Advancing brain tumor segmentation in smart healthcare via T1-generated virtual multimodal MRI fusion

Precision segmentation of brain tumors is crucial for diagnosis, treatment, and prognosis in smart healthcare. While traditional single-modality methods have performance limitations, multimodal segmentation, though more accurate, increases patient scanning time and strains medical resources. This paper introduces a novel approach using a denoising diffusion probabilistic models to generate virtual T2, T1CE, and FLAIR images from single-modality T1, enhancing segmentation without additional scans. By concurrently training the generation and segmentation networks, we achieve realistic multimodal images and precise tumor segmentation. Experiments show our method significantly outperforms single-modality techniques and rivals real multimodal segmentation, reducing the need for multimodal images, optimizing resource use, and offering an alternative for patients unable to receive contrast enhancers.