Desenvolvimento de método em cascata para realização do cálculo do ângulo de Cobb em imagens de raio-x da coluna vertebral

Abstract

Scoliosis is a spinal deformity that affects approximately 2 to 4% of the global population, and its early diagnosis is essential to prevent disease progression and avoid severe cases, which may only be reversible through corrective surgery. The diagnosis is usually performed using x-ray images, in which the specialist measures the Cobb angle, a parameter used to determine the severity of the deformity. In order to assist specialists in this process, this work presents a method for the automatic estimation of the Cobb angle based on vertebral segmentation. Initially, the region of interest (ROI) is extracted through spine segmentation, followed by vertebrae segmentation from the image delimited by the ROI and, finally, the Cobb angle is estimated for each obtained vertebral mask. The proposed method achieved Dice coefficients of 92.14% and 83.23% for spine and vertebrae segmentation, respectively, and a mean absolute error (MAE) of 8.72° in Cobb angle calculation.