Search papers, labs, and topics across Lattice.
The paper introduces PASTA, a conditional diffusion model for translating MRI scans into PET scans, addressing the need for pathology awareness in cross-modality medical image translation. PASTA employs a dual-arm architecture and multi-modal condition integration to preserve both structural and pathological details. Results show that PASTA generates high-quality 3D PET images, improving Alzheimer's diagnosis performance by 4% compared to MRI, approaching the performance of real PET scans.
Synthesized PET scans from MRI can nearly match the diagnostic accuracy of real PET for Alzheimer's, potentially unlocking wider access to crucial functional insights.
Positron emission tomography (PET) is a widely recognized technique for diagnosing neurodegenerative diseases, offering critical functional insights. However, its high costs and radiation exposure hinder its widespread use. In contrast, magnetic resonance imaging (MRI) does not involve such limitations. While MRI also detects neurodegenerative changes, it is less sensitive for diagnosis compared to PET. To overcome such limitations, one approach is to generate synthetic PET from MRI. Recent advances in generative models have paved the way for cross-modality medical image translation; however, existing methods largely emphasize structural preservation while neglecting the critical need for pathology awareness. To address this gap, we propose PASTA, a novel image translation framework built on conditional diffusion models with enhanced pathology awareness. PASTA surpasses state-of-the-art methods by preserving both structural and pathological details through its highly interactive dual-arm architecture and multi-modal condition integration. Additionally, we introduce a novel cycle exchange consistency and volumetric generation strategy that significantly enhances PASTA's ability to produce high-quality 3D PET images. Our qualitative and quantitative results demonstrate the high quality and pathology awareness of the synthesized PET scans. For Alzheimer's diagnosis, the performance of these synthesized scans improves over MRI by 4%, almost reaching the performance of actual PET. Our code is available at https://github.com/ai-med/PASTA.
