An algorithm for using deep learning convolutional neural networks with three dimensional depth sensor imaging in scoliosis detection

• Published in: The spine journal Vol. 21; no. 6; pp. 980 - 987
• Main Authors: Kokabu, Terufumi, Kanai, Satoshi, Kawakami, Noriaki, Uno, Koki, Kotani, Toshiaki, Suzuki, Teppei, Tachi, Hiroyuki, Abe, Yuichiro, Iwasaki, Norimasa, Sudo, Hideki
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.06.2021
• Subjects: Accuracy; Adolescent idiopathic scoliosis; Cobb angle; Convolutional neural network for regression; Correlation coefficient analyses; Deep learning algorithm; Mean absolute error; Noncontact and noninvasive system; Three-dimensional depth sensor; Accuracy; Correlation coefficient analyses; Noncontact and noninvasive system; Mean absolute error; Adolescent idiopathic scoliosis; Convolutional neural network for regression; Three-dimensional depth sensor; Cobb angle; Deep learning algorithm
• ISSN: 1529-9430; 1878-1632; 1878-1632
• DOI: 10.1016/j.spinee.2021.01.022

Timely intervention in growing individuals, such as brace treatment, relies on early detection of adolescent idiopathic scoliosis (AIS). To this end, several screening methods have been implemented. However, these methods have limitations in predicting the Cobb angle. This study aimed to evaluate the performance of a three-dimensional depth sensor imaging system with a deep learning algorithm, in predicting the Cobb angle in AIS. Retrospective analysis of prospectively collected, consecutive, nonrandomized series of patients at five scoliosis centers in Japan. One hundred and-sixty human subjects suspected to have AIS were included. Patient demographics, radiographic measurements, and predicted Cobb angle derived from the deep learning algorithm were the outcome measures for this study. One hundred and sixty data files were shuffled into five datasets with 32 data files at random (dataset 1, 2, 3, 4, and 5) and five-fold cross validation was performed. The relationships between the actual and predicted Cobb angles were calculated using Pearson's correlation coefficient analyses. The prediction performances of the network models were evaluated using mean absolute error and root mean square error between the actual and predicted Cobb angles. The shuffling into five datasets and five-fold cross validation was conducted ten times. There were no study-specific biases related to conflicts of interest. The correlation between the actual and the mean predicted Cobb angles was 0.91. The mean absolute error and root mean square error were 4.0° and 5.4°, respectively. The accuracy of the mean predicted Cobb angle was 94% for identifying a Cobb angle of ≥10° and 89% for that of ≥20°. The three-dimensional depth sensor imaging system with its newly innovated convolutional neural network for regression is objective and has significant ability to predict the Cobb angle in children and adolescents. This system is expected to be used for screening scoliosis in clinics or physical examination at schools.