A machine learning approach to assess changes in scoliosis

Area of Science:

• Medical Imaging
• Machine Learning
• Orthopedics

Background:

• Scoliosis management requires accurate measurement of spinal curve changes.
• Automated systems offer potential for efficient measurement but require validation.
• Assessing the reliability of automated measurements is crucial for clinical decision-making.

Purpose of the Study:

• To develop and evaluate a machine learning approach for validating automated scoliosis measurements.
• To assess the accuracy of different machine learning classifiers in differentiating valid from invalid results.

Main Methods:

• A dataset of 141 vertebral endplate inclinations from spine radiographs of scoliosis patients was used.
• Three machine learning classifiers were trained: Support Vector Classifier (SVC), Decision Tree (DT), and Logistic Regression (LR).
• Classifier performance was evaluated on a separate test set, focusing on accuracy in discriminating results with <3 degrees error.

Main Results:

• The Support Vector Classifier (SVC) achieved the highest accuracy at 86% in distinguishing 'Good Results' (error < 3 degrees) from 'Bad Results'.
• Logistic Regression (LR) achieved 76% accuracy, and Decision Tree (DT) achieved 68% accuracy.
• The machine learning approach successfully differentiated between good and bad results from the automated system.

Conclusions:

• Machine learning, particularly SVC, can effectively validate automated measurements of scoliotic curves.
• This approach enhances the reliability of automated systems used in scoliosis assessment.
• The findings support the use of machine learning for quality control in medical imaging analysis for scoliosis.