Generalized N-Dimensional Sparse Coding and Its Application to Computational Anatomy Models

2014 Fiscal Year Final Research Report

• Project/Area Number: 24300076
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Partial Multi-year Fund
• Section: 一般
• Research Field: Perception information processing/Intelligent robotics
• Research Institution: Ritsumeikan University
• Principal Investigator: CHEN YAN WEI 立命館大学, 情報理工学部, 教授 (60236841)
• Co-Investigator(Kenkyū-buntansha): TANAKA T, Hiromi 立命館大学, 情報理工学部, 教授 (10268154) ; HAN Xian-Hua 立命館大学, 立命館グローバルイノベーション研究機構, 准教授 (60469195) ; SATO Yoshinobu 奈良先端科学技術大学院大学, 情報科学研究科, 教授 (70243219) ; FURUKAWA Akira 首都大学東京, 人間健康科学研究科, 教授 (80199421) ; MORIKAWA Shigehiro 滋賀医科大学, 医学部, 教授 (60220042) ; TATEYAMA Tomoko 立命館大学, 情報理工学部, 助手 (90550153)
• Project Period (FY): 2012-04-01 – 2015-03-31
• Keywords: 多重線形 / 腹部複数臓器 / スパース / Low-rank / 局所解析 / ボリューム / 医用画像 / テンプレートマッチング

Outline of Final Research Achievements

Recently, sparse coding is a hot topic for efficient data representation, and has been widely used in computer vision field. In this project, we proposed a generalized ND sparse coding based on multi-linear algebra, for direct analysis of multi-dimensional data without unfolding process. Experiments results on noise reduction demonstrated that the proposed method can achieve better results compared with the conventional sparse coding. We also proposed a framework for local morphological analysis (local statistical shape models) of 3D organs based on sparse and low rank matrix decomposition and applied our proposed method to computer-aided diagnostics of liver cirrhosis. The local abnormal regions can be detected by estimating the sparse components. The norm of the sparse components can be used as a measure for classification of the normal and abnormal livers. The classification accuracy by our proposed method is improved to 95%.

Free Research Field

医用画像