The management of phantom limb pain (PLP) is still challenging due to a partial understanding of its neurophysiological mechanisms. Structural neuroimaging features are potential biomarkers. However, only a few studies assessed their correlations with clinical severity and treatment response. This study aims to explore the association between brain gray matter volume (GMV) with phantom limb manifestations severity and PLP improvement after neuromodulatory treatments (transcranial direct current stimulation and mirror therapy). Voxel-based morphometry analyses and functional decoding using a reverse inference term-based meta-analytic approach were used. We included 24 lower limb traumatic amputees with moderate to severe PLP. We found that alterations of cortical GMV were correlated with PLP severity but not with other clinical manifestations. Less PLP severity was associated with larger brain clusters GMV in the non-affected prefrontal, insula (non-affected mid-anterior region), and bilateral thalamus. However, only the insula cluster survived adjustments. Moreover, the reverse inference meta-analytic approach revealed that the found insula cluster is highly functionally connected to the contralateral insula and premotor cortices, and the decoded psychological processes related to this cluster were "rating," "sustained attention," "impulsivity, " and "suffering." Moreover, we found that responders to neuromodulatory treatment have higher GMV in somatosensory areas (total volume of S1 and S2) in the affected hemisphere at baseline, compared to non-responders, even after adjustments.