人工智能在乳腺癌组织病理诊断领域的研究进展

Research progress of artificial intelligence in the field of breast cancer histopathological diagnosis

人工智能可应用于乳腺癌原发肿瘤的良恶性判别、病理分级以及有丝分裂和肿瘤相关间质的鉴别。同时,它也在淋巴结转移诊断上达到病理学家的水平。在分子病理诊断上,人工智能不仅可用于ER、HER2免疫组织化学染色结果的判读,还可以用于推测PAM50分型及预后评分。目前,此类研究主要局限于技术领域,以准确率为主要的研究目标,缺乏对固定临床场景、临床要求下人工智能在病理诊断方面的转化性研究。此类研究应更注重计算结果的及时性、稳定性以及各种计算方法的性价比。今后,在此方面的人工智能研究,需要有人工智能的企业在病理医师给出的临床场景和临床要求下对现有技术进行整合创新,提出符合病理诊断需求的人工智能产品,并对其进行大样本、多中心、前瞻性试验,进一步推动人工智能在乳腺癌病理诊断实践中的落地。

Artificial intelligence can be used to distinguish benign and malignant primary breast cancer, rate pathological grading, identify mitosis and tumor-related stroma. Meanwhile, it has reached the level of pathologist on the diagnosis of lymph node metastasis. In molecular pathological diagnosis, artificial intelligence can not only be used to interpret the results of ER and HER2 immunohistochemical staining, but also to infer PAM50 typing and prognostic score. At present, such research is mainly limited to the technical field, with accuracy as the main research goal, and there is a lack of transformative research on pathological diagnosis by artificial intelligence under fixed clinical scenarios and clinical requirements. Such research should pay more attention to the timeliness and stability of calculation results and the cost-effectiveness of various calculation methods. In the future, artificial intelligence research in this area requires enterprise to carry out technological integration and innovation under the clinical scenarios and clinical requirements given by pathologists, propose artificial intelligence products that meet the needs of pathological diagnosis, and carry out large-sample, multi-center and prospective trials to further promote the application of artificial intelligence in the practice of breast cancer pathological diagnosis.