Automatic framework for multiple sclerosis follow-up by magnetic resonance imaging for reducing contrast agents

An automatic framework for multiple sclerosis (MS) follow-up by Magnetic Resonance Imaging (MRI) is presented. It is based on the identification and segmentation of lesions by using convolutional neural network (CNN) architecture applied to the volumes collected by different imaging modalities and on the registration of the volumes obtained by two consecutive examinations. The resulting binary masks obtained from the identification/segmentation strategy on each examination are used to calculate the volume of each lesions, their status (chronic or active) and, hence, to estimate the progression of the disease. Preliminary results are reported demonstrating that the calculations performed by the proposed framework are capable, when the disease is stable, to gather the same information obtainable when the contrast agent (CA) is administered to the patient.