Convolutional neural network measures Cobb angle for patients with idiopathic scoliosis
Click Here to Manage Email Alerts
Key takeaways:
- A convolutional neural network accurately measured Cobb angles of patients with adolescent idiopathic scoliosis.
- The artificial intelligence method was comparable with manual assessment from six doctors.
Published results showed a convolutional neural network method successfully detected vertebrae and measured Cobb angles on X-ray imaging of patients with adolescent idiopathic scoliosis.
Researchers compared the ability of an artificial intelligence (AI) convolutional neural network to assess Cobb angles on X-ray images of patients with adolescent idiopathic scoliosis (AIS) vs. two doctors who specialize in scoliosis treatment, two doctors who were spine specialists and two doctors who were in year 3 of their post-graduate studies. According to the study, researchers compared intraclass correlation coefficients (ICCs) between the groups.
Overall, the average Cobb angle difference between AI assessment and doctor assessment ranged from 2.8° to 4.6°. The smallest difference was seen on standing X-ray with a minor 2 curve type, while the largest difference was seen on supine side-bending X-ray with a minor 2 curve type. Researchers noted ICCs for both the AI and the doctors were “excellent or good” with ICCs greater than 0.96 on standing X-ray for all curve types.
According to the study, the mean error between the AI and the doctors was not affected by Cobb angle size, with AI measuring 1.7° to 2.2° smaller than that measured by the doctors.
“The proposed method showed a high correlation with the doctors’ measurements, regardless of the [Cobb angle] size, doctors’ experience and patient posture,” the researchers wrote in the study. “The proposed method showed excellent reliability, indicating that it is a promising automated method for measuring [Cobb angle] in patients with AIS,” they concluded.
Collapse
Read more about
Click Here to Manage Email Alerts