人工智能在乳腺癌诊断及预测模型中的应用

Artificial Intelligence in Breast Cancer Diagnostic and Predictive Modeling

人工智能(AI)在乳腺癌影像及病理的诊断,骨转移或淋巴转移病灶的寻找,乳腺实质组织密度的判定中,不仅能提高医生阅片的效率及准确性,还可缓解医生的工作压力,尤其是当图像质量不佳时。AI具有强大的数据分析及模拟再现能力,被广泛运用于乳腺癌的生存期和药物反应预测、临床分期、风险及预后评估等等。AI与临床检查相结合,以无创方式展示出对乳腺癌分子亚型诊断的巨大潜力;深度学习与纳米基因组学相结合,给乳腺癌的精确诊断带来了新的可能。然而,在AI数据库的建立、来源、规模及安全性,临床实践中的验收测试、质量保证、实际操作中,如何实现AI和医疗决策支持工具的泛化特性,如何处理AI造成的错误决策和管理不善,以及如何证明AI模型长期的稳定性和安全性仍是有待解决的问题。

Artificial intelligence(AI)is utilized in the diagnosis of pathology and imaging of breast cancer,in the search for lymphatic or bone metastases,and in the assessment of breast parenchymal tissue density.It improves the efficacy and accuracy of breast cancer image readings while simultaneously relieving some of workload for physicians,particularly in situations where image quality is insufficient for diagnosis.Additionally,with its excellent data analysis and simulation reproduction capabilities,AI is frequently employed insurvival prediction,therapeutic response prediction,clinical staging,risk and prognosis assessment in breast cancer.The integration of AI with clinical examination has demonstrated the great potential of non-invasive diagnosis of molecular subtypes of breast cancer;the combination of deep learning and nano genomics has brought new possibilities for accurate diagnosis of breast cancer.Nevertheless,in the establishment,source,scale,and security of AI databases,acceptance testing,quality assurance,and implementation operations in clinical practice,how to realize the generalization properties of AI and medical decision support tools,how to deal with erroneous decision making and mismanagement caused by AI,and how to prove the stability and safety of AI models in the long term remain open issues at present.