J022063 SVMを用いた末梢血中の白血病細胞の自動分類法に関する研究

Abstract

The purpose of the research is to build is to build a system, which enables to quantitatively evaluate the feature quantity , which inspecting engineer observed, in the smear image. This paper examined the technique of classifying normal leukocytes, and an abnormal cell (ATL cell) of (adult T-cell leukemia) in peripheral blood , using the form feature quantity and color feature quantity of a granule and nucleus .As a classification method, we employ SVM with the feature quantity of a color and morphology quantity. As a result, the normal leukocyte was able to be classified according to the reliability of ninety-five percentage. On the other hand, the classification between ATL and normal cells was not completed only color feature quantity (B<_>MAX) and form feature quantity (the number of area, the degree of circular, and outlines)