HistoMapr™: An Explainable AI (xAI) Platform for Computational Pathology Solutions

Pathologists are adopting whole slide images (WSIs) for diagnostic purposes. While doing so, pathologists should have all the information needed to make best diagnoses rapidly, while supervising computational pathology tools in real-time. Computational pathology has great potential for augmenting pathologists’ accuracy and efficiency, but concern exists regarding trust for ‘black-box AI’ solutions. Explainable AI (xAI) can reveal underlying reasons for its results, to promote safety, reliability, and accountability for critical tasks such as pathology diagnosis. Built on a hierarchy of computational and traditional image analysis algorithms, we present the development of our proprietary xAI software platform, HistoMapr, for pathologists to improve their efficiency and accuracy when viewing WSIs. HistoMapr and xAI represent a powerful and transparent alternative to ‘black-box’ AI. HistoMapr previews WSIs then presents key diagnostic areas first in an interactive, explainable fashion. Pathologists can access xAI features via a “Why?” button in the interface. Furthermore, two critical early application examples are presented: 1) Intelligent triaging that involves xAI estimation of difficulty for new cases to be forwarded to subspecialists or generalist pathologists; 2) Retrospective quality assurance entails detection of potential discrepancies between finalized results and xAI reviews. Finally, a prototype is presented for atypical ductal hyperplasia, a diagnostic challenge in breast pathology, where xAI descriptions were based on computational pipeline image results.

[1]  Wojciech Samek,et al.  Methods for interpreting and understanding deep neural networks , 2017, Digit. Signal Process..

[2]  Wenyu Liu,et al.  DeepMitosis: Mitosis detection via deep detection, verification and segmentation networks , 2018, Medical Image Anal..

[3]  Jeffrey L. Fine.  21st century workflow: A proposal , 2014, Journal of pathology informatics.

[4]  Donald L Weaver,et al.  The diagnostic challenge of low-grade ductal carcinoma in situ. , 2017, European journal of cancer.

[5]  Andrew Janowczyk,et al.  Deep learning for digital pathology image analysis: A comprehensive tutorial with selected use cases , 2016, Journal of pathology informatics.

[6]  Andrew H. Beck,et al.  Computational Pathology to Discriminate Benign from Malignant Intraductal Proliferations of the Breast , 2014, PloS one.

[7]  Tahsin Kurc,et al.  Twenty Years of Digital Pathology: An Overview of the Road Travelled, What is on the Horizon, and the Emergence of Vendor-Neutral Archives , 2018, Journal of pathology informatics.

[8]  Michael C. Montalto,et al.  An industry perspective: An update on the adoption of whole slide imaging , 2016, Journal of pathology informatics.

[9]  Gustavo K. Rohde,et al.  A Computer-Based Automated Algorithm for Assessing Acinar Cell Loss after Experimental Pancreatitis , 2014, PloS one.

[10]  Andreas Holzinger,et al.  Augmenting Statistical Data Dissemination by Short Quantified Sentences of Natural Language , 2018, Journal of Official Statistics.

[11]  Surabhi Bhargava,et al.  A Dataset and a Technique for Generalized Nuclear Segmentation for Computational Pathology , 2017, IEEE Transactions on Medical Imaging.

[12]  Fiona Ginty,et al.  A spatial analytics computational and systems biology platform predicts risk of colorectal cancer recurrence and identifies emergent spatial domain networks associated with recurrence , 2019, bioRxiv.

[13]  D. Lansing Taylor,et al.  Spatial Statistics for Segmenting Histological Structures in H&E Stained Tissue Images , 2017, IEEE Transactions on Medical Imaging.

[14]  Alexis B. Carter,et al.  Computational Pathology: A Path Ahead. , 2016, Archives of pathology & laboratory medicine.

[15]  Linda G. Shapiro,et al.  Assessment of Machine Learning of Breast Pathology Structures for Automated Differentiation of Breast Cancer and High-Risk Proliferative Lesions , 2019, JAMA network open.

[16]  Toby C. Cornish,et al.  US Food and Drug Administration Approval of Whole Slide Imaging for Primary Diagnosis: A Key Milestone Is Reached and New Questions Are Raised. , 2018, Archives of pathology & laboratory medicine.

[17]  Luong Nguyen,et al.  Architectural patterns for differential diagnosis of proliferative breast lesions from histopathological images , 2017, 2017 IEEE 14th International Symposium on Biomedical Imaging (ISBI 2017).

[18]  A. B. Tosun,et al.  Detection of malignant mesothelioma using nuclear structure of mesothelial cells in effusion cytology specimens , 2015, Cytometry. Part A : the journal of the International Society for Analytical Cytology.

[19]  Luong Nguyen,et al.  Histological Detection of High-Risk Benign Breast Lesions from Whole Slide Images , 2017, MICCAI.

[20]  Joachim M. Buhmann,et al.  Computational Pathology: Challenges and Promises for Tissue Analysis , 2015, Comput. Medical Imaging Graph..

[21]  J. Simpson,et al.  Differential Diagnosis of Proliferative Breast Lesions. , 2009, Surgical pathology clinics.

[22]  H. Chad Lane,et al.  Building Explainable Artificial Intelligence Systems , 2006, AAAI.

[23]  Fiona Ginty,et al.  Spatial domain analysis predicts risk of colorectal cancer recurrence and infers associated tumor microenvironment networks , 2019 .

[24]  Sidra Nawaz,et al.  Computational pathology: Exploring the spatial dimension of tumor ecology. , 2016, Cancer letters.

[25]  Catarina Eloy,et al.  BACH: Grand Challenge on Breast Cancer Histology Images , 2018, Medical Image Anal..

[26]  Cynthia Rudin,et al.  Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead , 2018, Nature Machine Intelligence.

[27]  Jason Yosinski,et al.  Deep neural networks are easily fooled: High confidence predictions for unrecognizable images , 2014, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[28]  Anant Madabhushi,et al.  Quantitative nuclear histomorphometric features are predictive of Oncotype DX risk categories in ductal carcinoma in situ: preliminary findings , 2019, Breast Cancer Research.

[29]  Liron Pantanowitz,et al.  Artificial Intelligence and Digital Pathology: Challenges and Opportunities , 2018, Journal of pathology informatics.

[30]  Guang-Zhong Yang,et al.  XAI—Explainable artificial intelligence , 2019, Science Robotics.

[31]  Luciano Floridi,et al.  Establishing the rules for building trustworthy AI , 2019, Nature Machine Intelligence.

[32]  Jia Guo,et al.  Cancer diagnosis by nuclear morphometry using spatial information , 2014, Pattern Recognit. Lett..

[33]  Nico Karssemeijer,et al.  Context-aware stacked convolutional neural networks for classification of breast carcinomas in whole-slide histopathology images , 2017, Journal of medical imaging.

[34]  Joachim M. Buhmann,et al.  Computational Pathology Analysis of Tissue Microarrays Predicts Survival of Renal Clear Cell Carcinoma Patients , 2008, MICCAI.

[35]  Carlos Guestrin,et al.  "Why Should I Trust You?": Explaining the Predictions of Any Classifier , 2016, ArXiv.

[36]  Zheng Liu,et al.  Pathologists' Computer-Assisted Diagnosis: A Mock-up of a Prototype Information System to Facilitate Automation of Pathology Sign-out. , 2017, Archives of pathology & laboratory medicine.

[37]  Cenk Sokmensuer,et al.  Unsupervised Tissue Image Segmentation through Object-Oriented Texture , 2010, 2010 20th International Conference on Pattern Recognition.

[38]  Andreas Holzinger,et al.  Measuring the Quality of Explanations: The System Causability Scale (SCS) , 2020, KI - Künstliche Intelligenz.

[39]  J. Elmore,et al.  Diagnostic concordance among pathologists interpreting breast biopsy specimens. , 2015, JAMA.