Scoliosis is a musculoskeletal disorder where early detection and precise evaluation are critical. However, traditional diagnosis using X-ray images has relied heavily on manual measurements, often resulting in time-consuming processes and inconsistent results among clinicians.

 

To address these challenges, the HealthHub research team has developed a novel AI-powered pipeline that automates scoliosis diagnosis and standardizes clinical reporting. This groundbreaking work was recently published in the prestigious international journal Scientific Reports.

 

The proposed pipeline consists of three core capabilities:

  1. Automated vertebral segmentation from X-ray images to identify individual spinal segments.

  2. Automatic calculation of Cobb angles, the key metric used in scoliosis diagnosis.

  3. Structured clinical reporting using SNOMED CT-based templates to standardize documentation across cases.

 

By automating these steps, the system significantly reduces inter-observer variability and improves diagnostic consistency. It also shortens the diagnostic process, thereby enhancing workflow efficiency for clinicians.

 

The research team evaluated the system using the public SpineWeb dataset. The results showed outstanding accuracy, with a Circular Mean Absolute Error (CMAE) of 3.50 degrees and a Symmetric Mean Absolute Percentage Error (SMAPE) of 7.35%, outperforming traditional manual measurements and demonstrating strong clinical applicability.

 

This study is expected to have a significant impact on the field of scoliosis diagnosis. In addition to improving diagnostic accuracy, the use of standardized SNOMED CT-based reports enhances data quality and facilitates large-scale clinical research and inter-hospital data comparison.

 

Ultimately, this innovation offers a robust foundation for delivering faster, more accurate, and more consistent care to scoliosis patients.

 

The publication marks a meaningful step forward in redefining diagnostic workflows in musculoskeletal radiology and highlights the transformative potential of AI in clinical practice.

 

 

 

Read the full paper:
https://link.springer.com/article/10.1038/s41598-025-01952-w

 

Download PDF:
https://link.springer.com/content/pdf/10.1038/s41598-025-01952-w.pdf