Inter-modality MRI Image Translation using Deep Learning

Magnetic resonance imaging (MRI) is one of the most predominantly used Medical Imaging in neurology and neurosurgery. In medical image translation, one medical image modality is translated into another medical image modality. Image translation and deep learning techniques in the medical domain reduce workload and total acquisition timing by removing the need to take some imagining sequences such are those taken in MRI called T1WI, T2WI, and Flair (most common sequence). Acquisition timing for Flair image modality is much longer than T1WI and T2WI. The objective of this work is to minimize the total acquisition time of MRIs by doing image translation on different MRI modalities called T1WI, T2WI, and Flair with the help of deep learning models. The proposed U-net Based model generates Flair image modality from T1WI and T2WI from the Deep Learning model in standard medical image format Digital Imaging and Communications in Medicine (DICOM) which significantly reduces the total acquisition timing of MRIs and thus reduces motion artefacts.