一般注記Cancer is a significant health problems around the world. Pathology is a microscopic study of tissue structure to examine disease. The nuclear properties have significant representation for cancer diagnosis. Except traditional color image, this study investigate multispectral images and 3D images to analyze cancer. These imaging systems may provide more information than traditional color image and obtain an additional option to pathologists. The objective of this study is to implement a computational method to describe nuclear characteristics of pathology images and classify cancer in hepatocellular carcinoma and thyroid follicular lesion. The system mainly contains two parts. First, nuclei segmentation is performed based on pixel-based classification. Lastly, Bag-of-visual-word model and random forest classifier are investigated in classification step.
identifier:oai:t2r2.star.titech.ac.jp:50310575
一次資料へのリンクURLhttp://t2r2.star.titech.ac.jp/rrws/file/CTT100708142/ATD100000413/thesis_12D53292_OranitBoonsiri.pdf （fulltext）
連携機関・データベース国立情報学研究所 : 学術機関リポジトリデータベース（IRDB）（機関リポジトリ）