Anatomical Features of the Sacroiliac Joint and Machine Learning-Based Classification of Disease Types

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Distinguishing inflammatory from degenerative sacroiliac joint (SIJ) disorders is crucial for effective treatment.
• Understanding structural differences in the SIJ is key to accurate diagnosis.

Purpose of the Study:

• To evaluate disease-related morphological patterns and morphometric characteristics of the SIJ.
• To apply machine learning models for classifying inflammatory, degenerative, and control SIJ groups based on imaging features.

Main Methods:

• Retrospective analysis of 209 individuals' MRI scans (418 SIJs).
• Assessment of SIJ morphology (joint surface, erosion, sclerosis, inflammation) and morphometrics (joint space, joint length).
• Classification using machine learning models (SVM, XGBoost) and deep neural networks with 5-fold cross-validation.

Main Results:

• Degenerative group showed significantly higher mean age.
• Both inflammatory and degenerative groups had significantly narrower SIJ spaces compared to controls.
• Machine learning models (SVM, XGBoost) achieved high accuracy (0.9518) and macro-F1 scores (0.9509) in classification.

Conclusions:

• MRI-based assessment of SIJ morphological and morphometric changes reliably differentiates inflammatory and degenerative disorders.
• Machine learning models effectively interpret these features for consistent and objective diagnosis.
• Detailed examination of anatomical features is vital for accurate SIJ disorder diagnosis.