AI‐driven Amyloid Pathology Detection Using Multi‐modal MRI: A Proof of Concept Study

基于多模态磁共振成像的AI驱动淀粉样蛋白病理检测:概念验证研究

阅读:3

Abstract

BACKGROUND: Traditional methods for detecting Alzheimer's disease (AD) pathology are often expensive, invasive, and inaccessible. Early identification of AD biomarkers is critical to enable timely and effective interventions that improve patient outcomes. To address these limitations, we developed a method to detect amyloid pathology, a hallmark of AD, using multi‐modal MRI—a cost‐effective, accessible, and non‐invasive imaging approach compared to traditional detection methods. METHOD: A deep learning model was trained using 1,488 paired AV‐45 amyloid PET and MRI scans from the Alzheimer's Disease Neuroimaging Initiative (ADNI), encompassing participants across the AD spectrum. The dataset was split into training, testing, and validation subsets (80:10:10). The model was validated on 149 participants across the AD continuum and cognitively healthy adults. T1‐weighted and T2‐weighted MRI scans were provided as inputs to synthesize an image representative of amyloid pathology (Figure 1). Standard Uptake Value Ratio (SUVR) scores were calculated for both the original PET images (ground truth) and the synthesized images. Sensitivity, specificity, and accuracy were assessed using an SUVR threshold of 1.55 to evaluate the model's ability to detect amyloid positivity. RESULT: On evaluation using the validation dataset, the model achieved a sensitivity of 0.70, specificity of 0.72, and accuracy of 0.71 in detecting amyloid positivity. The synthesized images demonstrated strong alignment with the original PET images, as reflected by similar SUVR scores and clinically relevant accuracy metrics. CONCLUSION: This proof of concept work provides a preliminary method to make amyloid detection safer, more affordable, and more accessible for patients and the healthcare industry. The technology leverages routinely acquired MRI scans to extract additional clinical insights, offering a non‐invasive tool for identifying candidates eligible for early disease‐modifying therapies. While further improvements to the model accuracy are ongoing, this approach has the potential to improve early detection and patient prognosis.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。