Uncovering cortical layers with multi-exponential analysis: a region of interest study

Abstract

Pathologies of the cerebral cortex often manifest at resolutions outside of the scope of conventional magnetic resonance imaging (MRI). Two different pathways aiming to overcome this limitation have emerged in recent years. One is focused on the direct imaging of the cortical layers achieved by increasing the MRI spatial resolution. The other approach relies on low-resolution images acquired at 3 T and represents the cortical layers in the domain of $T_{1}$ spin-lattice relaxation. In this work, we follow the $T_{1}$-mapping-based approach and explore two possible methods to achieve the representation of cortical layers: (1) modeling using a multi-exponential model, and (2) inverse Laplace transformation (ILT). Several regions of interest (ROI) across the cerebral cortex were measured and later used to create the ground-truth dataset. Using this data, the performance of the two models was evaluated. The ILT method proved superior to the multi-exponential model, yielding separation of all components with an average estimation error of 2.52 %. This method may enrich the low-resolution imaging framework by providing a more precise estimation of the spin-lattice spectrum.