Related Experiment Videos
Peripheral blood smear image analysis: A comprehensive review.
Emad A Mohammed1, Mostafa M A Mohamed2, Behrouz H Far1
1Department of Electrical and Computer Engineering, Schulich School of Engineering, University of Calgary, Calgary, Alberta, Canada.
Journal of Pathology Informatics
|May 21, 2014
Summary
Automated analysis of peripheral blood smear images using machine learning algorithms like artificial neural networks (ANNs) and support vector machines (SVMs) can overcome the limitations of manual examination, improving efficiency and consistency in laboratory diagnostics.
Area of Science:
- Medical Laboratory Science
- Computational Pathology
- Artificial Intelligence in Medicine
Background:
- Manual peripheral blood smear examination is a standard laboratory procedure.
- Manual analysis is labor-intensive, time-consuming, and prone to inter-observer variability.
- Automation is needed to enhance efficiency and reliability in blood smear analysis.
Purpose of the Study:
- To explore automated methods for peripheral blood smear image analysis.
- To address the challenges of image segmentation, feature extraction, and pattern classification.
- To improve the accuracy and consistency of laboratory diagnostics.
Main Methods:
- Image segmentation using Support Vector Machines (SVM) and Artificial Neural Networks (ANNs).
- Feature extraction and selection to identify distinguishing characteristics of cellular components.
- Pattern classification employing supervised and unsupervised learning algorithms (ANN, SVM, decision tree, K-nearest neighbor).
Main Results:
- Development of algorithms for automated peripheral blood smear image analysis.
- Demonstration of SVM and ANNs as effective methods for image segmentation.
- Exploration of various classification algorithms for accurate pattern recognition.
Conclusions:
- Automated analysis of peripheral blood smears offers a promising solution to the drawbacks of manual methods.
- Machine learning approaches, including ANNs and SVMs, are crucial for accurate image segmentation and classification.
- Combining multiple classifiers can potentially enhance diagnostic discrimination in hematology.