Hongyi Duanmu1, Shristi Bhattarai2, Hongxiao Li3, Zhan Shi1, Fusheng Wang1,4, George Teodoro5, Keerthi Gogineni6,7,8, Preeti Subhedar8, Umay Kiraz9, Emiel A M Janssen9,10, Ritu Aneja11, Jun Kong3,12. 1. Stony Brook University, NY, USA. 2. Department of Biology, Georgia State University, Atlanta, GA, USA. 3. Department of Mathematics and Statistics and Computer Science, Georgia State University, GA, USA. 4. Department of Biomedical Informatics, Stony Brook University, NY, USA. 5. Department of Computer Science, Federal University of Minas Gerais, MG, Brazil. 6. Department of Hematology-Medical Oncology, Winship Cancer Institute, Emory University School of Medicine, GA, USA. 7. Department of Surgery, Winship Cancer Institute, Emory University School of Medicine, GA, USA. 8. Georgia Cancer Center for Excellence, Grady Health System. 9. Department of Pathology, Stavanger University Hospital, Stavanger, Norway. 10. Department of Chemistry Bioscience and Environmental Engineering, University of Stavanger, Norway. 11. School of Health Professions, University of Alabama at Birmingham, AL, USA. 12. Winship Cancer Institute Emory University, GA, USA.
Abstract
MOTIVATION: Predicting pathological complete response (pCR) to neoadjuvant chemotherapy (NAC) in triple-negative breast cancer (TNBC) patients accurately is direly needed for clinical decision making. pCR is also regarded as a strong predictor of overall survival. In this work, we propose a deep learning system to predict pCR to NAC based on serial pathology images stained with hematoxylin and eosin (H&E) and two immunohistochemical biomarkers (Ki67 and PHH3). To support human prior domain knowledge based guidance and enhance interpretability of the deep learning system, we introduce a human knowledge derived spatial attention mechanism to inform deep learning models of informative tissue areas of interest. For each patient, three serial breast tumor tissue sections from biopsy blocks were sectioned, stained in three different stains, and integrated. The resulting comprehensive attention information from the image triplets is used to guide our prediction system for prognostic tissue regions. RESULTS: The experimental dataset consists of 26,419 pathology image patches of 1,000×1,000 pixels from 73 TNBC patients treated with NAC. Image patches from randomly selected 43 patients are used as a training dataset and images patches from the rest 30 are used as a testing dataset. By the maximum voting from patch-level results, our proposed model achieves a 93% patient-level accuracy, outperforming baselines and other state-of-the-art systems, suggesting its high potential for clinical decision making. AVAILABILITY: The codes, the documentation, and example data are available on an open source at: https://github.com/jkonglab/PCR_Prediction_Serial_WSIs_biomarkers.
MOTIVATION: Predicting pathological complete response (pCR) to neoadjuvant chemotherapy (NAC) in triple-negative breast cancer (TNBC) patients accurately is direly needed for clinical decision making. pCR is also regarded as a strong predictor of overall survival. In this work, we propose a deep learning system to predict pCR to NAC based on serial pathology images stained with hematoxylin and eosin (H&E) and two immunohistochemical biomarkers (Ki67 and PHH3). To support human prior domain knowledge based guidance and enhance interpretability of the deep learning system, we introduce a human knowledge derived spatial attention mechanism to inform deep learning models of informative tissue areas of interest. For each patient, three serial breast tumor tissue sections from biopsy blocks were sectioned, stained in three different stains, and integrated. The resulting comprehensive attention information from the image triplets is used to guide our prediction system for prognostic tissue regions. RESULTS: The experimental dataset consists of 26,419 pathology image patches of 1,000×1,000 pixels from 73 TNBC patients treated with NAC. Image patches from randomly selected 43 patients are used as a training dataset and images patches from the rest 30 are used as a testing dataset. By the maximum voting from patch-level results, our proposed model achieves a 93% patient-level accuracy, outperforming baselines and other state-of-the-art systems, suggesting its high potential for clinical decision making. AVAILABILITY: The codes, the documentation, and example data are available on an open source at: https://github.com/jkonglab/PCR_Prediction_Serial_WSIs_biomarkers.
Authors: Jennifer F De Los Santos; Alan Cantor; Keith D Amos; Andres Forero; Mehra Golshan; Janet K Horton; Clifford A Hudis; Nola M Hylton; Kandace McGuire; Funda Meric-Bernstam; Ingrid M Meszoely; Rita Nanda; E Shelley Hwang Journal: Cancer Date: 2013-02-21 Impact factor: 6.860
Authors: H Raza Ali; Aliakbar Dariush; Elena Provenzano; Helen Bardwell; Jean E Abraham; Mahesh Iddawela; Anne-Laure Vallier; Louise Hiller; Janet A Dunn; Sarah J Bowden; Tamas Hickish; Karen McAdam; Stephen Houston; Mike J Irwin; Paul D P Pharoah; James D Brenton; Nicholas A Walton; Helena M Earl; Carlos Caldas Journal: Breast Cancer Res Date: 2016-02-16 Impact factor: 6.466