Deep learning model provided mobile assessment of adolescent idiopathic scoliosis
Click Here to Manage Email Alerts
Key takeaways:
- An app-based deep learning model was able to diagnose and monitor adolescent idiopathic scoliosis.
- The model distinguished severity and curve type with a higher sensitivity compared with two spine surgeons.
Published results showed a smartphone app-based validated deep learning model was able to diagnose and monitor adolescent idiopathic scoliosis at minimal cost with no additional radiation.
AlignProCARE (Aimed) is an app-based spinal evaluation platform that uses a validated, deep neural network model to assess adolescent idiopathic scoliosis (AIS) from smartphone photographs of patients’ backs. To determine the efficacy of the AlignProCARE platform, researchers performed a diagnostic study of 2,158 patients with AIS and blindly compared the results with those from two spine surgeons. Outcomes included classification of AIS severity, curve type and progression during a 6-month follow-up period.
Overall, the model recommended follow-up for patients with an area under receiver operating characteristic curve (AUC) of 0.839 and recommended considering surgery for patients with an AUC of 0.902. Additionally, the model demonstrated the ability to distinguish among thoracic (AUC = 0.777), thoracolumbar or lumbar (AUC = 0.760) and mixed (AUC = 0.860) curve types.
Researchers also noted the model was able to recognize severity and curve type with a higher sensitivity compared with the two surgeons. The model and the two surgeons were comparable in sensitivity for distinguishing curve progression.
“With no extra radiation and minimal cost, this model could provide continuous monitoring with prompt interventions triggered when progression is detected,” the researchers wrote in the study. “This could contribute to further treatment planning and monitoring for the patient by providing computer-aided real-time assessments to aid physicians’ management [and] decision-making,” they concluded.