Application of text mining techniques and image analysis in pathology diagnosis

2017 Fiscal Year Final Research Report

• Project/Area Number: 15K08386
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Human pathology
• Research Institution: Kitasato University
• Principal Investigator: Hara Atsuko 北里大学, 医学部, 准教授 (10276123)
• Co-Investigator(Kenkyū-buntansha): 三枝 信 北里大学, 医学部, 教授 (00265711)
• Research Collaborator: Ishibashi Yuichi (株)スタットラボ, 代表
• Project Period (FY): 2015-04-01 – 2018-03-31
• Keywords: 病理診断 / テキストマイニング解析 / 画像解析 / 深層学習法 / 診断支援システム

Outline of Final Research Achievements

We have developed a pathological information data base system and a diagnostic processing model with the use of pathology reports and images of digitized specimens. We first described an algorithm to enable the numeric transformation of pathology reports (breast, gastrointestinal, and esophageal disease) using both text mining and statistical analysis, and we attempted to develop an inspection program that point out the medical inconsistency and/or correct the erroneous description. Images of digitized specimen (breast disease) were divided into many small images, then Wavelet transformation and cluster analysis was performed. They were taken as training data, and identified by pattern recognition by the deep learning method or the K-nearest neighbor method. The result of this identification was used as a feature vector for a specimen image. Similar images were retrieved by comparing the feature vectors of the targeted images and the specimen images in the database.

Free Research Field

病理学