MRI-to-PET Image Synthesis for Alzheimer’s Disease — MIT “Artificial Intelligence and Machine Learning” Offline Laboratory Research Project (Core Team Member)
Fine-tuningServed as a core team member in an offline AI laboratory project to develop and optimize a diffusion model for MRI-to-PET image synthesis for Alzheimer’s disease research. The work involved creating training-ready synthetic medical images and validating generation quality against established academic benchmarks. Responsibilities included model development with medical image inputs and assessing registration performance between MRI and PET modalities. • Developed a novel diffusion model using Transformer architectures. • Implemented image generation and fusion workflows with Python and U-Net/GAN components. • Achieved sub-millimeter registration accuracy (<0.5mm) for MRI/PET fusion. • Validated model performance against leading academic benchmarks and supported publication efforts.