Explainable nucleus classification using Decision Tree Approximation of Learned Embeddings
Bibliographic Information
Authors: Amgad M.; Atteya L.A.; Hussein H.; Mohammed K.H.; Hafiz E.; Elsebaie M.A.T.; Mobadersany P.; Manthey D.; Gutman D.A.; Elfandy H.; Cooper L.A.D.
Journal: Bioinformatics
Publisher: Oxford University Press
Publication Date: 29 September 2021
Volume / Issue: Volume 38 / Issue 2
Pages: 513–519
ISSN: 13674803
DOI: 10.1093/bioinformatics/btab670
Scopus: View on Scopus
PubMed: 34586355
Document Type: Article
Access: All Open Access; Green Open Access
Authors and Affiliations
Amgad M., Department of Pathology, Northwestern University, Chicago, IL, United States; Atteya L.A., Egyptian Ministry of Health, Cairo, Egypt; Hussein H., Department of Pathology, Nasser Institute for Research and Treatment, Cairo, Egypt; Mohammed K.H., Department of Pathology and Laboratory Medicine, University of Pennsylvania, Philadelphia, PA, United States; Hafiz E., Department of Clinical Laboratory Research, Theodor Bilharz Research Institute, Giza, Egypt; Elsebaie M.A.T., Department of Medicine, Cook County Hospital, Chicago, IL, United States; Mobadersany P., Department of Pathology, Northwestern University, Chicago, IL, United States; Manthey D., Kitware Inc., Clifton Park, NY, United States; Gutman D.A., Department of Neurology, Emory University, Atlanta, GA, United States; Elfandy H., Department of Pathology, National Cancer Institute, Cairo, Egypt; Cooper L.A.D., Department of Pathology, Northwestern University, Chicago, IL, United States
Abstract
Motivation: Nucleus detection, segmentation and classification are fundamental to high-resolution mapping of the tumor microenvironment using whole-slide histopathology images. The growing interest in leveraging the power of deep learning to achieve state-of-the-art performance often comes at the cost of explainability, yet there is general consensus that explainability is critical for trustworthiness and widespread clinical adoption. Unfortunately, current explainability paradigms that rely on pixel saliency heatmaps or superpixel importance scores are not well-suited for nucleus classification. Techniques like Grad-CAM or LIME provide explanations that are indirect, qualitative and/or nonintuitive to pathologists. Results: In this article, we present techniques to enable scalable nuclear detection, segmentation and explainable classification. First, we show how modifications to the widely used Mask R-CNN architecture, including decoupling the detection and classification tasks, improves accuracy and enables learning from hybrid annotation datasets like NuCLS, which contain mixtures of bounding boxes and segmentation boundaries. Second, we introduce an explainability method called Decision Tree Approximation of Learned Embeddings (DTALE), which provides explanations for classification model behavior globally, as well as for individual nuclear predictions. DTALE explanations are simple, quantitative, and can flexibly use any measurable morphological features that make sense to practicing pathologists, without sacrificing model accuracy. Together, these techniques present a step toward realizing the promise of computational pathology in computer-aided diagnosis and discovery of morphologic biomarkers. © 2021 The Author(s) 2021. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
Keywords
Cell Nucleus; Decision Trees; decision tree
Citation Information
Scopus Citations: 18
For comprehensive information about the Theodor Bilharz Research Institute (TBRI), its institutional activities, scientific and research achievements, clinical and hospital services, and the diverse expertise offered through its 22 specialized research and clinical departments, as well as opportunities for professional training, specialized workshops, and scientific conferences, readers are invited to visit the Institute’s official website.
The website provides regularly updated information on the Institute’s latest news, research activities, scientific initiatives, clinical services, institutional programs, and academic and professional opportunities.
English Website: https://www.tbri.sci.eg/en/
Arabic Website: https://www.tbri.sci.eg/ar/
Prepared and Uploaded by:
Abdalla F. Abdalla
Electronic Portal Unit
