Using Spectral Clustering to Segment Prostate Images
Author Information
Author(s): Mohamed Samar, Salama Magdy MA
Primary Institution: University of Waterloo
Hypothesis
Can spectral clustering effectively segment prostate gland images from TRUS without user interaction?
Conclusion
The spectral clustering algorithm provides fast and accurate estimates for prostate volume and internal gland segmentation without requiring user input.
Supporting Evidence
- The algorithm achieved 93% average overlap areas compared to expert radiologist segmented images.
- The method does not require any prior knowledge or user interaction for segmentation.
- It effectively segments both the prostate gland and internal regions, including cancerous areas.
Takeaway
This study shows a new way to look at prostate images that helps doctors see important details without needing to draw outlines by hand.
Methodology
The study used spectral clustering based on graph theory to segment prostate images from TRUS without requiring initial contours or user input.
Limitations
The study acknowledges that more images are needed for generalization and that the algorithm may not perform well with atypical prostate shapes.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website