arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2305.03056eess.IVq-bio.NC
收藏

通过 XAI 利用 MRI 的多种脑测量指标进行阿尔茨海默病分类中的生物标志物研究

Biomarker Investigation using Multiple Brain Measures from MRI through XAI in Alzheimer's Disease Classification

Davide Coluzzi, Valentina Bordin, Massimo Walter Rivolta, Igor Fortel, Liang Zhang, Alex Leow, Giuseppe Baselli

AI总结:

该研究训练 ResNet18 与 BC-GCN-SE 分别分析结构 MRI 和脑连接矩阵,并用 Grad-CAM 验证其关注内侧颞叶等病理区域,证明多模态融合可提升 AD 分类可靠性。

摘要:

阿尔茨海默病(AD)是全球导致痴呆的首要原因,这是一种进行性损害的疾病,会造成高住院率和高死亡率。为优化诊断流程,已有大量工作致力于开发用于自动 AD 分类的深度学习(DL)方法。然而,其典型的黑箱特征导致临床框架内信任度低、使用稀少。在本研究中,我们提出两个最先进的 DL 模型,分别在结构 MRI(ResNet18)和由弥散数据派生的脑连接矩阵(BC-GCN-SE)上训练。模型首先依据分类准确率进行评估。随后,使用可解释人工智能(XAI)方法(Grad-CAM)分析结果,以衡量两个模型的可解释性水平。XAI 评估在 132 个脑区上进行,这些脑区提取自 Harvard-Oxford 与 AAL 脑图谱的组合,并与公认的病理区域进行比较,以衡量其与领域知识的符合程度。结果显示,与现有文献相比,分类性能可接受(ResNet18:TPR中位数=0.817,TNR中位数=0.816;BC-GCN-SE:TPR中位数=0.703,TNR中位数=0.738)。通过统计检验(p < 0.05)和对最相关脑区(前 15%)的排序评估,Grad-CAM 揭示 ResNet18 与 BC-GCN-SE 模型均涉及目标脑区:内侧颞叶和默认模式网络。所得可解释性并非没有局限。尽管如此,结果表明结合不同成像模态可能提升分类性能和模型可靠性。这有可能增强人们对 DL 模型的信任,并促进其作为辅助诊断工具的广泛应用。

英文摘要:

Alzheimer's Disease (AD) is the world leading cause of dementia, a progressively impairing condition leading to high hospitalization rates and mortality. To optimize the diagnostic process, numerous efforts have been directed towards the development of deep learning approaches (DL) for the automatic AD classification. However, their typical black box outline has led to low trust and scarce usage within clinical frameworks. In this work, we propose two state-of-the art DL models, trained respectively on structural MRI (ResNet18) and brain connectivity matrixes (BC-GCN-SE) derived from diffusion data. The models were initially evaluated in terms of classification accuracy. Then, results were analyzed using an Explainable Artificial Intelligence (XAI) approach (Grad-CAM) to measure the level of interpretability of both models. The XAI assessment was conducted across 132 brain parcels, extracted from a combination of the Harvard-Oxford and AAL brain atlases, and compared to well-known pathological regions to measure adherence to domain knowledge. Results highlighted acceptable classification performance as compared to the existing literature (ResNet18: TPRmedian = 0.817, TNRmedian = 0.816; BC-GCN-SE: TPRmedian = 0.703, TNRmedian = 0.738). As evaluated through a statistical test (p < 0.05) and ranking of the most relevant parcels (first 15%), Grad-CAM revealed the involvement of target brain areas for both the ResNet18 and BC-GCN-SE models: the medial temporal lobe and the default mode network. The obtained interpretabilities were not without limitations. Nevertheless, results suggested that combining different imaging modalities may result in increased classification performance and model reliability. This could potentially boost the confidence laid in DL models and favor their wide applicability as aid diagnostic tools.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑