Precision Through Detail: Radiomics and Windowing Techniques as Key for Detecting Dens Axis Fractures in CT Scans.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Radiology

Background:

• Dens axis fractures are critical injuries requiring accurate detection.
• Computed tomography (CT) imaging is standard for diagnosing these fractures.
• Current detection methods can be limited by image quality and interpretation variability.

Purpose of the Study:

• To evaluate the impact of advanced windowing techniques on CT image analysis.
• To compare a pure deep learning (DL) model with a combined DL segmentation, windowing, and radiomics approach for dens axis fracture detection.
• To assess the diagnostic performance of computational models in identifying upper cervical spine fractures.

Main Methods:

• Retrospective analysis of 366 patient CT datasets.
• Development of two models: a pure DL model (CNN/FNN) and a hybrid model (U-Net segmentation, radiomics, ML classifier).
• Evaluation of classification accuracy, focusing on windowing parameters and ML strategies.

Main Results:

• The pure DL model (M1) achieved a maximum classification accuracy of 93.7%.
• The hybrid model (M2) using ROI-based windowing and radiomics reached up to 95.7% accuracy.
• The combined approach demonstrated superior performance in dens axis fracture detection.

Conclusions:

• Integrating advanced windowing, U-Net segmentation, and radiomics enhances dens axis fracture detection in CT imaging.
• This hybrid computational approach shows potential for improving diagnostic accuracy.
• Further clinical integration and exploration of this method are warranted.