计算机辅助诊断(CADx)系统培训非肠镜专家

Computer-Aided Diagnostic (CADx) Systems Train Non-Colonoscopy Specialists

摘要: 结直肠癌筛查在早期发现和切除癌前腺瘤方面发挥着重要作用，能够显著降低结直肠癌的发病率和死亡率。在结肠镜检查中，大多数发现的息肉为小息肉或微小息肉，其恶性病变的可能性较低。因此，提出了“保留并观察”和“切除后丢弃”的策略，以减少不必要的干预，从而显著节省成本并提高诊疗效率。然而，准确地鉴别诊断性能很困难，尤其是非结肠镜专家，其诊断准确性往往难以达到实施上述策略所需的标准，导致这些策略在临床实践中未能广泛推广。因此，对经验不足的结肠镜医生进行有效培训显得至关重要。本文综述了人工智能通过计算机辅助诊断(CADx)系统提高息肉诊断准确性的潜力，并探讨了其在培训经验不足的结肠镜医生中的应用前景。回顾的多项研究表明，尽管人工智能在结直肠息肉诊断中的表现不一定优于专家内镜医师，但始终优于非专家结肠镜医师。人工智能系统可作为有效的辅助工具，帮助提高内镜医师的光学诊断准确性和信心。虽然目前CADx系统尚未广泛应用于对非专家结肠镜医师的光学诊断培训，但其展现出的潜力令人期待。

Abstract: Colorectal cancer screening plays a vital role in the early detection and resection of precancerous adenoma, which can significantly reduce the morbidity and mortality of colorectal cancer. During colonoscopy, most polyps found are microscopic polyps with a low likelihood of malignant lesions. Therefore, “keep and observe” and “discard after excision” are proposed to reduce unnecessary interventions, thereby significantly saving costs and improving the efficiency of diagnosis and treatment. However, accurate differential diagnostic performance is difficult, especially for non-colonoscopists, whose diagnostic accuracy often struggles to meet the standards required to implement the above strategies, failing these strategies to be widely generalized in clinical practice. Therefore, effective training of inexperienced colonoscopists is essential. This article reviews the potential of artificial intelligence to improve the accuracy of polyp diagnosis through computer-aided diagnosis (CADx) systems and discusses its application prospects in colonoscopists with little training experience. Multiple studies reviewed have shown that AI consistently outperforms non-expert colonoscopists in colorectal polyp diagnosis, although it does not necessarily perform better than expert endoscopists. Artificial intelligence systems can be used as an effective adjunct to help endoscopists improve the accuracy and confidence of their optical diagnosis. Although CADx systems are not yet widely used to train non-expert colonoscopists in optical diagnostics, the potential of the system is promising.