摘要
乳腺癌分子分型对乳腺癌的治疗具有决定性的参考作用,传统的分型方法有创且可能存在假阳性问题,而已有的基于影像学的分型方法准确率较低。本文提出一种利用迁移学习提取特征并结合支持向量机的分型预测方法,对乳腺癌PET/CT标记图像进行融合和归一化,再使用Xception迁移学习网络进行特征提取,最后使用支持向量机进行分类实现分型。对样本测试集进行性能评估表明,Xception+SVM模型的准确率达到0.687,AUC为0.787,优于现有基于影像学的方法,验证了本文方法的有效性。
Molecular typing of breast cancer plays a decisive reference role in the treatment of breast cancer.Traditional typing methods are invasive and may have the problem of false positive,and the accuracy of existing imaging-based typing methods is unsatisfactory.Therefore,a feature extraction method using transfer learning,combined with typing prediction method of support vector machine(SVM),is proposed in the study.After the fusion and normalization of breast cancer PET/CT marker images,Xception transfer learning network is used for feature extraction,and finally SVM is used for classification and typing.The performance evaluation on test set shows that the accuracy of Xception+SVM model is 0.687,and the AUC is 0.787,better than the existing imaging-based methods,which verifies the effectiveness of the proposed method.
作者
赵清一
林勇
ZHAO Qingyi;LIN Yong(School of Health Science and Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)
出处
《中国医学物理学杂志》
CSCD
2022年第5期635-639,共5页
Chinese Journal of Medical Physics
基金
国家自然科学基金(81801797)。
