A new artificial intelligence program for the automatic evaluation of scoliosis on frontal spinal radiographs: Accuracy, advantages and limitations

Abstract

BACKGROUND: Scoliosis is one of the most common spinal deformations that is usually diagnosed on frontal radiographs using Cobb’s method. The use of automatic measurement methods based on artificial intelligence can overcome many drawbacks of the usual method and can significantly save the time of the radiologist. AIM: to analyze the accuracy, advantages and disadvantages of a newly developed AI program for automatic diagnosis of scoliosis and measurement of Cobb’s angle on frontal radiographs. Methods: 112 digital radiographs were used to test the agreement of Cobb’s angle measurements between the new automatic method and the radiologist using Blant-Altman method on Microsoft Excel. A limited clinical accuracy test was also conducted using 120 radiographs. The accuracy of the system in defining the grade of scoliosis was evaluated by calculating sensitivity; specificity; accuracy; and area under the ROC curve (ROC AUC). RESULTS: The agreement of Cobb’s angle measurement between the system and the radiologist was found mostly in scoliosis with grades 1 and 2. Only 2.8% of the results were found to be unsatisfying with an angle variability of more than 5°. Diagnostic accuracy metrics of the limited clinical trial in Mariinsky city hospital had also proved the reliability of the system, with sensitivity = 0.97; specificity = 0.88; accuracy (general validity) = 0.93; area under the ROC curve (ROC AUC) = 0.93. CONCLUSION: Overall, the AI program can automatically and accurately define the grade of scoliosis and measure the angles of spinal curvatures on frontal radiographs.