Prognosis Prediction of Alzheimer's Disease Based on Multi-Modal Diffusion Model

Alzheimer's Disease is a complex and currently one of the most prevalent illnesses. Due to these factors there is a growing emphasis on the early diagnosis of Alzheimer's Disease and our approach involves leveraging deep learning techniques to overcome this challenge. In this paper, we introduce a deep learning model designed to predict the progression of Alzheimer's Disease. Our model is based on the diffusion model and utilizes a multi-modal dataset that includes Magnetic Resonance Imaging data (image) and biospecimen data (clinical non-image) associated with Alzheimer's Disease. The proposed model operates through image-to-image translation based on a conditional diffusion process. Our findings validate that our model can generate images that faithfully capture the structural changes in the brains of Alzheimer's patients. Moreover, it outperforms other models according to various evaluation metrics such as PSNR, SSIM, and FID. Additionally, we demonstrate the superiority of a multi-modal dataset over a single modality dataset. We anticipate that the adoption of our proposed model will facilitate the early diagnosis of Alzheimer's Disease, thereby making a significant contribution to the medical field.