人工智能在神经肿瘤学中的应用:总体综述
本文由RANO各工作组联合完成,综述人工智能在神经影像、神经病理、检验医学、神经外科、药物研发和临床试验六大领域的应用。影像分割与定量、AI辅助图像重建、基于DNA甲基化的分类及工作流程优化最接近临床使用,影像基因组学、全切片图像分子预测等仍需更强前瞻性验证。
要点
- 影像分割与定量、AI辅助图像重建、DNA甲基化分类及流程优化最接近临床。
- 影像基因组学、全切片图像分子预测、AI引导手术靶向、液体活检解读等仍需前瞻性验证。
- 主要挑战包括大型多样数据集有限、数据采集与参考标准差异及外部验证不足。
核心结论
- AI可捕捉中枢神经系统肿瘤的异质性并提供定量、可重复的诊断、预测与预后信息,但负责实施需标准化、可重复、多机构验证与人工监督。
研究设计与样本
- RANO各工作组联合开展的总体综述,覆盖神经影像、神经病理、检验医学、神经外科、药物研发与临床试验六个领域。
主要发现
- 最接近临床的应用包括影像分割与定量、AI辅助图像重建、基于DNA甲基化的分类和工作流程优化。
- 影像基因组学、基于全切片图像的分子预测、AI引导手术靶向、液体活检解读、AI引导药物研发、数字孪生和计算机模拟试验仍有前景但需更强前瞻性验证。
临床怎么用
- 可将影像分割与定量、AI辅助图像重建、DNA甲基化分类及流程优化视为较接近临床落地的方向。
- 对影像基因组学、全切片图像分子预测、液体活检解读等尚需前瞻性验证的应用应谨慎用于临床决策。
局限性
- 大型多样数据集可及性有限,数据采集与参考标准存在差异,外部验证有限。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
Artificial intelligence (AI) is increasingly being applied across neuro-oncology domains to analyze complex, high-dimensional, multimodal data. Despite major advances in the biological understanding and classification of CNS tumors, clinical outcomes for many of them remain poor, reflecting intra- and inter-tumor heterogeneity and limitations of conventional reductionist approaches. AI offers a paradigm shift to capture this heterogeneity and provide quantitative, reproducible diagnostic, predictive, and prognostic insights. This article describes a joint effort across the Response Assessment in Neuro-Oncology (RANO) working groups to provide an overarching review of the AI use across six key domains of neuro-oncology: neuroimaging, neuropathology, laboratory medicine, neurosurgery, drug discovery, and clinical trials. Applications closest to clinical use include imaging segmentation and quantification, AI-assisted image reconstruction, DNA methylation-based classification, and workflow optimization. Radiogenomics, WSI-based molecular prediction, AI-guided surgical targeting, liquid biopsy interpretation, AI-guided drug discovery, digital twins, and in silico trials remain promising but require stronger prospective validation. Key challenges include limited availability of large diverse datasets, variability in data acquisition and reference standards, and limited external validation. This review provides a unifying framework for understanding the role of AI across neuro-oncology and highlights key considerations for its responsible implementation, emphasizing standardization, reproducibility, multi-institutional validation, and human oversight. Artificial intelligence (AI) is increasingly being used in brain tumor care, for example to measure tumors on scans, classify tumor samples, inform surgical planning, and support the development of new treatments. Some applications are already close to routine care, while others remain experimental. This review brings together progress across the main areas of neuro-oncology and compares how ready different applications are for clinical use. It also highlights remaining challenges, including reliable validation across hospitals, safe integration into care, and continued human oversight so that AI tools can support clinical decisions and ultimately improve patient care.
引用
- Pitarch-Abaigar C, Lohmann P, Aloy P, et al. AI in Neuro-Oncology: An Overarching Review. Neuro Oncol. 2026. PMID: 42867147. doi:10.1093/neuonc/noag257
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。