Emely Rosbach
I am a research assistant at the Technical University of Applied Sciences Ingolstadt, where I also received my Master’s degree in User Experience Design. My current research focuses on understanding how medical professionals interact with AI-based decision support systems in digital pathology and radiology. I am particularly interested in how human cognition, specifically cognitive biases, shapes human-machine collaboration in critical environments such as healthcare.
Publications:
2024
2.
Ganz, Jonathan; Marzahl, Christian; Ammeling, Jonas; Rosbach, Emely; Richter, Barbara; Puget, Chloé; Denk, Daniela; Demeter, Elena A.; Tăbăran, Flaviu A.; Wasinger, Gabriel; Lipnik, Karoline; Tecilla, Marco; Valentine, Matthew J.; Dark, Michael J.; Abele, Niklas; Bolfa, Pompei; Erber, Ramona; Klopfleisch, Robert; Merz, Sophie; Donovan, Taryn A.; Jabari, Samir; Bertram, Christof A.; Breininger, Katharina; Aubreville, Marc
Information mismatch in PHH3-assisted mitosis annotation leads to interpretation shifts in H&E slide analysis Journal Article
In: Scientific Reports, vol. 14, no. 1, pp. 26273, 2024, ISSN: 2045-2322.
Abstract | Links | BibTeX | Tags:
@article{ganz_information_2024,
title = {Information mismatch in PHH3-assisted mitosis annotation leads to interpretation shifts in H&E slide analysis},
author = {Jonathan Ganz and Christian Marzahl and Jonas Ammeling and Emely Rosbach and Barbara Richter and Chloé Puget and Daniela Denk and Elena A. Demeter and Flaviu A. Tăbăran and Gabriel Wasinger and Karoline Lipnik and Marco Tecilla and Matthew J. Valentine and Michael J. Dark and Niklas Abele and Pompei Bolfa and Ramona Erber and Robert Klopfleisch and Sophie Merz and Taryn A. Donovan and Samir Jabari and Christof A. Bertram and Katharina Breininger and Marc Aubreville},
url = {https://www.nature.com/articles/s41598-024-77244-6},
doi = {10.1038/s41598-024-77244-6},
issn = {2045-2322},
year = {2024},
date = {2024-11-01},
urldate = {2024-11-04},
journal = {Scientific Reports},
volume = {14},
number = {1},
pages = {26273},
abstract = {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.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
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.
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.
1.
Aubreville, Marc; Ganz, Jonathan; Ammeling, Jonas; Rosbach, Emely; Gehrke, Thomas; Scherzad, Agmal; Hackenberg, Stephan; Goncalves, Miguel
Prediction of tumor board procedural recommendations using large language models Journal Article
In: European Archives of Oto-Rhino-Laryngology, 2024, ISSN: 1434-4726.
Abstract | Links | BibTeX | Tags:
@article{aubreville_prediction_2024,
title = {Prediction of tumor board procedural recommendations using large language models},
author = {Marc Aubreville and Jonathan Ganz and Jonas Ammeling and Emely Rosbach and Thomas Gehrke and Agmal Scherzad and Stephan Hackenberg and Miguel Goncalves},
url = {https://doi.org/10.1007/s00405-024-08947-9},
doi = {10.1007/s00405-024-08947-9},
issn = {1434-4726},
year = {2024},
date = {2024-09-01},
journal = {European Archives of Oto-Rhino-Laryngology},
abstract = {Multidisciplinary tumor boards are meetings where a team of medical specialists, including medical oncologists, radiation oncologists, radiologists, surgeons, and pathologists, collaborate to determine the best treatment plan for cancer patients. While decision-making in this context is logistically and cost-intensive, it has a significant positive effect on overall cancer survival.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Multidisciplinary tumor boards are meetings where a team of medical specialists, including medical oncologists, radiation oncologists, radiologists, surgeons, and pathologists, collaborate to determine the best treatment plan for cancer patients. While decision-making in this context is logistically and cost-intensive, it has a significant positive effect on overall cancer survival.