自闭症谱系障碍（ASD）是一类神经发育障碍性疾病，其核心临床表现为社交行为障碍、语言沟通困难以及重复刻板行为。早期诊断并进行早期行为干预，是目前能改善患者临床症状且减少终身残疾的主要措施。但早期诊断困难，因为目前所采用的行为量表，比如ABC（异常行为检查表）等，在儿童生命早期阶段，如2岁左右，很难准确地反应出被试者的社交行为或者语言交流能力；而且，其判断也非常依赖于医生的经验，具有一定的主观性。近年来，随着脑影像技术，比如核磁共振（MRI），特别是功能性核磁共振（fMRI），以及人工智能技术（AI）的飞速发展，为ASD的早期诊断，从无创、客观、以及神经活动异常-行为异常偶联的角度，展示了一种新的机遇。本文比较系统地介绍了ASD的临床基础与诊断挑战，并详细阐述了AI在ASD早期诊断中的数据来源、核心算法和典型研究案例。通过分析当前技术面临的挑战与局限性，提出未来研究方向与临床转化路径，以期进一步加强AI在ASD早期诊断中的临床使用研究，使之最终能高效的服务于临床。

   Autism spectrum disorder（ASD）is a group of neurodevelopmental disorders clinically characterized by deficits in social interaction and verbal communication，along with repetitive behaviors．Early diagnoses，followed by early behavioral interventions，are currently effective strategies to attenuate clinical symptoms and improve patient outcomes in ASD; however，early diagnosis remains challenging at present．Current diagnostic tools，such as the Aberrant Behavior Checklist（ABC），have difficulty in accurately evaluating social engagement and verbal communication in early childhood（e.g. ，at 2 years of age）．Furthermore，such evaluations may be somewhat subjective，as they rely，at least partially，on clinicians’ experience．Recently，with the rapid development of brain imaging techniques（e.g. ，MRI，especially fMRI）and artificial intelligence（AI），a significant opportunity has emerged to develop novel early-diagnostic tools for ASD．These tools exhibit clear advantages，including being non-invasive，more objective，and focusing on the coupling of abnormalities in both neuronal activity and behavioral responses．Here，we systematically introduce the clinical features of ASD，discuss the challenges in early diagnosis，and review the current status of AI-driven diagnostic tool development in detail，including data acquisition methods，computational architectures，and typical clinical applications．Through a detailed analysis of the technical challenges，advantages，and limitations of these tools，we propose future research directions，aiming to eventually establish highly valuable AI-based diagnostic approaches for clinical use in ASD care．

       本文系统探讨了人工智能（AI）技术在模拟医学培训中的应用现状、优势与挑战。AI通过虚拟患者系统、手术模拟评估、医学影像诊断培训及结构化报告优化四大核心场景,显著提升培训的智能化与个性化水平。研究表明,AI驱动的实时反馈机制（如手术技能评估系统）在随机对照试验中表现优于传统专家指导, 并具备大规模推广潜力, 可降低人力成本。然而,技术仍面临算法透明性、数据隐私伦理及临床转化效果验证等挑战。未来需深化跨学科合作, 结合增强现实（AR）等技术创新, 构建全球资源共享的智能认证体系, 推动医学教育范式转型。

       This review summarizes and discusses the application status, advantages,and challenges of artificial intelligence（AI）technology in simulated medical training．AI significantly enhances the intelligence and personalization of training through four core scenarios：virtual patient systems, surgical simulation assessment, medical imaging diagnosis training, and structured reporting optimization．Researches demonstrates that AI-driven real-time feedback mechanisms（e.g., surgical skill assessment systems）outperform traditional expert guidance in randomized controlled trials（P<0.001）and exhibit potential for large-scale implementation to reduce labor costs．However, challenges remain regarding algorithmic transparency, data privacy ethics, and clinical translation validation．Future efforts require deepened interdisciplinary collaboration, integration with innovations like augmented reality, and the establishment of a globally shared intelligent certification system to advance the transformation of medical education paradigms．

乳腺癌是女性最常见的原发恶性肿瘤之一。目前,通过采用综合治疗手段,包括手术、新辅助治疗、辅助放化疗等多种手段,乳腺癌已成为疗效最佳的实体肿瘤之一。其中,新辅助治疗（NAT）包括新辅助化疗、靶向治疗和内分泌治疗,目的是使肿瘤降期、保乳、保腋窝,并可观察药物敏感性,是当前乳腺癌综合治疗中非常重要的组成部分,其治疗疗效对患者手术方式的选择和预后至关重要。尽管病理学评价在评估NAT疗效方面被公认为金标准,但其局限性在于只能通过有创手段在治疗后进行,无法在治疗前对患者做出准确预测。磁共振成像（MRI）作为一项广泛使用的乳腺成像技术,在评估NAT疗效中扮演着关键角色。近年来,人工智能技术,尤其是影像组学（Radiomics）和深度学习（Deep Learning）,在医学影像分析领域取得了显著进展。这些技术能够从医学图像中提取大量肉眼难以识别的特征,揭示病变内部的微观结构和生物学行为,全面反映肿瘤的异质性,这不仅有助于临床医生更准确地区分良、恶性肿瘤,还能对恶性肿瘤的预后进行更为精确的评估。本文系统综述了近年来基于MRI的人工智能技术在乳腺癌新辅助治疗疗效评估中的应用研究,旨在促进人工智能在NAT临床实践中的应用和发展,为乳腺癌NAT治疗策略的优化和个性化医疗的实现提供科学依据。

Breast cancer is one of the most common primary malignant tumors in women．Currently,breast cancer has become one of the most effective solid tumors by using comprehensive treatment methods,including surgery,neoadjuvant therapy,adjuvant radiotherapy and chemotherapy．Among them,neoadjuvant therapy（NAT）,including neoadjuvant chemotherapy,targeted therapy and endocrine therapy,is a very important part of the current comprehensive treatment of breast cancer．It aims to reduce the tumor stage,preserve the breast,preserve the armpit,and observe the drug sensitivity．Its therapeutic effect is crucial to the choice of surgical methods and prognosis of patients．Although pathological evaluation is recognized as the gold standard in evaluating the response to NAT,its limitation is that it can only be performed after treatment by invasive means,and cannot accurately predict response before treatment．As a widely used breast imaging technology,magnetic resonance imaging（MRI）plays a key role in evaluating the response to NAT．However,traditional MRI evaluation methods are limited by the individual differences of interobserver and the low repeatability of evaluation results,which affects the accuracy of efficacy evaluation to a certain extent．In recent years,artificial intelligence technology,especially radiomics and deep learning,has made significant progress in the field of medical image analysis．These techniques can extract a large number of features that are difficult to be recognized by the naked eye from medical images,reveal the internal microstructure and biological behavior of the lesion,and fully reflect the heterogeneity of the tumor．This not only helps clinicians to distinguish benign and malignant tumors more accurately,but also makes a more accurate assessment of the prognosis of malignant tumors．This article reviews the application and progress of MRI-based artificial intelligence technology in evaluating the response to neoadjuvant therapy for breast cancer in the past five years,aiming to promote the application and development of artificial intelligence in NAT clinical practice,and provide a scientific basis for the optimization of NAT treatment strategy and the realization of personalized medicine for breast cancer．

       人工智能（AI）这一新兴技术的出现和应用给炎症性肠病（IBD）的诊断带来了巨大的变革。越来越多的研究着手于开发基于机器学习（ML）和深度学习（DL）的诊断模型，并获得了良好的诊断性能，尤其是在IBD的图像诊断，卷积神经网络（CNN）等模型由于其出色的图像分析能力，在内镜检查和组织病理检查等方面具有十分可观的发展前景。近年来AI诊断模型的应用越发广泛，但与此同时，关于算法、数据库及其应用方面仍存在一些难以忽视的局限性。本文将主要就图像识别方面对AI在IBD诊断中的应用进行综述，以期为IBD精准图像诊断领域下步研究提供参考。

        As an emerging technology，artificial intelligence（AI）has brought great changes to the precise diagnosis of inflammatory bowel disease（IBD）．More and more  researches have developed diagnostic models which are based on machine learning（ML）and deep learning（DL）and obtained satisfactory diagnostic performance．Especially in the image  diagnosis of IBD，convolutional neural network（CNN）and  other models  have  considerable  development  prospects in  endoscopy  and histopathology due to their excellent image analysis capabilities．In recent years，the application of AI diagnostic models has become more and more widespread，but at the same time，there are still some limitations about algorithms，databases and their applications that cannot be ignored．This review mainly focused on the application of AI in IBD diagnosis from the aspect of image recognition，to provide a reference for IBD diagnosis towards precision medicine．