Automatic Spine Curvature Estimation from X-ray Images

Automatic Spine Curvature Estimation from X-ray Images

Idiopathic scoliosis, a common spinal deformity, poses significant health risks if not detected and treated early. The Cobb angle, a standard clinical metric for measuring spinal curvature, plays a crucial role in scoliosis diagnosis and treatment planning. However, manual measurement of the Cobb angle is prone to inter-operator variability and is a time-consuming task for surgeons and radiologists.

Problem:

Manual measurement of the Cobb angle suffers from inter-operator variability and is laborious for clinicians. This variability and time-consuming nature hinder timely and accurate diagnosis of scoliosis, potentially delaying crucial interventions. Additionally, existing measurement methods may not generalize well to unseen X-ray images, as they may come from different distributions than the training set.

Research Aim:

The primary aim of this research is to develop an automated method for measuring the Cobb angle directly from X-ray images of the spine. The focus is on exploring robust methods capable of adapting to test set images from distributions different from the training set. By addressing the challenges of inter-operator variability and laborious manual measurement, the goal is to provide clinicians with a reliable and efficient tool for accurate Cobb angle measurement in scoliosis diagnosis and treatment planning.

Outcome So Far:

Initial research efforts have begun to explore and develop automated methods for measuring the Cobb angle from X-ray images. Strategies for robust adaptation to test set images with different distributions are being investigated. While the project is ongoing, preliminary findings suggest promising potential for developing an automated system that addresses the challenges of inter-operator variability and tedious manual measurement. Ongoing research aims to further refine and validate the proposed methods, with the ultimate goal of improving diagnostic accuracy and treatment outcomes for patients with idiopathic scoliosis.

Relevant publications and links:

  1. Khanal, B., Dahal, L., Adhikari, P., & Khanal, B. (2019, October). Automatic Cobb Angle Detection using Vertebra Detector and Vertebra Corners Regression. In International Workshop and Challenge on Computational Methods and Clinical Applications for Spine Imaging (pp. 81-87). Springer, Cham. pdf   

Team: Bishesh Khanal, Bidur Khanal, Arnav Chavan (remote intern, IIT Dhanbad, India), Risav Tiwari (remote intern, IIT Dhanbad, India), Aryan Raj (remote intern, IIT Dhanbad, India)

Research Themes: Transforming Global health with AI (TOGAI)
Project Category: Computer Vision, Machine Learning, Medical Imaging