001-23 – Reconstruction of 2D MRI using SynthSR and DL+DiReCT for Subcortical Shape Analysis in Epilepsy

Individuals with Idiopathic Generalised Epilepsy (IGE) have previously presented with morphometric subcortical structure abnormalities, suggesting pathomechanistic involvement. 2D MRI is often acquired during the neuroradiological evaluation of IGE, whereas 3D MRI is preferable for morphometry. We compared the performance of 3D images synthesised from 2D scans against 3D scans in detecting morphometric abnormalities in an IGE cohort. During the same scanning session, 2D and 3D MRI were acquired from 39 healthy controls and 31 individuals with IGE at the Walton Centre NHS Foundation Trust, Liverpool. 3D images were synthesised from 2D scans using DL+DiReCT and SynthSR. Scans and synthetic images were processed comparably. Subcortical surface shape differences between the patients and the controls were evaluated. Similarities between the parcellations and shapes were measured. Dice coefficients were low across the 176 cortical parcellations but higher across the 15 subcortical structures (cortical/subcortical: 2D = .215/.361, DL+DiReCT = .198/.318, SynthSR = .217/.349). Globally, volumes were highly correlated (volume/thickness: 2D = .990/.903, DL+DiReCT = .989/.922, SynthSR = .985/.803, p < .001). In the 3D scans, surface deflations were detected in the left accumbens, right thalamus, and the caudate and putamen bilaterally of the IGE group, compared to controls (IGE/controls: males = 14/16, mean age = 32.16/32.13). The 2D scans yielded no significant shape changes, whereas the synthesised images did, some of which were not found in the 3D scans. IGE patients presented subcortical shape changes suggestive of regional atrophy. Images synthesised from 2D scans led to the identification of more IGE-related subcortical shape abnormalities than the unprocessed 2D scans, alongside abnormalities not detected in the 3D scans. Synthesis potentially facilitates morphometry with reduced MRI requirements in epilepsy, but quantitative analyses should be interpreted cautiously.