Qualitative Classification of Proximal Femoral Bone Using Geometric Features and Texture Analysis in Collected MRI...

Area of Science:

• Biomedical Imaging
• Machine Learning
• Orthopedics

Background:

• Osteoporosis and related fractures are significant health concerns.
• Accurate assessment of proximal femoral bone (PFB) health is crucial for fracture risk prediction.
• Current diagnostic methods may benefit from advanced analytical techniques.

Purpose of the Study:

• To differentiate between healthy and unhealthy proximal femoral bone (PFB) using geometric and texture features from MRI.
• To identify the most influential features for classifying PFB health.
• To evaluate the performance of machine learning models in this classification task.

Main Methods:

• Acquired MRI and dual-energy X-ray absorptiometry (DEXA) data from 284 participants.
• Extracted 204 geometric and texture features from PFB MRI scans.
• Employed machine learning algorithms including Support Vector Machine (SVM), decision tree, and logistic regression.
• Utilized a Genetic Algorithm (GA) for feature selection to optimize classification accuracy.

Main Results:

• The Support Vector Machine (SVM) with radial basis function kernel achieved the highest classification performance at 89.08%.
• Geometric features were identified as the most influential in discriminating between healthy and unhealthy PFB.
• The study successfully classified 185 participants as healthy and 99 as unhealthy based on DEXA T-scores.

Conclusions:

• Machine learning, particularly SVM with geometric features from MRI, shows high potential for classifying proximal femoral bone (PFB) health.
• This approach offers a novel, non-invasive method for qualitative assessment of PFB.
• Further research with larger cohorts is recommended to validate and refine the model.