The Advanced Toolbox for Multitask Medical Imaging Consistency (ATOMMIC): A framework to facilitate Deep Learning in Magnetic Resonance Imaging

Dimitrios Karkalousos, Ivana Isgum, Henk Marquering, Matthan W. A. Caan

Keywords: Deep Learning, Magnetic Resonance Imaging, Image reconstruction, Image segmentation, Multitask Learning

Abstract: Integrating Deep Learning (DL) into medical imaging, particularly in Magnetic Resonance Imaging (MRI), has marked a significant advancement in the field, enhancing the efficiency and accuracy of tasks such as image reconstruction, segmentation, and quantitative parameter map estimation. Despite these advancements, existing frameworks have limited support to perform multiple tasks simultaneously, essential for optimizing the workflow from data acquisition to analysis. Addressing this gap, we introduce the Advanced Toolbox for Multitask Medical Imaging Consistency (ATOMMIC), a novel open-source toolbox designed to facilitate the integration of multiple MRI tasks within a unified MultiTask Learning (MTL) framework. ATOMMIC supports a wide range of DL models and datasets, allowing for seamless and consistent execution of multiple tasks. By enabling joint task execution and supporting complex and real-valued data, ATOMMIC allows to streamline various DL applications in MRI reconstruction and analysis.

Submission Number: 161