Jasodanand Varuna H, Kowshik Sahana S, Puducheri Shreyas, Romano Michael F, Xu Lingyi, Au Rhoda, Kolachalama Vijaya B
Department of Medicine, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA.
Faculty of Computing & Data Sciences, Boston University, Boston, MA, USA.
Nat Commun. 2025 Aug 11;16(1):7407. doi: 10.1038/s41467-025-62590-4.
Alzheimer's disease (AD) diagnosis hinges on detecting amyloid beta (Aβ) plaques and neurofibrillary tau (τ) tangles, typically assessed using PET imaging. While accurate, these modalities are expensive and not widely accessible, limiting their utility in routine clinical practice. Here, we present a multimodal computational framework that integrates data from seven distinct cohorts comprising 12, 185 participants to estimate individual PET profiles using more readily available neurological assessments. Our approach achieved an AUROC of 0.79 and 0.84 in classifying Aβ and τ status, respectively. Predicted PET status was consistent with various biomarker profiles and postmortem pathology, and model-identified regional brain volumes aligned with known spatial patterns of tau deposition. This approach can support scalable pre-screening of candidates for anti-amyloid therapies and clinical trials targeting Aβ and τ, offering a practical alternative to direct PET imaging.
阿尔茨海默病（AD）的诊断依赖于检测淀粉样β（Aβ）斑块和神经纤维缠结tau（τ），通常使用正电子发射断层扫描（PET）成像进行评估。虽然这些方法准确，但成本高昂且无法广泛应用，限制了它们在常规临床实践中的效用。在此，我们提出了一个多模态计算框架，该框架整合了来自七个不同队列、共12185名参与者的数据，以使用更容易获得的神经学评估来估计个体的PET图像。我们的方法在分类Aβ和τ状态时的受试者工作特征曲线下面积（AUROC）分别达到了0.79和0.84。预测的PET状态与各种生物标志物特征及死后病理学结果一致，且模型识别出的区域脑容量与已知的tau沉积空间模式相符。这种方法可以支持对淀粉样蛋白治疗候选者以及针对Aβ和τ的临床试验进行可扩展的预筛选，为直接PET成像提供了一种实用的替代方案。
