Frequent monitoring is essential for assessing the progression of adolescent idiopathic scoliosis (AIS), especially during years of growth. Standard practice involves radiographic measurement of Cobb angles every 4–6 months; however, this may not align with curve progression given unpredictable growth patterns. The emergence of clinical artificial intelligence (AI) tools, such as the Momentum SpineTM smartphone application, have been developed to address monitoring concerns. Video-imaging technology is used to measure extra-spinal, topographic markers that are correlated to the magnitude of spinal deformity. The purpose of the current study is 1) to validate the AI-generated Cobb angle produced via Momentum SpineTM and 2) to assess the AI's ability to track disease progression overtime. Eligible AIS patients (n=131) consented to partake in the study and underwent same-day radiographic imaging and AI scan at the baseline visit. Subsequent same-day AI scans were performed at any follow-up visits with radiographic imaging, with 17 patients having completed an additional follow-up visit thus far. Paired t-test was used to compare all same-day radiographic and AI-generated Cobb angles. Two-way ANOVA was used to assess the accuracy of AI-generated Cobb angle compared to radiographic Cobb angle overtime. Paired t-test results of all scans demonstrated that mean Momentum SpineTM main curve Cobb angle differed by an average of 6.72° compared to the radiographic measurement ((MAI = 26.8±12.7° vs. Mradiograph = 33.5±14.2°). Despite this difference being statistically significant, (−6.72°, 95% CI −8.19°,−5.26°, t(127)=9.09, p < 0 .0001), it falls within the accepted range for inter-rater variability of ~7° between clinicians. In the population with follow-up, post-hoc analysis showed the agreement between AI-predicted Cobb angle and radiographic measurement improved to an average difference of 4.05° (AI:24.24±13.4° vs radiograph:28.29±14.0°, p=0.6788). Momentum SpineTM can estimate Cobb angles within the accepted range of inter-rater variability between clinicians when compared to same-day radiographic Cobb angles. Therefore, it can be used to detect and monitor AIS. The next step in this study is to evaluate at-home AI scans performed on a monthly basis and determine potential practice change by scheduling clinic visits when progression is detected, as well as eliminating planned visits when the curve is reported to be stable.
Liu et al. (Wed,) studied this question.