Purpose: The coronal Cobb angle is commonly used for assessing the adolescent idiopathic scoliosis (AIS); however, it may underestimate the severity of AIS while the plane of maximum curvature (PMC) could be a promising descriptor for three-dimensional assessment of AIS. This study aimed to develop a computational method (CM) for estimating the PMC based on the coronal and sagittal images of the spine, and to verify the results with computed tomography (CT). Methods: Twenty-eight thoracic and 24 lumbar curves from 30 subjects with AIS were analysed. For the CM, PMC was estimated via identifying the two corner points at the superior endplate of upper-end vertebra and the inferior endplate of lower-end vertebra in the coronal and sagittal CT images separately (eight corner points in total). For the CT, PMC was determined through rotating the spine images axially until the maximum Cobb angle was found. Intraclass correlation coefficient (ICC), Bland–Altman method and linear regression analysis were used for the statistical analyses. Results: The high ICC values (intra- > 0.91; inter- > 0.84) suggested very good intra- and inter-rater reliability of the CM in PMC estimation. The high ICC values (> 0.91) and assessment of Bland–Altman method demonstrated a good agreement between the PMC acquired using the CM and CT. The generated linear regression equations (R2 > 0.69) could allow to estimate the PMC (originally measured through the CT) via the CM. Conclusion: The developed computational method could estimate reliable and valid PMC for the patients with AIS, and become feasible for three-dimensional assessment of AIS. Level of evidence: III. [ABSTRACT FROM AUTHOR]