Related Experiment Videos
Breast cancer classification based on advanced multi dimensional fuzzy neural network.
Somayeh Naghibi1, Mohammad Teshnehlab, Mahdi Aliyari Shoorehdeli
1KNT University of Technology, Tehran, Iran. s_naghibi_63@yahoo.com
Journal of Medical Systems
|July 2, 2011
Summary
New fuzzy neural network models effectively detect breast cancer, overcoming the "curse of dimensionality" for earlier diagnosis. These methods offer accurate and interpretable results in classifying benign and malignant lesions.
Area of Science:
- Computational Intelligence
- Medical Informatics
- Biomedical Engineering
Background:
- Breast cancer remains a leading cause of cancer death in women, highlighting the critical need for effective early detection methods.
- Traditional Fuzzy Neural Networks (FNN) face computational challenges due to the "curse of dimensionality" as input numbers increase.
- Developing advanced intelligent systems is crucial for improving decision-making in medical diagnostics.
Purpose of the Study:
- To address the "curse of dimensionality" in Fuzzy Neural Networks (FNN) for breast cancer detection.
- To evaluate the efficacy of Hierarchical Fuzzy Neural Network (HFNN) and Fuzzy Gaussian Potential Neural Network (FGPNN) in classifying breast cancer.
- To compare the performance of HFNN and FGPNN against traditional FNN using a new training algorithm.
Main Methods:
- Application of Hierarchical Fuzzy Neural Network (HFNN) with low-dimensional fuzzy neural networks in a hierarchical structure.
- Implementation of Fuzzy Gaussian Potential Neural Network (FGPNN) utilizing Gaussian Potential Functions (GPF) as membership functions.
- Utilizing a novel training algorithm to apply HFNN and FGPNN to the Wisconsin Breast Cancer Database for lesion classification.
Main Results:
- HFNN and FGPNN demonstrated effectiveness in classifying breast cancer lesions as benign or malignant.
- These models achieved comparable accuracy to traditional FNN but with significantly fewer rules and parameters.
- The proposed methods showed reduced computational complexity, mitigating the "curse of dimensionality".
Conclusions:
- HFNN and FGPNN offer a viable solution to the "curse of dimensionality" in fuzzy neural networks for breast cancer diagnosis.
- These advanced models maintain high accuracy while improving interpretability for human experts.
- The study validates the effectiveness of HFNN and FGPNN in early breast cancer detection and classification.