BACKGROUND: Early-onset major depressive disorder (EO-MDD) is characterized by its significant heterogeneity, hindering progress in research. Traditional case-control studies, like group-level structural covariance network, struggle to capture individual heterogeneity among EO-MDD patients. METHODS: In this study, T1-weighted structural magnetic resonance imaging was obtained from 185 participants, including 103 EO-MDD patients and 82 healthy controls. A subject-level individual differential structural covariance network (IDSCN) was constructed for each patient based on the concept of normative model. Semi-supervised clustering algorithms were then employed to classify EO-MDD subtypes, followed by validation analyses to assess clustering stability. RESULTS: Our study identified two neuroanatomical subtypes. The low-covariance subtype is characterized by significant neural maturation gaps across the whole brain and more pronounced anxiety somatization symptoms. Conversely, the high-covariance subtype demonstrates simultaneous mature of brain structures. CONCLUSION: Our findings provide valuable insights into the neuroanatomical heterogeneity of EO-MDD patients, highlighting the importance of considering individual symptom profiles in subtype classification. These findings have substantial clinical implications for personalized treatment and precision medicine, offering more effective treatment choices and accurate diagnoses.