Deep learning generates synthetic cancer histology for explainability and education
Abstract Artificial intelligence methods including deep neural networks (DNN) can provide rapid molecular classification of tumors from routine histology with accuracy that matches or exceeds human pathologists. Discerning how neural networks make their predictions remains a significant challenge, b...
| 出版年: | npj Precision Oncology |
|---|---|
| 主要な著者: | , , , , , , , , , , , , , , , , , , , , , , |
| フォーマット: | 論文 |
| 言語: | 英語 |
| 出版事項: |
Nature Portfolio
2023-05-01
|
| オンライン･アクセス: | https://doi.org/10.1038/s41698-023-00399-4 |
| _version_ | 1850354645763883008 |
|---|---|
| author | James M. Dolezal Rachelle Wolk Hanna M. Hieromnimon Frederick M. Howard Andrew Srisuwananukorn Dmitry Karpeyev Siddhi Ramesh Sara Kochanny Jung Woo Kwon Meghana Agni Richard C. Simon Chandni Desai Raghad Kherallah Tung D. Nguyen Jefree J. Schulte Kimberly Cole Galina Khramtsova Marina Chiara Garassino Aliya N. Husain Huihua Li Robert Grossman Nicole A. Cipriani Alexander T. Pearson |
| author_facet | James M. Dolezal Rachelle Wolk Hanna M. Hieromnimon Frederick M. Howard Andrew Srisuwananukorn Dmitry Karpeyev Siddhi Ramesh Sara Kochanny Jung Woo Kwon Meghana Agni Richard C. Simon Chandni Desai Raghad Kherallah Tung D. Nguyen Jefree J. Schulte Kimberly Cole Galina Khramtsova Marina Chiara Garassino Aliya N. Husain Huihua Li Robert Grossman Nicole A. Cipriani Alexander T. Pearson |
| author_sort | James M. Dolezal |
| collection | DOAJ |
| container_title | npj Precision Oncology |
| description | Abstract Artificial intelligence methods including deep neural networks (DNN) can provide rapid molecular classification of tumors from routine histology with accuracy that matches or exceeds human pathologists. Discerning how neural networks make their predictions remains a significant challenge, but explainability tools help provide insights into what models have learned when corresponding histologic features are poorly defined. Here, we present a method for improving explainability of DNN models using synthetic histology generated by a conditional generative adversarial network (cGAN). We show that cGANs generate high-quality synthetic histology images that can be leveraged for explaining DNN models trained to classify molecularly-subtyped tumors, exposing histologic features associated with molecular state. Fine-tuning synthetic histology through class and layer blending illustrates nuanced morphologic differences between tumor subtypes. Finally, we demonstrate the use of synthetic histology for augmenting pathologist-in-training education, showing that these intuitive visualizations can reinforce and improve understanding of histologic manifestations of tumor biology. |
| format | Article |
| id | doaj-art-90c8d0f480c347d78a0bdd3a1cd68a11 |
| institution | Directory of Open Access Journals |
| issn | 2397-768X |
| language | English |
| publishDate | 2023-05-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| spelling | doaj-art-90c8d0f480c347d78a0bdd3a1cd68a112025-08-19T23:08:08ZengNature Portfolionpj Precision Oncology2397-768X2023-05-017111310.1038/s41698-023-00399-4Deep learning generates synthetic cancer histology for explainability and educationJames M. Dolezal0Rachelle Wolk1Hanna M. Hieromnimon2Frederick M. Howard3Andrew Srisuwananukorn4Dmitry Karpeyev5Siddhi Ramesh6Sara Kochanny7Jung Woo Kwon8Meghana Agni9Richard C. Simon10Chandni Desai11Raghad Kherallah12Tung D. Nguyen13Jefree J. Schulte14Kimberly Cole15Galina Khramtsova16Marina Chiara Garassino17Aliya N. Husain18Huihua Li19Robert Grossman20Nicole A. Cipriani21Alexander T. Pearson22Section of Hematology/Oncology, Department of Medicine, University of Chicago MedicineDepartment of Pathology, University of Chicago MedicineSection of Hematology/Oncology, Department of Medicine, University of Chicago MedicineSection of Hematology/Oncology, Department of Medicine, University of Chicago MedicineTisch Cancer Institute, Icahn School of Medicine at Mount SinaiGhost Autonomy, Inc.Section of Hematology/Oncology, Department of Medicine, University of Chicago MedicineSection of Hematology/Oncology, Department of Medicine, University of Chicago MedicineDepartment of Pathology, University of Chicago MedicineDepartment of Pathology, University of Chicago MedicineDepartment of Pathology, University of Chicago MedicineDepartment of Pathology, University of Chicago MedicineDepartment of Pathology, University of Chicago MedicineDepartment of Pathology, University of Chicago MedicineDepartment of Pathology and Laboratory Medicine, University of Wisconsin at MadisonDepartment of Pathology, University of Chicago MedicineDepartment of Pathology, University of Chicago MedicineSection of Hematology/Oncology, Department of Medicine, University of Chicago MedicineDepartment of Pathology, University of Chicago MedicineDepartment of Pathology, University of Chicago MedicineUniversity of Chicago, Center for Translational Data ScienceDepartment of Pathology, University of Chicago MedicineSection of Hematology/Oncology, Department of Medicine, University of Chicago MedicineAbstract Artificial intelligence methods including deep neural networks (DNN) can provide rapid molecular classification of tumors from routine histology with accuracy that matches or exceeds human pathologists. Discerning how neural networks make their predictions remains a significant challenge, but explainability tools help provide insights into what models have learned when corresponding histologic features are poorly defined. Here, we present a method for improving explainability of DNN models using synthetic histology generated by a conditional generative adversarial network (cGAN). We show that cGANs generate high-quality synthetic histology images that can be leveraged for explaining DNN models trained to classify molecularly-subtyped tumors, exposing histologic features associated with molecular state. Fine-tuning synthetic histology through class and layer blending illustrates nuanced morphologic differences between tumor subtypes. Finally, we demonstrate the use of synthetic histology for augmenting pathologist-in-training education, showing that these intuitive visualizations can reinforce and improve understanding of histologic manifestations of tumor biology.https://doi.org/10.1038/s41698-023-00399-4 |
| spellingShingle | James M. Dolezal Rachelle Wolk Hanna M. Hieromnimon Frederick M. Howard Andrew Srisuwananukorn Dmitry Karpeyev Siddhi Ramesh Sara Kochanny Jung Woo Kwon Meghana Agni Richard C. Simon Chandni Desai Raghad Kherallah Tung D. Nguyen Jefree J. Schulte Kimberly Cole Galina Khramtsova Marina Chiara Garassino Aliya N. Husain Huihua Li Robert Grossman Nicole A. Cipriani Alexander T. Pearson Deep learning generates synthetic cancer histology for explainability and education |
| title | Deep learning generates synthetic cancer histology for explainability and education |
| title_full | Deep learning generates synthetic cancer histology for explainability and education |
| title_fullStr | Deep learning generates synthetic cancer histology for explainability and education |
| title_full_unstemmed | Deep learning generates synthetic cancer histology for explainability and education |
| title_short | Deep learning generates synthetic cancer histology for explainability and education |
| title_sort | deep learning generates synthetic cancer histology for explainability and education |
| url | https://doi.org/10.1038/s41698-023-00399-4 |
| work_keys_str_mv | AT jamesmdolezal deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT rachellewolk deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT hannamhieromnimon deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT frederickmhoward deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT andrewsrisuwananukorn deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT dmitrykarpeyev deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT siddhiramesh deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT sarakochanny deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT jungwookwon deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT meghanaagni deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT richardcsimon deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT chandnidesai deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT raghadkherallah deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT tungdnguyen deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT jefreejschulte deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT kimberlycole deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT galinakhramtsova deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT marinachiaragarassino deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT aliyanhusain deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT huihuali deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT robertgrossman deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT nicoleacipriani deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation AT alexandertpearson deeplearninggeneratessyntheticcancerhistologyforexplainabilityandeducation |