机器学习辅助肿瘤诊断

Machine learning aided tumor diagnosis

近年来,高质量数字病理切片在病理诊断中的应用改变了传统的病理阅片方式,大量的定量分析算法应运而生,其中机器深度学习算法对大数据样本分析的能力普遍强于其他算法,在病理切片分析中表现出巨大潜力,取得了显著成果。机器分析病理图片的过程为提取病理图片特征并据此进行分类,从而判断肿瘤的性质、分级和预后等,可以提升病理诊断的客观性和准确率。目前,机器学习辅助病理图片分析应用较为成熟的领域包括乳腺癌的诊断和预后、皮肤癌的性质判断、肺癌的诊断和预后以及前列腺癌和子宫颈上皮内瘤变的分级等。本文就上述领域的进展进行综述并展开讨论。

In recent years, the application of high-quality digital slides in pathological diagnosis has changed the traditional reading methods. As a result, a large number of quantitative analysis algorithms come into being. Among them, the machine deep learning algorithm has outperformed other algorithms in the analysis of large data, showing great potential in the analysis of pathological sections. The process of pathological image analysis based on machine learning consists of feature extraction and classification to determine the nature, grading and prognosis of tumors, which can improve the objectivity and accuracy of pathological diagnosis. At present, those fields in which machine learning aided pathological image analysis presents as a relatively mature diagnostic tool include the diagnosis and prognosis of breast cancer, the determination of the nature of skin cancer, the diagnosis and prognosis of lung cancer, and the grading of prostate cancer and cervical intraepithelial neoplasia. In this paper, the research progresses in these fields are reviewed and discussed.