【摘要】 提出一种基于磁共振成像的脑肿瘤与健康大脑的区分方法,即从磁共振成像生成的图像中计算离散Shmaliy矩(DSMs)提取特征信息,通过逐步回归进行变量选择后采用支持向量机(SVM)方法建立DSM-SVM分类模型.对于外部独立测试集,预测准确率(89.3%)、精密度(93.8%)、F1评分(90.9%)、马修相关系数(0.78)、曲线下面积(0.96)、敏感性(88.6%)、特异性(92.9%)均表明该模型具有令人满意的性能,与几种专门的分类算法相比,具有更高的准确性和可靠性.更多还原

【Abstract】 A new approach to distinguish brain tumors from healthy ones was proposed based on their magnetic resonance imaging for the first time, in which the discrete Shmaliy moments（DSMs） were calculated from the magnetic resonance imaging scan images to extract the feature information and selected by stepwise regression, and then the support vector machine（SVM） method was employed to establish the DSMSVM classification model. For the external independent test set, the prediction accuracy was 89.3%; the precision（93.8%）, F₁-score（90.9%）, Matthew’s correlation coefficient（0.78）, area under the curve（0.96）,sensitivity（88.6%） and specificity（92.9%）, indicating that the proposed model was satisfactory. Compared with several specialized classification algorithms, the proposed DSM-SVM approach has a higher accuracy and reliability.更多还原