Comprehensive Summary
This study evaluates whether applying deep learning reconstruction (DLR) to a fast gray-matter T1 inversion recovery (FGATIR) imaging improves visualization and delineation of the centromedian nucleus (CM) of the thalamus. This is to enable direct targeting for neuromodulation in idiopathic epilepsy and Lennox-Gastaut syndrome. The analysis of the patients who underwent thalamic neuromodulation using DLR-FGATIR images compared to the Morel histologic atlas showed that the centromedian nucleus appeared as a hypointense region which was distinct to its neighboring nuclei. Through different brain axial images, it was quantified that DLR-FGATIR slices achieved a Dice score of 0.49 as an experienced rater, and successfully traversed the Morel atlas CM. These demonstrate that DLR-FGATIR provided high-resolution, high-contrast imaging of medial and posterior CM boundaries, aligning well with expert segmentation at high field strength.
Outcomes and Implications
This research is important because precise targeting of the CM is critical for effective neuromodulation, yet has been a longstanding challenge due to lack of reliable radiographic or electrophysiologic landmarks. By showing that DLR-FGATIR can visualize the CM with sufficient contrast and accuracy, the study suggests a safer, more efficient alternative to indirect atlas-based targeting, microelectrode recordings, or advanced MRI methods that require lengthy acquisition times and postprocessing. This means surgeons can easily access and utilize this sequence for clinical practices to directly target and localize the centromedian nucleus for deep brain stimulation. While DLR-FGATIR hasn’t been measured or accessed for its effects on neuromodulation, it can continue to work as guidance for electrode placement within the CM, and possibly increase its use for direct targeting in epilepsy.