Radiomics for classification of bone mineral loss: A machine learning study.

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Medicine
• Biomedical Engineering

Background:

• Osteoporosis and osteopenia are significant public health concerns, necessitating accurate and early diagnostic methods.
• Current diagnostic tools may have limitations, driving the need for novel approaches in bone mineral deficiency assessment.

Purpose of the Study:

• To develop and evaluate predictive models for classifying osteoporosis, osteopenia, and normal bone density.
• To utilize radiomics and machine learning techniques on bone mineral densitometry (BMD) images for classification.

Main Methods:

• A retrospective study included 147 patients (12 men, 135 women).
• Seven regions (four lumbar, three femoral) were segmented on BMD images, and 54 texture features were extracted.
• Evaluated feature selection methods (CLAE, ORAE, GRAE, PRCA) and classifiers (RF, RC, KN, LB) for classifying bone density states.

Main Results:

• Area Under the Curve (AUC) values ranged from 0.50 to 0.78.
• The highest performance (AUC=0.78) was achieved by Random Forest (RF) with Classifier Attribute Evaluation (CLAE) or One Rule Attribute Evaluation (ORAE), and Random Committee (RC) with ORAE, distinguishing osteoporosis from normal in the trochanteric region.
• RF with Principal Components Analysis (PRCA) and Logit-Boost (LB) with PRCA achieved the highest performance (AUC=0.76) in differentiating osteoporosis from normal in the femoral neck region.

Conclusions:

• Radiomics combined with machine learning offers a novel method for classifying bone mineral deficiency diseases.
• The study demonstrates the potential of using BMD image features for improved diagnostic classification.