血液细胞形态识别研究

Study on recognition of blood cell morphology

血液细胞形态识别在白血病检查与诊断中起重要作用,细胞形态特征提取对于识别性能有很大影响,通过提取反映血液细胞形态几何、纹理及对几何变形不敏感的二维不变矩构成特征矢量,基于SVM分类器研究骨髓血液细胞形态的识别,特别研究了急性淋巴细胞白血病L3型细胞与骨髓正常细胞的识别,得到正确识别率为97%.

Recognition of blood cell morphology plays an important role in the diagnosis of leukemia. The cell morphology feature extraction has significant impacts on the property identification. In this paper, the blood cell geometric shape, texture and two-dimensional moment invari- ants which are not sensitive to the geometric deformation are extracted to form the feature vector. The identification of bone marrow blood call morphology based on SVM is studied. Specially, the recognition of Acute lymphocytic leukemia L3 style＇s cell and normal cell from bone marrow are studied. Using the feature of blood cell morphology and texture, correct recognition rate achieves 97 percent.