmals der Drehwegfehler als Bewertungskriterium herangezogen. Zahlreiche Forschungsvorhaben haben sich in diesem Zusammenhang intensiv mit dem Drehwegfehler als parametererregte Schwingungsanregung auseinandergesetzt, um diesen im Hinblick auf das Systemverhalten zu optimieren. Dabei bleibt der Zahnfußbereich, welcher außerhalb der Evolvente liegt, und dessen Einfluss auf den Drehwegfehler oftmals unberücksichtigt.
Dieser Artikel untersucht daher den Einfluss des Zahnfußbereiches auf den statischen Drehwegfehler, indem relevante Größen wie das spezifische Kopfspiel und der Fußrundungsradius an einer typischen und real eingesetzten Stirnradverzahnung variiert werden.
Ferner wird betrachtet, inwiefern der Drehwegfehler durch unterschiedliche Geometriedefinitionen (Fertigungssimulation, Zahngeometrie mit vereinfachter sowie stark vereinfachter Modellierung des Zahnfußbereichs) beeinflusst wird. Die Autoren nutzen hierfür zwei FE-Ansätze (adaptives und nicht-adaptives Netz) und bewerten diese hinsichtlich ihrer Gültigkeit, Modellgüte und Praktikabilität.
Die Ergebnisse der Untersuchungen werden hinsichtlich einer minimal erforderlichen Geometrie des Zahnfußbereiches interpretiert. Für den Praxistransfer wird einerseits der maximale Fehler bei Nichtberücksichtigung des gesamten Zahnfußbereichs beziffert.
Andererseits wird eine Empfehlung zur Anpassung einer vereinfachten Geometrie, an eine Geometrie, ähnlich der aus einer Fertigungssimulation, gegeben. Weiterhin wird der Artikel diskutieren, ob man ein vereinfachtes Zahnmodell zur Berechnung der Getriebedynamik nutzen kann, wenn der Fertigungsprozess noch nicht definiert ist.
Es gibt eine Vielzahl von „Wie-erstelle-ich-ein Finite-Element-Programm?“-Lehrbüchern, aber nur recht wenige Veröffentlichungen zur Frage „Wie wende ich ein Finite-Element-Programm an?“. Dieses Buch legt den Schwerpunkt auf die zweite Fragestellung. Es basiert auf den Vorlesungen zur Anwendung der Finite-Element-Methode, die der Autor seit 1998 an der OTH Regensburg hält. Deren Inhalte kommen aus seiner Tätigkeit als Prüfingenieur für Baustatik in einem großen Münchener Ingenieurbüro. Behandelt werden sowohl Fragestellungen, mit denen sich jeder Ingenieur konfrontiert sieht, wenn er Berechnungen mit einem Finite-Element-Programm erstellen will, als auch Problempunkte, die im Büro des Autors im Zuge einer Projektbearbeitung auftraten und auf den ersten Blick gar nicht so klar waren. Der 2.Teil des zweibändigen Werkes befasst sich mit Themen aus der nichtlinearen Statik.
Generative deep learning approaches for the design of dental restorations: A narrative review
(2024)
Objectives:
This study aims to explore and discuss recent advancements in tooth reconstruction utilizing deep learning (DL) techniques. A review on new DL methodologies in partial and full tooth reconstruction is conducted.
Data/Sources:
PubMed, Google Scholar, and IEEE Xplore databases were searched for articles from 2003 to 2023.
Study selection:
The review includes 9 articles published from 2018 to 2023. The selected articles showcase novel DL approaches for tooth reconstruction, while those concentrating solely on the application or review of DL methods are excluded. The review shows that data is acquired via intraoral scans or laboratory scans of dental plaster models. Common data representations are depth maps, point clouds, and voxelized point clouds. Reconstructions focus on single teeth, using data from adjacent teeth or the entire jaw. Some articles include antagonist teeth data and features like occlusal grooves and gap distance. Primary network architectures include Generative Adversarial Networks (GANs) and Transformers. Compared to conventional digital methods, DL-based tooth reconstruction reports error rates approximately two times lower.
Conclusions:
Generative DL models analyze dental datasets to reconstruct missing teeth by extracting insights into patterns and structures. Through specialized application, these models reconstruct morphologically and functionally sound dental structures, leveraging information from the existing teeth. The reported advancements facilitate the feasibility of DL-based dental crown reconstruction. Beyond GANs and Transformers with point clouds or voxels, recent studies indicate promising outcomes with diffusion-based architectures and innovative data representations like wavelets for 3D shape completion and inference problems.
Clinical significance:
Generative network architectures employed in the analysis and reconstruction of dental structures demonstrate notable proficiency. The enhanced accuracy and efficiency of DL-based frameworks hold the potential to enhance clinical outcomes and increase patient satisfaction. The reduced reconstruction times and diminished requirement for manual intervention may lead to cost savings and improved accessibility of dental services.
The present paper takes a novel approach to production of fibre-reinforced thermoplastic tubes. The method begins with the raw materials, reinforcing fibre and thermoplastic granulate which are processed to tapes through a newly developed direct impregnation process. It is followed by consolidation of fibre-reinforced thermoplastic tubes using infrared (IR) emitters in the filament winding process. This process employs various angles and utilizes a rotatable consolidation axis. The winding process operates at a constant speed, addressing the challenge of bending the fibre-reinforced tapes in the angle reversal areas near the tube ends. Experiments have confirmed that the process can run at speeds reaching approximately 470 mm/min. The design of the impregnation line takes into account the properties of the thermoplastic and the roving, allowing for a speed of up to 1 m/s.
Background:
With the prevalence of burnout among surgeons posing a significant threat to healthcare outcomes, the mental toughness of medical professionals has come to the fore. Mental toughness is pivotal for surgical performance and patient safety, yet research into its dynamics within a global and multi-specialty context remains scarce. This study aims to elucidate the factors contributing to mental toughness among surgeons and to understand how it correlates with surgical outcomes and personal well-being.
Methods:
Utilizing a cross-sectional design, this study surveyed 104 surgeons from English and German-speaking countries using the Mental Toughness Questionnaire (MTQ-18) along with additional queries about their surgical practice and general life satisfaction. Descriptive and inferential statistical analyses were applied to investigate the variations in mental toughness across different surgical domains and its correlation with professional and personal factors.
Results:
The study found a statistically significant higher level of mental toughness in micro-surgeons compared to macro-surgeons and a positive correlation between mental toughness and surgeons' intent to continue their careers. A strong association was also observed between general life satisfaction and mental toughness. No significant correlations were found between the application of psychological skills and mental toughness.
Conclusion:
Mental toughness varies significantly among surgeons from different specialties and is influenced by professional dedication and personal life satisfaction. These findings suggest the need for targeted interventions to foster mental toughness in the surgical community, potentially enhancing surgical performance and reducing burnout. Future research should continue to explore these correlations, with an emphasis on longitudinal data and the development of resilience-building programs.
Digital Twins (DT) implementation in the Built Environment (BE) industry is still in its early stages. Aiming to increase the knowledge about DT, this study analyzes how DT can be understood in the BE sector and investigates its different potential benefits and expected challenges. To do so, the Systematic Literature Review (SLR) approach was employed. Using 228 publications, the current study presents a proposed definition and structure for DT systems. The proposed structure is based on four main layers: physical, digital, application, and user layers. The study also classified the applications of DT into six groups: sustainability and environmental, facility management, safety, health, and risk management, structural performance, construction management, and architectural and urban-related applications. The challenges of DT implementation were also grouped based on industry-related, social and organizational, economic, technological, and political and legal challenges. Based on the results, future research directions and practical recommendations were presented to support the successful deployment of the technology.
Künstliche Intelligenz (KI) stellt eine Schlüsseltechnologie des gesellschaftlichen Wandels im 21. Jahrhundert dar. Mittlerweile werden zahlreiche technologische Anwendungen genutzt, die auf maschinellem Lernen und den damit verbundenen Möglichkeiten der Datensamm¬lung, -nutzung und -verwertung aufbauen. Indem KI große Datenmengen beherrschbar und verborgene Muster und Zusammenhänge sichtbar macht, wird vieles schneller, einfacher und effizienter – sei es im Alltag, in der Arbeit oder in Organisationen. Offen bleibt jedoch nach wie vor die Frage, welche tiefgreifenden und teilweise latenten Folgen für den Menschen als soziales Wesen und das gesellschaftliche Zusammenleben mit dem Einsatz und der Entwick¬lung von KI verbunden sind. Wie wandelt sich das Verhältnis von Mensch und Technik durch KI und wie ist dieser Wandel zu bewerten? Welche Chancen, aber auch Risiken eröffnen sich durch den Einsatz und die Entwicklung von KI für Mensch und Gesellschaft? Welchen Grenzen unterliegt der Wandel und welche Gestaltungsmöglichkeiten bieten sich? Und nicht zuletzt: Was und wer bestimmt die Entwicklungspfade, die KI nimmt – mit welchen Folgen und für wen?
Um den zunehmenden Anforderungen an die Beherrschung digitaler Techniken und an die Fähigkeit zur interdisziplinären Zusammenarbeit an Studierende aller Fachrichtungen zu begegnen wurde das interdisziplinäre Lehrformat Digitalisierungskollegs für Studierende entwickelt. Das in vielen Fachbereichen ausbaufähige Angebot von Digitalthemen in der Hoch- schullehre wird hiermit dauerhaft erweitert. Ein Digitalisierungskolleg besteht aus einer Vorlesungs- reihe mit angrenzendem Seminar, in denen Studierende interdisziplinäre Lösungen für Fragen der digitalen Transformation entwickeln. Geleitet werden sie von etablierten Wissenschaftlerinnen und Wissenschaftlern, aktiv betreut und ausgestaltet von ein bis zwei Coaches. Kernelement sowohl des Kollegs als auch der einzelnen Projekte ist die Interdisziplinarität. Eine*r der beteiligten Projekt- leiter*innen hat einen direkten Bezug zur Technik und kommt aus der Informatik, der Wirtschafts- informatik, der Elektrotechnik oder vergleichbaren Disziplinen. Zielgruppe der Projekte sind Studierende verschiedener Disziplinen im Masterstudium oder in den letzten Semestern eines Bachelorstudiums. Durch die Teilnahme erwerben auch Studierende aus digitalisierungsfernen Fächern frühzeitig umfangreiche IT-Kenntnisse. Als Begleiteffekt der umfangreichen Vernetzung zwischen den Digitalisierungskollegs (Studierende, Coaches und Projektleitende) entsteht bereits zu Beginn einer wissenschaftlichen Karriere eine große digitale Community. Alle Teilnehmenden lernen frühzeitig die interdisziplinäre Zusammenarbeit und verbessern erheblich ihre Karriere- chancen innerhalb und außerhalb der Wissenschaft.
Automated use of wired measures in grid planning for solving current and voltage band problems
(2019)
This work deals with the use of cable exchange and parallel cabling as network expansion measures in the context of an automated grid planning for the elimination of current and voltage band problems in a distribution network. Inter alia, reference is made to a higher-level program, which is being developed in the EU project "CrossEnergy", and the procedures for remedying the limit value violations are presented.
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.
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.
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.
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.
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.