DTI with Minimal Data: Image Translation Based Distortion Correction and FA Map Generation for Clinical Efficiency

Motivation: Current DTI processing requires around 30 directions per shell to ensure data quality, which is difficult to obtain in certain vulnerable patient groups. Goal(s): To reduce number of DWI directions needed in clinical study. Approach: This study proposes a pipeline with translation-based registration method to correct both distortions and motion artifacts, followed by a generative adversarial network (GAN) to generate fractional anisotropy map, leveraging high quality data from HCP, with fine-tuning for applications. Results: We validate this approach in a subset of HCP DTI, healthy volunteer, and on epilepsy patient data, using only 6 DWI images while preserving lesion details. Impact: Success of the proposed pipeline will enable much shorter DTI acquisition time for patients who cannot stay still throughout a multidirectional DTI scans, which holds great potential to becoming a promising tool for clinical applications.