几种机器学习方法在黑色素瘤计算机辅助诊断中的性能比较

黑色素瘤的计算机辅助诊断是基于激光共聚焦扫描显微镜（CLSM）皮肤图像纹理特征, 并引入机器学习的技术, 为临床应用研发的一种能够准确、有效地识别在体恶性黑色素瘤新医学诊断方法, 将常用的基于机器学习的ID3、分类与回归树（CART）和AdaBoost三种算法应用于良恶性黑色素瘤图像的特征识别, 并对各种学习方法的性能进行比较。实验结果表明, AdaBoost算法具有较好的分类识别性能, 不但提高了恶性黑色素瘤早期诊断的准确度, 降低了良性黑色素瘤的误诊率, 而且为临床上早期发现和诊断提供了客观依据。

Performance comparison of several machine learning methods forcomputer-aided diagnosis of melanoma

Computer-aided diagnosis of melanoma is a new medical diagnosis method, based on the texture features of confocal laser scanning microscopy (CLSM) skin images, it uses the machine learning methods, for the accuracy of diagnosis of malignant melanoma and valid identification of malignant melanoma. The paper applied currently popular machine learning methods including the ID3 algorithm, classification and regression tree(CART) algorithm, AdaBoost algorithm, to features detection between benign and malignant melanoma images. And then it compared the performance of each learning method. Experimental results show that the AdaBoost algorithm has the best discrimination performance. It improves the early malignant melanoma diagnosis accuracy, and reduces the misdiagnosis rate of benign common nevi, also it provides an objective basis for clinical diagnosis and early detection.