Correction to: Machine learning-based reconstruction of 2D MRI for quantitative morphometry in epilepsy

In the original article, we report a clinical cohort comprised of healthy controls (HC), and two phenotypic subgroups of people with idiopathic generalised epilepsy: 42 people with drug-resistant idiopathic generalised epilepsy, and 33 (31 following exclusions) people with drug-sensitive idiopathic generalised epilepsy—pwDRIGE and pwDSIGE in the original manuscript, respectively. Following a review of archived data, it was flagged that these phenotypic labels had been mistakenly assigned at an early stage in this study. The pwDRIGE subgroup was instead comprised of 42 people with drug-resistant focal epilepsy, and the pwDSIGE group 31 people with IGE (11 of whom were drug-sensitive). Group definitions have been changed to reflect this, but importantly, subgroup characteristics remain unchanged and are consistent with the original manuscript. The updated demographic information is available in Table 1 (a revision of Table 2 from the original manuscript) and Table 2 (a revision of Table S1 from the original Supplementary Material). An update to Supplementary Figure 7 shows the relabelled results of subgroup comparisons (see Fig. 1).This error does not impact on our primary conclusions, which regarded the reliability of image processing methods and was agnostic to the demographics of the included participants. Furthermore, the groupings used in our secondary clinical analyses remain the same. Nonetheless, we interpreted several of our findings in the context of a homogenous idiopathic generalised epilepsy cohort, whereby systematic differences could be reasonably attributed to drug resistance—see section 4.2. of the original manuscript. These interpretations are no longer wholly valid, and we therefore offer an update to the clinical summary from section 4.2.As stated in the original article, data is available on request from the authors, now with corrected metadata. Additionally, on request from the authors, a version of the original manuscript is available where all errors have been highligted. Code (unchanged) freely available on GitHub: https://github.com/C-Ratcliffe/221216_Proj-IS.With modern analysis techniques, imaging abnormalities in pwE are reported with increasing frequency and evidence suggests that associated structural abnormalities extend beyond putative atrophy, potentially including abnormal orientations or regional subcortical thickening (hypertrophy) (Caciagli et al., 2019; Zubal et al., 2025). Limited by a heterogenous epilepsy sample of a modest size, and the relative heterogeneity of the epilepsy phenotype, we opted to focus on examining subcortical morphometry for the clinical arm of this project. (McKavanagh et al., 2023; Whelan et al., 2018) We also present clinical interpretations tentatively.A subset of the data analysed in this paper has been examined previously by Brownhill et al. (2021), who found evidence of pallidal, putaminal, and thalamic volumetric reductions (relative to controls) in the truncated IGE-only cohort. IGE is a heterogenous subtype of epilepsy, the phenotypes of which the present sample is insufficient to represent reliably—especially when combined into a mixed-aetiology pwE group. Nonetheless, our clinical analyses indicate convergent structural abnormalities in individuals with IGE and FE, consistent with the concept of a core epileptogenic network. This study therefore recapitulates and expands on the findings of previous structural imaging analyses into the neural correlates of epilepsy.Firstly by extending the findings of Brownhill et al. (2021) and identifying bilateral pallidal volume reductions in a mixed-aetiology epilepsy cohort, but also by demonstrating that whilst volumetric analyses might lack the necessary sensitivity to identify pathology-specific abnormalities in other key regions, subcortical shape analyses offer a means to identify pathology-related focal atrophy and hypertrophy. Specifically, we highlight abnormal inwards surface deflation in the accumbens, caudate, and thalamus (concordant with volumetric abnormalities found in Whelan et al. (2018)), and outwards regional inflation in the caudate-despite the lack of any volumetric abnormalities, such as those specific to the pallidum.Despite the inclusion of a focal epilepsy subgroup, hypertrophy of the amygdalae was not present (and was not indicated in supplementary subgroup analyses). Whilst commonly associated with mTLE, there is no consistent evidence suggesting relating increased amygdalae volumes to IGE, suggesting instead some potential influence on the distinct cognitive profiles of mTLE and IGE (Ratcliffe et al., 2020; Zubal et al., 2025). Our findings support the hypothesis that the interaction between structural abnormalities and phenotype in epilepsy are multifaceted and complex, and that further focused study could deliver insights into the impact of subcortical/projection network reorganisation on seizure presentation and comorbid cognitive impairment (Rodriguez-Cruces et al., 2022).