Related Experiment Videos
Breast cancer diagnosis using self-organizing map for sonography.
1Department of General Surgery, China Medical College and Hospital, Taichung, Taiwan. dlchen88@ms13.hinet.net
Ultrasound in Medicine & Biology
|April 25, 2000
Summary
This study shows that a neural network model called self-organizing maps (SOM) accurately classifies breast lesions from sonograms. The high negative predictive value suggests it could help avoid unnecessary biopsies for benign tumors.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Distinguishing benign from malignant breast lesions on sonography is crucial for patient management.
- Accurate classification can reduce unnecessary invasive procedures.
- Computer-aided diagnosis systems show promise in improving diagnostic accuracy.
Purpose of the Study:
- To evaluate the performance of a self-organizing map (SOM) neural network model for classifying benign and malignant breast lesions using sonographic images.
- To assess the diagnostic capabilities of an SOM-based computer-aided diagnosis (CAD) system.
Main Methods:
- Retrospective analysis of 243 breast tumors (161 benign, 82 malignant) from digitized sonographic images.
- Feature extraction using 24 autocorrelation texture features.
- Classification using a self-organizing map (SOM) neural network.
- Performance evaluation with k-fold cross-validation and receiver operating characteristic (ROC) curves.
Main Results:
- The SOM system achieved an ROC area index of 0.9357 +/- 0.0152.
- Overall accuracy was 85.6%, with high sensitivity (97.6%) and negative predictive value (98.5%).
- Specificity was 79.5% and positive predictive value was 70.8%.
Conclusions:
- The developed SOM-based computer-aided diagnosis system demonstrates high performance in classifying sonographic breast lesions.
- The system's high negative predictive value indicates its potential utility in reducing unnecessary biopsies for benign breast conditions.