Anatomical plane multi-view learning for detecting spinal deformities with surface topography

Abstract

Detecting functional and structural spinal deformities in children, such as rounded back and scoliosis, is essential to enable timely treatment and reduce the risk of reduced quality of life in adulthood. This study proposes a multi-view binary classifier based on anatomical planes for identifying postural deviations. Tested on 1,477 surface topographical measurements from school children, the model achieved 75% accuracy, 84% specificity, and 64% sensitivity. With further optimization and validation on broader cohorts, this approach shows promosing as a scalable tool for school-based screening programs.