AI Can Now Predict Which Kids' Scoliosis Will Get Worse
Could spare teens from unnecessary braces, X-rays, and surgery.
Researchers trained a transformer model that takes in the curve of the spine to predict curve progression in adolescent idiopathic scoliosis — a lateral curvature of the spine that develops during adolescence, without known cause. The objective is to predict progression from ages 9 to 24, as measured from a sequence of DXA scans. The model was trained using a large-scale synthetic dataset of spine curves and their time series, covering different curve types and different progression patterns. It was able to generalise from synthetic to real data when evaluated on a dataset of real DXA scans covering multiple time points, and fine-tuning on real data gave a significant boost to performance. The model accurately predicted spine curve progression in both scoliosis and normal cases. The condition can result in significant pain and disability, and often progresses rapidly during adolescence.
- The AI reads spine scans and forecasts how much a child's curve will worsen over the following years
- It first learned on computer-generated spine data, then improved sharply when trained on real patient scans
- Better forecasts could mean fewer X-rays and braces, and catch the kids who really need treatment sooner
Why It Matters
Fewer unnecessary X-rays and braces for kids, and earlier help for the ones who truly need it.