Detecting and Segmenting Vertebrae in X-Ray Images Using Heterogeneous Computing
Author Information
Author(s): Lecron Fabian, Mahmoudi Sidi Ahmed, Benjelloun Mohammed, Mahmoudi Saïd, Manneback Pierre
Primary Institution: University of Mons
Hypothesis
Can a parallel hybrid implementation improve the efficiency of vertebra detection and segmentation in X-ray images?
Conclusion
The proposed method significantly enhances the speed and accuracy of vertebra segmentation in X-ray images using a hybrid computing approach.
Supporting Evidence
- The method achieved a global speedup ranging from 3 to 22 times compared to CPU implementations.
- Segmentation errors were measured in pixels, with a mean error of approximately 2.90 px.
- The study utilized a sample of 51 radiographs from the NHANES II database.
Takeaway
This study shows how computers can help doctors quickly find and outline bones in X-ray pictures, making it easier to check for problems.
Methodology
The study used a parallel hybrid implementation combining CPU and GPU processing to enhance vertebra segmentation from X-ray images.
Limitations
The method may struggle with images where vertebrae are closely merged, leading to potential misidentification.
Participant Demographics
The sample consisted of X-ray images from patients aged 25 to 74, focusing on cervical vertebrae C3 to C7.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website