Precision Synthesis of Multi-Tracer PET via VLM-Modulated Rectified Flow for Stratifying Mild Cognitive Impairment

2026-04-13

AI Summary

This study addresses the limitations of early Alzheimer’s disease screening imposed by the high cost and radiation exposure of PET imaging, as well as the inability of existing MRI-to-PET synthesis methods to achieve personalized, high-fidelity generation. To overcome these challenges, the authors propose DIReCT++, a novel framework that integrates 3D rectified flow with the domain-adapted vision-language model BiomedCLIP. Leveraging both structural MRI and clinical text, DIReCT++ enables accurate synthesis of multi-tracer PET images—specifically $^{18}$F-AV-45 and $^{18}$F-FDG—with high fidelity. The method demonstrates strong cross-center generalizability and faithfully recapitulates disease-related patterns, significantly improving stratification accuracy for mild cognitive impairment across multicenter datasets. This approach offers a highly effective, minimally invasive tool for the early diagnosis of Alzheimer’s disease.

Technology Category

• Computer Vision: Multi-modal Vision
• Humans and AI: Brain-Sensing and Analysis
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Search and Retrieval-Augmented AI: Personalized, context-aware and across-device search
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

The biological definition of Alzheimer's disease (AD) relies on multi-modal neuroimaging, yet the clinical utility of positron emission tomography (PET) is limited by cost and radiation exposure, hindering early screening at preclinical or prodromal stages. While generative models offer a promising alternative by synthesizing PET from magnetic resonance imaging (MRI), achieving subject-specific precision remains a primary challenge. Here, we introduce DIReCT$++$, a Domain-Informed ReCTified flow model for synthesizing multi-tracer PET from MRI combined with fundamental clinical information. Our approach integrates a 3D rectified flow architecture to capture complex cross-modal and cross-tracer relationships with a domain-adapted vision-language model (BiomedCLIP) that provides text-guided, personalized generation using clinical scores and imaging knowledge. Extensive evaluations on multi-center datasets demonstrate that DIReCT$++$ not only produces synthetic PET images ($^{18}$F-AV-45 and $^{18}$F-FDG) of superior fidelity and generalizability but also accurately recapitulates disease-specific patterns. Crucially, combining these synthesized PET images with MRI enables precise personalized stratification of mild cognitive impairment (MCI), advancing a scalable, data-efficient tool for the early diagnosis and prognostic prediction of AD. The source code will be released on https://github.com/ladderlab-xjtu/DIReCT-PLUS.

Problem

• multi-tracer PET synthesis
• mild cognitive impairment stratification
• Alzheimer's disease early diagnosis
• precision medical imaging
• cross-modal generation

Innovation

• Rectified Flow
• Vision-Language Model
• Multi-Tracer PET Synthesis
• Personalized Generation
• Mild Cognitive Impairment Stratification