Lesion-Specific Prediction with Discriminator-Based Supervised Guided Attention Module Enabled GANs in Multiple Sclerosis

• DOI: 10.48550/arxiv.2208.14533
• Publication Date: 2022-01-01
• Authors: Jueqi Wang; Derek Berger; Erin L. Mazerolle; Jean-Alexis Delamer; Jacob Levman
• Keywords: Fluid-attenuated inversion recovery; Discriminator

Multiple Sclerosis (MS) is a chronic neurological condition characterized by the development of lesions in white matter brain. T2-fluid attenuated inversion recovery (FLAIR) brain magnetic resonance imaging (MRI) provides superior visualization and characterization MS lesions, relative to other MRI modalities. Follow-up FLAIR helpful information for clinicians towards monitoring disease progression. In this study, we propose novel modification generative adversarial networks (GANs) predict future lesion-specific at fixed time intervals. We use supervised guided attention dilated convolutions discriminator, which supports making an informed prediction whether generated images are real or not based on lesion area, turn has potential help improve generator area examinations more accurately. compared our method several baselines one state-of-art CF-SAGAN model [1]. conclusion, results indicate that proposed achieves higher accuracy reduces standard deviation errors with models similar overall performance.