Impact of blurs on machine-learning aided digital pathology image analysis

作     者：Maki Ogura Tomoharu Kiyuna Hiroshi Yoshida 

作者机构：Digital Healthcare Business Development OfficeNEC CorporationTokyo 108-8001Japan Department of Diagnostic PathologyNational Cancer Center HospitalTokyo 104-0045Japan 

出 版 物：《Artificial Intelligence in Cancer》 (癌症中的人工智能（英文）)

年 卷 期：2020年第1卷第1期

页      面：31-38页

学科分类：08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

主　　题：Machine learning Digital pathology image Automated image analysis Blur Color Reproducibility 

摘      要：BACKGROUND Digital pathology image(DPI)analysis has been developed by machine learning(ML)***,little attention has been paid to the reproducibility of ML-based histological classification in heterochronously obtained DPIs of the same hematoxylin and eosin(HE)*** To elucidate the frequency and preventable causes of discordant classification results of DPI analysis using ML for the heterochronously obtained *** We created paired DPIs by scanning 298 HE stained slides containing 584 tissues twice with a virtual slide *** paired DPIs were analyzed by our MLaided classification *** defined non-flipped and flipped groups as the paired DPIs with concordant and discordant classification results,*** compared differences in color and blur between the non-flipped and flipped groups by L1-norm and a blur index,*** We observed discordant classification results in 23.1%of the paired DPIs obtained by two independent scans of the same microscope *** detected no significant difference in the L1-norm of each color channel between the two groups;however,the flipped group showed a significantly higher blur index than the non-flipped *** Our results suggest that differences in the blur-not the color-of the paired DPIs may cause discordant classification *** ML-aided classification model for DPI should be tested for this potential cause of the reduced reproducibility of the *** a future study,a slide scanner and/or a preprocessing method of minimizing DPI blur should be developed.