乳腺癌针吸细胞形态定量的人工神经网络诊断模型的建立及应用研究  

The establishment and application study based on the artificial neural network diagnostic model for differentiate diagnosis mammary carcinoma with the cell form stereological quantitative parameters by fine needle aspiration cytology

作　　者：龚平[1] 郭华雄[1] 王文清[1] 李春燕[1] 

机构地区：[1]华中科技大学同济医学院附属荆州医院病理科,434020

出　　处：《中国体视学与图像分析》2007年第3期198-201,共4页Chinese Journal of Stereology and Image Analysis

摘　　要：目的建立乳腺癌针吸细胞形态定量参数的人工神经网络诊断模型,并验证其在辅助FNA诊断乳腺癌的价值。方法利用MPIAS-2000系统对60例乳腺癌及30例乳腺良性病变的针吸细胞学涂片进行形态定量测定,对获得的29项形态参数进行人工神经网络建模分析,并用盲法对其鉴别诊断能力进行评价。结果所建立的网络模型经过14次训练后即可达到误差要求,诊断模型对乳腺癌及乳腺良性病变的诊断正确率为100%,其特异性和敏感性均为100%。结论乳腺良恶性病变的针吸细胞学涂片进行ANN分析所建立的诊断模型,对乳腺癌及良性病病变的鉴别诊断具有较高的应用价值,为辅助针吸细胞学诊断乳腺良恶性病变提供了新的思路。Objective To establish diagnostic models of cell form parameters in mammary carcinoma FNA＇s smears by an artificial neural network （ANN） methods, and apply the models to discriminate the mammary cancer and benign lesion. Method The cell form quantitative parameters of 60 breast cancer cases and 30 benign lesion cases were analysis by MPIAS-2000,all of cases were provided pathologically. The ANN diagnosis models of cancer and benign lesion were established combined with the 29 cell quantitative parameters. The blind-test set were used to confirm the models. Results The ANN model＇s meet performance goal by 14 times trainlm. ANN model of cell form quantitative parameters bad an accuracy and specialty of 100% for differentiation of breast carcinoma and benign lesion. Conclusion Method of erecting models based on cell form quantitative parameters with ANN could identify breast carcinoma and benign lesion. It may be valuable and new idea for FNA＇c in the differentiation of breast diseases.

关 键 词：针吸细胞学 乳腺癌 细胞形态定量参数 人工神经网络 诊断模型 

分 类 号：R737.9[医药卫生—肿瘤]