Park Chang-Hyun, Yoon Uicheul, Lee Phil Hyu, Kim Jinna, Lee Seung-Koo, Shin Na-Young
Division of Artificial Intelligence and Software, College of Artificial Intelligence, Ewha Womans University, Seoul, Republic of Korea.
Department of Biomedical Engineering, College of Bio and Medical Sciences, Daegu Catholic University, Gyeongsan, Republic of Korea.
Front Aging Neurosci. 2025 Jul 16;17:1579326. doi: 10.3389/fnagi.2025.1579326. eCollection 2025.
During the prodromal stage of Parkinson's disease (PD), brain structural alterations precede clinical diagnosis and offer opportunities for early detection. We investigated whether combining clinical non-motor markers with an MRI-based brain structural marker could enhance predictive performance for PD conversion.
Individuals with prodromal symptoms ( = 46, 63.5 ± 7.6 years, 24 males) were selected from the Parkinson's Progression Markers Initiative dataset. We developed a machine learning classifier to identify individuals with brain structural patterns similar to PD based on cortical thickness and white matter integrity. Its predictive performance for PD conversion was assessed alone and combined with clinical non-motor markers such as rapid eye movement sleep behavior disorder and olfactory dysfunction.
Six individuals converted to PD within 4 years. The MRI marker classified 21 individuals as having PD-like brain patterns, including all six converters. When combined with olfactory dysfunction, the approach achieved optimal performance with 100% sensitivity, 80% specificity, and 90% balanced accuracy, outperforming individual markers and other combinations.
MRI-quantified brain structural similarity to PD, particularly when combined with olfactory assessment, significantly enhances prediction of PD conversion in individuals with prodromal symptoms. This accessible, multimodal approach could facilitate early identification of high-risk individuals for targeted interventions and clinical trials.
在帕金森病（PD）的前驱期，脑结构改变先于临床诊断出现，为早期检测提供了机会。我们研究了将临床非运动标志物与基于MRI的脑结构标志物相结合是否能提高对PD转化的预测性能。
从帕金森病进展标志物计划数据集中选取有前驱症状的个体（n = 46，63.5±7.6岁，24名男性）。我们开发了一种机器学习分类器，根据皮质厚度和白质完整性来识别具有与PD相似脑结构模式的个体。分别评估其对PD转化的预测性能，以及与快速眼动睡眠行为障碍和嗅觉功能障碍等临床非运动标志物相结合时的预测性能。
4年内有6名个体转化为PD。MRI标志物将21名个体分类为具有类似PD的脑模式，其中包括所有6名转化者。当与嗅觉功能障碍相结合时，该方法达到了最佳性能，灵敏度为100%，特异度为80%，平衡准确度为90%，优于单个标志物及其他组合。
MRI量化的与PD相似的脑结构，特别是与嗅觉评估相结合时，能显著提高对有前驱症状个体PD转化的预测。这种易于采用的多模态方法有助于早期识别高危个体，以进行有针对性的干预和临床试验。
