目的  探讨人工智能（AI）技术在阿尔茨海默病（AD）神经病理特征自动识别中的教学应用可能。 方法  本研究纳入了34例涵盖对照及不同分期（早、中、晚期）AD样病变的脑组织样本。通过六胺银（M-Ag）、磷酸化tau蛋白（p-tau）免疫组织化学染色及Gallyas银染色，分别获取β-淀粉样蛋白沉积、神经原纤维缠结及神经炎性斑块的神经病理图像。采用Visiopharm等平台构建AI模型实现3类特征的自动识别。以神经病理学专家的人工标注为“金标准”，采用交并比评估模型算法在特征分割和量化中的准确率。 结果  Visiopharm软件的U-Net算法识别准确率高达86%，与专家人工标注基本一致。 结论  AI技术在一定程度能较准确识别AD病理特征，将来有望转化为教学工具，为构建AI+的智能教学模式提供实践方案。

Objective  To explore the application of artificial intelligence (AI) techniques in the automatic identification of neuropathological features of Alzheimer’s disease (AD) for potential teaching scenario.  Methods  A total of 34 brain tissue samples were included, comprising controls and Alzheimer’s disease (AD)-like lesions across various pathological stages (early, middle, and late). To characterize the hallmark features of the disease, high-resolution neuropathological images were acquired using three specific staining protocols, methenamine silver (M-Ag) for β-amyloid deposition, phosphorylated tau (p-tau) immunohistochemistry for neurofibrillary tangles, and Gallyas silver staining for neuritic plaques. AI models were primarily constructed using platforms such as Visiopharm to achieve automated segmentation and identification of the three pathological features. The performance of the AI models was validated against manual annotations by expert neuropathologists, which served as the “gold standard”. The intersection over union metric was employed to evaluate the accuracy of the algorithms in feature segmentation and quantitative analysis.  Results  Comparison with manual annotations by neuropathology experts showed that the U-Net algorithm in Visiopharm software achieved an identification accuracy of up to 86%, which was largely consistent with expert annotations.   Conclusion  AI technology can partially accurately identify AD pathological features and holds the potential for transformation into teaching tools. The findings provide a practical foundation for developing AI-enhanced intelligent teaching models in neuropathology education.