Related Experiment Video
Updated: May 29, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
An unsupervised strategy for biomedical image segmentation
Roberto Rodríguez1, Rubén Hernández
1Digital Signal Processing Group, Institute of Cybernetics, Mathematics, and Physics, Havana, Cuba.
Advances and Applications in Bioinformatics and Chemistry : AABC
|September 16, 2011
Summary
This study introduces an unsupervised method for biomedical image segmentation using mean shift filtering and entropy. The novel approach achieves high accuracy, with minimal false positives and no false negatives in real-world applications.
Area of Science:
- Medical Imaging
- Computer Vision
- Biomedical Engineering
Background:
- Image segmentation is crucial in biomedical applications.
- Existing methods are often application-specific or require prior knowledge.
- Unsupervised segmentation methods offer greater flexibility but are more challenging.
Purpose of the Study:
- To develop an unsupervised strategy for biomedical image segmentation.
- To utilize mean shift filtering and entropy as a stopping criterion.
- To evaluate the proposed method against manual segmentation.
Main Methods:
- An unsupervised segmentation algorithm based on recursively applying mean shift filtering.
- Entropy is employed as the stopping criterion for the segmentation process.
- Validation performed on diverse real biomedical images.
Main Results:
- The proposed unsupervised strategy demonstrated effectiveness in biomedical image segmentation.
- Achieved less than 20% for false positives.
- Attained 0% for false negatives when compared to manual segmentation.
Conclusions:
- The developed unsupervised method provides an accurate and efficient solution for biomedical image segmentation.
- The strategy overcomes limitations of supervised and application-specific techniques.
- Offers a robust alternative for segmenting complex biomedical images.
