@article{PugetGanzBertrametal.2025, author = {Puget, Chlo{\´e} and Ganz, Jonathan and Bertram, Christof and Conrad, Thomas and Baeblich, Malte and Voss, Anne and Landmann, Katharina and Haake, Alexander F. H. and Spree, Andreas and Hartung, Svenja and Aeschlimann, Leonore and Soto, Sara and de Brot, Simone and Dettwiler, Martina and Aupperle-Lellbach, Heike and Bolfa, Pompei and Bartel, Alexander and Kiupel, Matti and Breininger, Katharina and Aubreville, Marc and Klopfleisch, Robert}, title = {Artificial intelligence predicts c-KIT exon 11 genotype by phenotype in canine cutaneous mast cell tumors: Can human observers learn it?}, volume = {63}, journal = {Veterinary Pathology}, number = {2}, publisher = {Sage}, address = {London}, issn = {1544-2217}, doi = {https://doi.org/10.1177/03009858251380284}, pages = {369 -- 379}, year = {2025}, abstract = {Canine cutaneous mast cell tumors (ccMCTs) are frequent neoplasms with variable biological behaviors. Internal tandem duplication mutations in c-KIT exon 11 (c-KIT-11-ITD) are associated with poor prognosis but predict therapeutic response to tyrosine kinase inhibitors. In a previous work, deep learning algorithms managed to predict the presence of c-KIT-11-ITD on digitalized hematoxylin and eosin-stained histological slides (whole-slide images, WSIs) in up to 87\% of cases, suggesting the existence of morphological features characterizing ccMCTs carrying c-KIT-11-ITD. This 3-stage blinded study aimed to identify morphological features indicative of c-KIT-11-ITD and to evaluate the ability of human observers to learn this task. 17 untrained pathologists first classified 8 WSIs and 200 image patches (highly relevant for algorithmic classification) of ccMCTs as either positive or negative for c-KIT-11-ITD. Second, they self-trained to recognize c-KIT-11-ITD by looking at the same WSIs and patches correctly sorted. Third, pathologists classified 15 new WSIs and 200 new patches according to c-KIT-11-ITD status. In addition, participants reported microscopic features they considered relevant for their decision. Without training, participants correctly classified the c-KIT-11-ITD status of 63\%-88\% of WSIs and 43\%-55\% of patches. With self-training, 25\%-38\% of WSIs and 55\%-56\% of patches were correctly classified. High cellular pleomorphism, anisokaryosis, and sparse cytoplasmic granulation were commonly suggested as features associated with c-KIT-11-ITD-positive ccMCTs, none of which showed reliable predictivity in a follow-up study. The results indicate that transfer of algorithmic skills to the human observer is difficult. A c-KIT-11-ITD-specific morphological feature remains to be extracted from the artificial intelligence model.}, language = {en} } @unpublished{AubrevilleStathonikosDonovanetal.2023, author = {Aubreville, Marc and Stathonikos, Nikolas and Donovan, Taryn and Klopfleisch, Robert and Ganz, Jonathan and Ammeling, Jonas and Wilm, Frauke and Veta, Mitko and Jabari, Samir and Eckstein, Markus and Annuscheit, Jonas and Krumnow, Christian and Bozaba, Engin and Cayir, Sercan and Gu, Hongyan and Chen, Xiang and Jahanifar, Mostafa and Shephard, Adam and Kondo, Satoshi and Kasai, Satoshi and Kotte, Sujatha and Saipradeep, Vangala and Lafarge, Maxime W. and Koelzer, Viktor H. and Wang, Ziyue and Zhang, Yongbing and Yang, Sen and Wang, Xiyue and Breininger, Katharina and Bertram, Christof}, title = {Domain generalization across tumor types, laboratories, and species - Insights from the 2022 edition of the Mitosis Domain Generalization Challenge}, publisher = {arXiv}, address = {Ithaca}, doi = {https://doi.org/10.48550/arXiv.2309.15589}, year = {2023}, abstract = {Recognition of mitotic figures in histologic tumor specimens is highly relevant to patient outcome assessment. This task is challenging for algorithms and human experts alike, with deterioration of algorithmic performance under shifts in image representations. Considerable covariate shifts occur when assessment is performed on different tumor types, images are acquired using different digitization devices, or specimens are produced in different laboratories. This observation motivated the inception of the 2022 challenge on MItosis Domain Generalization (MIDOG 2022). The challenge provided annotated histologic tumor images from six different domains and evaluated the algorithmic approaches for mitotic figure detection provided by nine challenge participants on ten independent domains. Ground truth for mitotic figure detection was established in two ways: a three-expert consensus and an independent, immunohistochemistry-assisted set of labels. This work represents an overview of the challenge tasks, the algorithmic strategies employed by the participants, and potential factors contributing to their success. With an F1 score of 0.764 for the top-performing team, we summarize that domain generalization across various tumor domains is possible with today's deep learning-based recognition pipelines. When assessed against the immunohistochemistry-assisted reference standard, all methods resulted in reduced recall scores, but with only minor changes in the order of participants in the ranking.}, language = {en} } @unpublished{GanzMarzahlAmmelingetal.2024, author = {Ganz, Jonathan and Marzahl, Christian and Ammeling, Jonas and Richter, Barbara and Puget, Chlo{\´e} and Denk, Daniela and Demeter, Elena A. and Tabaran, Flaviu A. and Wasinger, Gabriel and Lipnik, Karoline and Tecilla, Marco and Valentine, Matthew J. and Dark, Michael and Abele, Niklas and Bolfa, Pompei and Erber, Ramona and Klopfleisch, Robert and Merz, Sophie and Donovan, Taryn and Jabari, Samir and Bertram, Christof and Breininger, Katharina and Aubreville, Marc}, title = {On the Value of PHH3 for Mitotic Figure Detection on H\&E-stained Images}, publisher = {arXiv}, address = {Ithaca}, doi = {https://doi.org/10.48550/arXiv.2406.19899}, year = {2024}, abstract = {The count of mitotic figures (MFs) observed in hematoxylin and eosin (H\&E)-stained slides is an important prognostic marker as it is a measure for tumor cell proliferation. However, the identification of MFs has a known low inter-rater agreement. Deep learning algorithms can standardize this task, but they require large amounts of annotated data for training and validation. Furthermore, label noise introduced during the annotation process may impede the algorithm's performance. Unlike H\&E, the mitosis-specific antibody phospho-histone H3 (PHH3) specifically highlights MFs. Counting MFs on slides stained against PHH3 leads to higher agreement among raters and has therefore recently been used as a ground truth for the annotation of MFs in H\&E. However, as PHH3 facilitates the recognition of cells indistinguishable from H\&E stain alone, the use of this ground truth could potentially introduce noise into the H\&E-related dataset, impacting model performance. This study analyzes the impact of PHH3-assisted MF annotation on inter-rater reliability and object level agreement through an extensive multi-rater experiment. We found that the annotators' object-level agreement increased when using PHH3-assisted labeling. Subsequently, MF detectors were evaluated on the resulting datasets to investigate the influence of PHH3-assisted labeling on the models' performance. Additionally, a novel dual-stain MF detector was developed to investigate the interpretation-shift of PHH3-assisted labels used in H\&E, which clearly outperformed single-stain detectors. However, the PHH3-assisted labels did not have a positive effect on solely H\&E-based models. The high performance of our dual-input detector reveals an information mismatch between the H\&E and PHH3-stained images as the cause of this effect.}, language = {en} } @unpublished{GanzAmmelingRosbachetal.2024, author = {Ganz, Jonathan and Ammeling, Jonas and Rosbach, Emely and Lausser, Ludwig and Bertram, Christof and Breininger, Katharina and Aubreville, Marc}, title = {Is Self-Supervision Enough? Benchmarking Foundation Models Against End-to-End Training for Mitotic Figure Classification}, publisher = {arXiv}, address = {Ithaca}, doi = {https://doi.org/10.48550/arXiv.2412.06365}, year = {2024}, abstract = {Foundation models (FMs), i.e., models trained on a vast amount of typically unlabeled data, have become popular and available recently for the domain of histopathology. The key idea is to extract semantically rich vectors from any input patch, allowing for the use of simple subsequent classification networks potentially reducing the required amounts of labeled data, and increasing domain robustness. In this work, we investigate to which degree this also holds for mitotic figure classification. Utilizing two popular public mitotic figure datasets, we compared linear probing of five publicly available FMs against models trained on ImageNet and a simple ResNet50 end-to-end-trained baseline. We found that the end-to-end-trained baseline outperformed all FM-based classifiers, regardless of the amount of data provided. Additionally, we did not observe the FM-based classifiers to be more robust against domain shifts, rendering both of the above assumptions incorrect.}, language = {en} } @inproceedings{GanzBertramKlopfleischetal.2022, author = {Ganz, Jonathan and Bertram, Christof and Klopfleisch, Robert and Jabari, Samir and Breininger, Katharina and Aubreville, Marc}, title = {Classification of visibility in multi-stain microscopy images}, booktitle = {Medical Imaging with Deep Learning: MIDL 2022 Short Papers}, url = {https://openreview.net/forum?id=-GsA-mUVmm}, year = {2022}, language = {en} } @unpublished{PugetGanzOstermaieretal.2024, author = {Puget, Chlo{\´e} and Ganz, Jonathan and Ostermaier, Julian and Konrad, Thomas and Parlak, Eda and Bertram, Christof and Kiupel, Matti and Breininger, Katharina and Aubreville, Marc and Klopfleisch, Robert}, title = {Deep Learning model predicts the c-Kit-11 mutational status of canine cutaneous mast cell tumors by HE stained histological slides}, publisher = {arXiv}, address = {Ithaca}, doi = {https://doi.org/10.48550/arXiv.2401.06169}, year = {2024}, abstract = {Numerous prognostic factors are currently assessed histopathologically in biopsies of canine mast cell tumors to evaluate clinical behavior. In addition, PCR analysis of the c-Kit exon 11 mutational status is often performed to evaluate the potential success of a tyrosine kinase inhibitor therapy. This project aimed at training deep learning models (DLMs) to identify the c-Kit-11 mutational status of MCTs solely based on morphology without additional molecular analysis. HE slides of 195 mutated and 173 non-mutated tumors were stained consecutively in two different laboratories and scanned with three different slide scanners. This resulted in six different datasets (stain-scanner variations) of whole slide images. DLMs were trained with single and mixed datasets and their performances was assessed under scanner and staining domain shifts. The DLMs correctly classified HE slides according to their c-Kit 11 mutation status in, on average, 87\% of cases for the best-suited stain-scanner variant. A relevant performance drop could be observed when the stain-scanner combination of the training and test dataset differed. Multi-variant datasets improved the average accuracy but did not reach the maximum accuracy of algorithms trained and tested on the same stain-scanner variant. In summary, DLM-assisted morphological examination of MCTs can predict c-Kit-exon 11 mutational status of MCTs with high accuracy. However, the recognition performance is impeded by a change of scanner or staining protocol. Larger data sets with higher numbers of scans originating from different laboratories and scanners may lead to more robust DLMs to identify c-Kit mutations in HE slides.}, language = {en} } @article{BertramAubrevilleDonovanetal.2021, author = {Bertram, Christof and Aubreville, Marc and Donovan, Taryn and Bartel, Alexander and Wilm, Frauke and Marzahl, Christian and Assenmacher, Charles-Antoine and Becker, Kathrin and Bennett, Mark and Corner, Sarah M. and Cossic, Brieuc and Denk, Daniela and Dettwiler, Martina and Garcia Gonzalez, Beatriz and Gurtner, Corinne and Haverkamp, Ann-Kathrin and Heier, Annabelle and Lehmbecker, Annika and Merz, Sophie and Noland, Erica L. and Plog, Stephanie and Schmidt, Anja and Sebastian, Franziska and Sledge, Dodd G. and Smedley, Rebecca C. and Tecilla, Marco and Thaiwong, Tuddow and Fuchs-Baumgartinger, Andrea and Meuten, Donald J. and Breininger, Katharina and Kiupel, Matti and Maier, Andreas and Klopfleisch, Robert}, title = {Computer-assisted mitotic count using a deep learning-based algorithm improves interobserver reproducibility and accuracy}, volume = {59}, journal = {Veterinary Pathology}, number = {2}, publisher = {Sage}, address = {London}, issn = {1544-2217}, doi = {https://doi.org/10.1177/03009858211067478}, pages = {211 -- 226}, year = {2021}, abstract = {The mitotic count (MC) is an important histological parameter for prognostication of malignant neoplasms. However, it has inter- and intraobserver discrepancies due to difficulties in selecting the region of interest (MC-ROI) and in identifying or classifying mitotic figures (MFs). Recent progress in the field of artificial intelligence has allowed the development of high-performance algorithms that may improve standardization of the MC. As algorithmic predictions are not flawless, computer-assisted review by pathologists may ensure reliability. In the present study, we compared partial (MC-ROI preselection) and full (additional visualization of MF candidates and display of algorithmic confidence values) computer-assisted MC analysis to the routine (unaided) MC analysis by 23 pathologists for whole-slide images of 50 canine cutaneous mast cell tumors (ccMCTs). Algorithmic predictions aimed to assist pathologists in detecting mitotic hotspot locations, reducing omission of MFs, and improving classification against imposters. The interobserver consistency for the MC significantly increased with computer assistance (interobserver correlation coefficient, ICC = 0.92) compared to the unaided approach (ICC = 0.70). Classification into prognostic stratifications had a higher accuracy with computer assistance. The algorithmically preselected hotspot MC-ROIs had a consistently higher MCs than the manually selected MC-ROIs. Compared to a ground truth (developed with immunohistochemistry for phosphohistone H3), pathologist performance in detecting individual MF was augmented when using computer assistance (F1-score of 0.68 increased to 0.79) with a reduction in false negatives by 38\%. The results of this study demonstrate that computer assistance may lead to more reproducible and accurate MCs in ccMCTs.}, language = {en} } @inproceedings{MarzahlWilmTharunetal.2021, author = {Marzahl, Christian and Wilm, Frauke and Tharun, Lars and Perner, Sven and Kr{\"o}ger, Christine and Voigt, J{\"o}rn and Klopfleisch, Robert and Maier, Andreas and Aubreville, Marc and Breininger, Katharina}, title = {Robust quad-tree based registration on whole slide images}, booktitle = {Proceedings of Machine Learning Research: Proceedings of COMPAY 2021}, number = {156}, publisher = {PMLR}, address = {[s. l.]}, url = {https://proceedings.mlr.press/v156/marzahl21a.html}, pages = {181 -- 190}, year = {2021}, language = {en} } @inproceedings{GanzKirschHoffmannetal.2021, author = {Ganz, Jonathan and Kirsch, Tobias and Hoffmann, Lucas and Maier, Andreas and Breininger, Katharina and Bl{\"u}mcke, Ingmar and Jabari, Samir and Aubreville, Marc}, title = {Automatic and explainable grading of meningiomas from histopathology images}, volume = {2021}, booktitle = {Proceedings of Machine Learning Research: Proceedings of COMPAY 2021}, number = {156}, publisher = {PMLR}, address = {[s. l.]}, url = {https://proceedings.mlr.press/v156/ganz21a.html}, pages = {69 -- 80}, year = {2021}, language = {en} } @unpublished{GanzMarzahlAmmelingetal.2024, author = {Ganz, Jonathan and Marzahl, Christian and Ammeling, Jonas and Rosbach, Emely and Richter, Barbara and Puget, Chlo{\´e} and Denk, Daniela and Demeter, Elena A. and Tabaran, Flaviu A. and Wasinger, Gabriel and Lipnik, Karoline and Tecilla, Marco and Valentine, Matthew J. and Dark, Michael and Abele, Niklas and Bolfa, Pompei and Erber, Ramona and Klopfleisch, Robert and Merz, Sophie and Donovan, Taryn and Jabari, Samir and Bertram, Christof and Breininger, Katharina and Aubreville, Marc}, title = {Information Mismatch in PHH3-Assisted Mitosis Annotation Leads to Interpretation Shifts in H\&E Slide Analysis}, titleParent = {Research Square}, publisher = {Research Square}, address = {Durham}, doi = {https://doi.org/10.21203/rs.3.rs-4900505/v1}, year = {2024}, abstract = {The count of mitotic figures (MFs) observed in hematoxylin and eosin (H\&E)-stained slides is an important prognostic marker, as it is a measure for tumor cell proliferation. However, the identification of MFs has a known low inter-rater agreement. In a computer-aided setting, deep learning algorithms can help to mitigate this, but they require large amounts of annotated data for training and validation. Furthermore, label noise introduced during the annotation process may impede the algorithms' performance. Unlike H\&E, where identification of MFs is based mainly on morphological features, the mitosis-specific antibody phospho-histone H3 (PHH3) specifically highlights MFs. Counting MFs on slides stained against PHH3 leads to higher agreement among raters and has therefore recently been used as a ground truth for the annotation of MFs in H\&E. However, as PHH3 facilitates the recognition of cells indistinguishable from H\&E staining alone, the use of this ground truth could potentially introduce an interpretation shift and even label noise into the H\&E-related dataset, impacting model performance. This study analyzes the impact of PHH3-assisted MF annotation on inter-rater reliability and object level agreement through an extensive multi-rater experiment. Subsequently, MF detectors, including a novel dual-stain detector, were evaluated on the resulting datasets to investigate the influence of PHH3-assisted labeling on the models' performance. We found that the annotators' object-level agreement significantly increased when using PHH3-assisted labeling (F1: 0.53 to 0.74). However, this enhancement in label consistency did not translate to improved performance for H\&E-based detectors, neither during the training phase nor the evaluation phase. Conversely, the dual-stain detector was able to benefit from the higher consistency. This reveals an information mismatch between the H\&E and PHH3-stained images as the cause of this effect, which renders PHH3-assisted annotations not well-aligned for use with H\&E-based detectors. Based on our findings, we propose an improved PHH3-assisted labeling procedure.}, language = {en} } @unpublished{AmmelingHeckerGanzetal.2023, author = {Ammeling, Jonas and Hecker, Moritz and Ganz, Jonathan and Donovan, Taryn and Klopfleisch, Robert and Bertram, Christof and Breininger, Katharina and Aubreville, Marc}, title = {Automated Volume Corrected Mitotic Index Calculation Through Annotation-Free Deep Learning using Immunohistochemistry as Reference Standard}, publisher = {arXiv}, address = {Ithaca}, doi = {https://doi.org/10.48550/arXiv.2311.08949}, year = {2023}, abstract = {The volume-corrected mitotic index (M/V-Index) was shown to provide prognostic value in invasive breast carcinomas. However, despite its prognostic significance, it is not established as the standard method for assessing aggressive biological behaviour, due to the high additional workload associated with determining the epithelial proportion. In this work, we show that using a deep learning pipeline solely trained with an annotation-free, immunohistochemistry-based approach, provides accurate estimations of epithelial segmentation in canine breast carcinomas. We compare our automatic framework with the manually annotated M/V-Index in a study with three board-certified pathologists. Our results indicate that the deep learning-based pipeline shows expert-level performance, while providing time efficiency and reproducibility.}, language = {en} } @unpublished{WilmFragosoGarciaBertrametal.2022, author = {Wilm, Frauke and Fragoso-Garcia, Marco and Bertram, Christof and Stathonikos, Nikolas and {\"O}ttl, Mathias and Qiu, Jingna and Klopfleisch, Robert and Maier, Andreas and Aubreville, Marc and Breininger, Katharina}, title = {Mind the Gap: Scanner-induced domain shifts pose challenges for representation learning in histopathology}, publisher = {arXiv}, address = {Ithaca}, doi = {https://doi.org/10.48550/arXiv.2211.16141}, year = {2022}, language = {en} }