Liu Jinshuo, Yang Pengbo, Wang Zihao, Ke Yang, Wang Kangxin, Ainivar Kaisai, Chu Kai, Yang Xinling
Xinjiang Medical University,No.567, Shangde North Road, Shuimogou District,Urumqi City, Xinjiang Uygur Autonomous Region, Urumchi 830000, PR China; The Second Affiliated Hospital of Xinjiang Medical University, No. 38, North Second Lane, Nanhu East Road, Urumqi, Xinjiang Uygur Autonomous Region, Urumchi 830000, PR China.
Xinjiang Medical University,No.567, Shangde North Road, Shuimogou District,Urumqi City, Xinjiang Uygur Autonomous Region, Urumchi 830000, PR China.
Arch Gerontol Geriatr. 2025 Nov;138:105983. doi: 10.1016/j.archger.2025.105983. Epub 2025 Aug 6.
Research on the pathological mechanisms of Parkinson's disease (PD) reveals a significant association between nigrostriatal dopaminergic neuron degeneration and abnormal metabolic networks. However, the substantial increase in clinical heterogeneity complicates early diagnosis and subtyping. Significant progress has been made in recent years through multimodal neuroimaging studies. Current research utilizing ¹⁸F-FDG PET/CT combined with multimodal neuroimaging techniques has systematically revealed characteristic PD metabolic patterns: hypermetabolism in the putamen is significantly associated with motor symptoms, while hypometabolism in the frontal/parietal lobes is closely linked to cognitive decline. In contrast, multiple system atrophy (MSA) manifests as hypometabolism in the cerebellar-pontine region, and progressive supranuclear palsy (PSP) exhibits functional dissociation within the midbrain-frontal network. Recent advancements demonstrate that artificial intelligence (AI)-driven multimodal radiomics analysis, by integrating metabolic and structural features, significantly enhances PD subtyping classification efficacy and differential diagnostic accuracy. Furthermore, multiple studies have confirmed that metabolic abnormalities precede morphological changes, suggesting their potential as early biomarkers. Collectively, current evidence indicates that distinctive metabolic network patterns-such as the contrast in cerebellar metabolism between PD and MSA-coupled with AI-driven deep mining of multimodal data, provide a critical foundation for the precise subtyping of neurodegenerative diseases and personalized therapeutic interventions.Future research should focus on establishing multicenter data-sharing frameworks and standardized metabolic databases. These initiatives are essential to further optimize the generalization capability of AI models and accelerate the clinical translation of metabolic biomarkers.
帕金森病（PD）病理机制的研究表明黑质纹状体多巴胺能神经元变性与异常代谢网络之间存在显著关联。然而，临床异质性的大幅增加使早期诊断和亚型分类变得复杂。近年来，通过多模态神经影像学研究取得了重大进展。目前利用¹⁸F-FDG PET/CT结合多模态神经影像学技术的研究系统地揭示了PD的特征性代谢模式：壳核代谢亢进与运动症状显著相关，而额叶/顶叶代谢减退与认知功能下降密切相关。相比之下，多系统萎缩（MSA）表现为脑桥小脑区域代谢减退，进行性核上性麻痹（PSP）表现为中脑-额叶网络内的功能分离。最近的进展表明，人工智能（AI）驱动的多模态放射组学分析通过整合代谢和结构特征，显著提高了PD亚型分类的效能和鉴别诊断的准确性。此外，多项研究证实代谢异常先于形态学改变，表明它们作为早期生物标志物的潜力。总体而言，目前的证据表明，独特的代谢网络模式——如PD和MSA在小脑代谢方面的差异——再加上AI驱动的多模态数据深度挖掘，为神经退行性疾病的精确亚型分类和个性化治疗干预提供了关键基础。未来的研究应专注于建立多中心数据共享框架和标准化代谢数据库。这些举措对于进一步优化AI模型的泛化能力和加速代谢生物标志物的临床转化至关重要。
