Local-Entropy Based Approach for X-Ray Image Segmentation and Fracture Detection

Area of Science:

• Medical Imaging
• Radiology
• Computer-Aided Diagnosis

Background:

• Fracture detection in pediatric radiography can be challenging, particularly for subtle fractures.
• Accurate and timely diagnosis of fractures is crucial to prevent long-term complications.

Purpose of the Study:

• To develop and validate a novel, rotation-invariant technique for segmenting and classifying fractures in X-ray images.
• To improve the detection of small, difficult-to-visualize fractures in pediatric ulna and radius bones.

Main Methods:

• A novel rotation-invariant method utilizing local Shannon entropy for de-noising and tissue removal.
• Image segmentation based on entropy representation, followed by graph theory for contour refinement.
• Fracture classification by comparing extracted bone contours with ideal healthy contours.

Main Results:

• The method demonstrated high efficiency and robustness on a dataset of 860 X-ray images.
• Achieved segmentation quality and classification accuracy up to 91.16% and precision up to 86.22%.
• Successfully detected small fractures often missed by visual inspection.

Conclusions:

• The proposed hybrid method offers a significant advancement in computer-aided fracture detection systems.
• A computerized warning system is essential for radiologists to prevent false-negative diagnoses.
• The technique shows promise for enhancing diagnostic accuracy in pediatric orthopedics.