Integrating Brain Morphological Features and Ionized Serum Magnesium to Identify Mild Tic Comorbidity in Children with Autism Spectrum Disorder.

Background: Autism spectrum disorder (ASD) frequently co-occurs with tic disorders, yet clinical differentiation remains challenging. This study developed and validated a predictive model combining brain morphological imaging and serum trace elements to distinguish ASD alone from ASD with comorbid mild tic disorders. Methods: This retrospective cross-sectional diagnostic study included 104 children aged 4-15 years (90 boys and 14 girls): 53 with ASD alone and 51 with ASD and mild tic disorders. Participants were randomly divided into training and internal validation cohorts at a 7:3 ratio. Candidate predictors were screened in the training cohort with correction for multiple comparisons and further selected using least absolute shrinkage and selection operator (LASSO) logistic regression. These features were incorporated into a multivariable regression equation and a nomogram. Model performance and internal validation were assessed via receiver operating characteristic (ROC) analysis, the Hosmer-Lemeshow test, and decision curve analysis (DCA). Results: Independent predictors included asymmetry indices of the caudate nucleus, nucleus accumbens, and paratenial thalamic nucleus; cortical curvatures of the left anterior cingulate cortex and right lateral occipital gyrus; and ionized serum magnesium levels (all p < 0.05). The model achieved the areas under the ROC curves (AUROCs) of 0.904 (95% CI: 0.834-0.975) in the training cohort and 0.826 (95% CI: 0.664-0.988) in the internal validation cohort, outperforming individual predictors. Calibration was acceptable, and DCA suggested potential clinical utility within this cohort. Conclusion: The nomogram prediction model accurately distinguishes between ASD and ASD-mT, showing strong discriminative power and clinical value. It may aid clinicians in early comorbidity detection and guide treatment decisions.