可解释人工智能在脑肿瘤的应用概述

A Review of the Applications of Explainable Artificial Intelligence in Brain Tumors

摘要: 脑肿瘤作为一种侵袭性强、复发率高的疾病，对患者的生存率和生活质量有着显著影响。现代医学影像技术如MRI和CT为脑肿瘤的检测和诊断提供了宝贵的支持，而近年来兴起的人工智能(AI)和深度学习技术更是推动了脑肿瘤分析的自动化与智能化发展。然而，深度学习模型的“黑箱”性质限制了它们在实际医疗环境中的广泛应用，临床医生难以理解或信任这些缺乏解释性的模型。可解释人工智能(XAI)应运而生，旨在提升深度学习模型的透明性和可理解性，从而增强其在脑肿瘤检测、分类和预后预测中的应用效果。本文综述了当前XAI在脑肿瘤诊断中的应用进展，探讨了不同XAI方法的优缺点及其在临床场景中的适用性，总结了XAI在医学影像分析中的挑战与未来发展方向。

Abstract: Brain tumors are highly aggressive and have high recurrence rates, significantly affecting patients’ survival and quality of life. Modern medical imaging technologies, such as magnetic resonance imaging (MRI) and computed tomography (CT), provide valuable support for the detection and diagnosis of brain tumors. In recent years, the emergence of artificial intelligence (AI) and deep learning has further promoted the automation and intelligent analysis of brain tumors. However, the “black-box” nature of deep learning models limits their widespread adoption in real clinical environments, as clinicians find it difficult to understand or trust models that lack interpretability. Explainable artificial intelligence (XAI) has therefore emerged to enhance the transparency and interpretability of deep learning models, thereby improving their applicability in brain tumor detection, classification, and prognosis prediction. This article reviews the current progress of XAI in brain tumor diagnosis, discusses the advantages and limitations of different XAI methods and their suitability for clinical scenarios, and summarizes the challenges and future directions of XAI in medical imaging analysis.