We aimed to determine the alterations in the subcortical structures of patients with idiopathic generalized epilepsy with tonic–clonic seizures (IGE-GTCS) via MRI volumetry and vertex-based shape analysis and to evaluate the relationships between MRI measures and drug responses. In a follow-up sample of 48 patients with IGE-GTCS and 48 matched normal controls (NCs), high-resolution 3D T1WI was performed at baseline. After 1 year of follow-up, 31 patients were classified as seizure free (SF) and 17 as drug resistant (DR). The volumes of subcortical structures were extracted, and vertex-based shape analysis was performed using FSL-Integrated Registration and Segmentation Toolbox (FSL-FIRST). Comparisons among groups were calculated adjusting for covariates [age, sex, and intracranial volume (ICV)]. Analysis of the relationships among imaging biomarkers along with frequency and duration was assessed using partial correlations. The differential imaging indicators were used as features in a linear support vector machine (LSVM). The DR group displayed significant regional atrophy in the volume of the left amygdala compared with NCs (p = 0.004, false discovery rate corrected) and SF patients (p = 0.029, uncorrected). Meanwhile, vertex-based shape analysis showed focal inward deformation in the basolateral subregion of the left amygdala in DR compared with the results for SF and NC (p < 0.05, FWE corrected). There were significant correlations between the volume changes and seizure frequency (r = −0.324, p = 0.030) and between shape (r = −0.438, p = 0.003) changes and seizure frequency. Moreover, the volume of the left thalamus in the DR group was significantly correlated with seizure frequency (r = −0.689, p = 0.006). The SVM results revealed areas under the receiver operating characteristic curve of 0.82, 0.68, and 0.88 for the classification between SF and DR, between SF and NC, and between DR and NC, respectively. This study indicates the presence of focal atrophy in the basolateral region of the left amygdala in patients with IGE drug resistance; this finding may help predict drug responses and suggests a potential therapeutic target.