Development of a Machine Learning Algorithm to Correlate Lumbar Disc Height on X-rays with Disc Bulging or Herniation

Area of Science:

• Orthopedics
• Radiology
• Medical Informatics

Background:

• Lumbar disc bulging or herniation (LDBH) is a primary cause of spinal stenosis and nerve compression, often dictating the need for spine surgery.
• Magnetic Resonance Imaging (MRI) is crucial for surgical evaluation but faces accessibility limitations.
• Spinal X-rays offer rapid assessment of bony structures, presenting an alternative diagnostic avenue.

Purpose of the Study:

• To identify factors associated with LDBH using lumbar X-ray findings.
• To develop a clinical diagnostic tool for LDBH prediction based on X-ray data.
• To leverage machine learning for enhanced diagnostic capabilities.

Main Methods:

• Analysis of 458 patients with collected clinical and imaging variables.
• Application of five machine learning methods: LASSO regression, MARS, decision tree, random forest, and extreme gradient boosting.
• Identification of key predictors for LDBH from lumbar spine X-rays.

Main Results:

• L4-5 posterior disc height, patient age, and L1-2 anterior disc height emerged as the most significant predictors of LDBH.
• A decision tree algorithm was successfully constructed to aid clinical decision-making.
• Machine learning effectively identified crucial variables from lumbar X-ray data.

Conclusions:

• Lumbar X-rays, combined with machine learning, show potential for diagnosing LDBH.
• The study highlights the predictive value of L1-2 disc height for LDBH.
• The developed decision tool can support surgeons in evaluating LDBH and surgical necessity.