Related Experiment Videos
Content-based ultrasound image retrieval using a coarse to fine approach
Dong-Min Kwak1, Bum-Soo Kim, Ock-Kyung Yoon
1School of Electrical Engineering and Computer Science, Kyungpook National University, Taegu, Korea. imis@palgong.knu.ac.kr
Annals of the New York Academy of Sciences
|February 21, 2003
Summary
This study introduces a two-stage content-based image retrieval (CBIR) method for clinical ultrasound images, improving diagnostic efficiency. The novel approach enhances medical image database searching capabilities.
Area of Science:
- Medical Imaging
- Computer Science
- Information Retrieval
Background:
- Picture Archiving and Communication Systems (PACS) require advanced retrieval for multimedia data.
- Efficient retrieval is crucial for medical diagnostics and pathology studies.
- Existing methods struggle with the complexity of clinical image databases.
Purpose of the Study:
- To develop a novel content-based image retrieval (CBIR) algorithm for clinical ultrasound image databases.
- To enhance the efficiency and accuracy of retrieving relevant medical images.
- To address the limitations of current retrieval technologies in PACS.
Main Methods:
- A two-stage retrieval process was designed for maximizing efficiency.
- Coarse retrieval utilizes statistical characteristics of wavelet coefficients, reducing the search space by up to 70%.
- Fine retrieval employs Legendre moments of the global histogram probability density function (pdf) on the pre-retrieved set.
Main Results:
- The proposed algorithm demonstrated promising results on an abdominal ultrasound image database.
- It significantly improved retrieval efficiency compared to other methods.
- The two-stage approach effectively narrows down search results.
Conclusions:
- The developed CBIR algorithm is effective for clinical ultrasound image retrieval.
- This method offers a viable solution for enhancing PACS capabilities in medical applications.
- Further application in pathology and diagnostics is supported by the promising results.