口腔疾病防治 ›› 2022, Vol. 30 ›› Issue (2): 77-82.DOI: 10.12016/j.issn.2096-1456.2022.02.001

• 专家论坛 • 上一篇    下一篇

基于深度学习的口腔癌预后分析

陶谦(), 袁哲   

  1. 中山大学光华口腔医学院·附属口腔医院口腔颌面外科,广东省口腔医学重点实验室,广东 广州(510055)
  • 收稿日期:2021-07-27 修回日期:2021-08-21 出版日期:2022-02-20 发布日期:2021-11-25
  • 通讯作者: 陶谦
  • 作者简介:陶谦,医学博士,中山大学光华口腔医学院•附属口腔医院口腔颌面外科教授,博士生导师,主任医师。擅长口腔颌面部肿瘤、外伤和唾液腺相关疾病的诊断与治疗。主持和参加国家自然科学基金及省、市科研基金等多项研究工作。现任广东省口腔医学会口腔颌面外科专业委员会常委,《中华口腔医学研究杂志》(电子版)和《口腔疾病防治》编委。主编专著《颌骨肿瘤的诊断与治疗》,在 SCI 杂志和国内专业杂志发表论文 70余篇。
  • 基金资助:
    广东省科技计划项目(2017A020211025)

Prognostic analysis of oral cancer based on deep learning

TAO Qian(), YUAN Zhe   

  1. Department of Oral and Maxillofacial Surgery, Guanghua School of Stomatology, Hospital of Stomatology, Sun Yat-sen University, Guangdong Provincial Key Laboratory of Stomatology, Guangzhou 510055, China
  • Received:2021-07-27 Revised:2021-08-21 Online:2022-02-20 Published:2021-11-25
  • Contact: TAO Qian
  • Supported by:
    Science and Technology Project of Guangdong Province(2017A020211025)

摘要:

TNM分期作为评估口腔癌患者预后的常用方法,多年临床应用证明其存在仅局限于分析患者临床病理数据的不足,难以适应现代医学的发展。深度学习(deep learning,DL)已广泛应用在人类生活的各个方面,具备高效、智能化的数据分析优势,可以充分挖掘和分析海量的医学数据,在医疗实践中的应用方兴未艾。在口腔癌预后分析方面,深度学习能够高效处理与分析分别以淋巴细胞、灰度协调矩阵(gray level coocrrencr matrix,GLCM)和基因图谱为代表的病理、放射影像和分子图像等患者资料,并据此进行准确的预后判断;通过辅助医师优化治疗方案,深度学习可以有效改善患者的生存情况。尽管目前深度学习在口腔癌患者预后研究中存在供给数据量不足、缺乏实际临床应用等缺陷,但其已展现出良好的临床应用前景。

关键词: 口腔癌, 深度学习, 预后, TNM分期, 医学影像学, 分子图像, 算法, 模型

Abstract:

TNM(tumor node metastasis)classification is a common way to evaluate the prognosis of patients with oral cancer; however, many years of application have proven this method to be confined merely in clinical and pathological data and it cannot be adapted to the development of modern medicine. Deep learning (DL) has been widely used in various aspects of human life, has advantages for conducting efficient and intelligent searches and can explore and analyze substantial medical information well. Additionally, the application of DL to medical practice is quickly increasing. In the field of oral cancer prognosis, DL can efficiently process and analyze the pathological, radiographic and molecular data of oral cancer patients represented by lymphocytes, gray level cooccurrence matrix (GLCM) and gene maps and make accurate prognostic judgments accordingly. By assisting physicians in optimizing treatment plans, DL can effectively improve patients’ survival. Although DL lacks sufficient data and practical clinical application in prognostic studies, it has shown good clinical application prospects.

Key words: oral cancer, deep learning, prognosis, TNM classification, medical imageology, molecular image, algorithm, model

中图分类号: