A novel and reliable computational intelligence system for breast cancer detection

Area of Science:

• Computational intelligence
• Medical informatics
• Biomedical engineering

Background:

• Breast cancer is a leading cause of morbidity and mortality in women.
• Accurate and reliable detection methods are crucial for effective treatment.
• Existing diagnostic tools can benefit from advanced computational approaches.

Purpose of the Study:

• To propose a hybrid computational intelligence model for enhanced breast cancer detection.
• To integrate unsupervised (SOM) and supervised (CVNN) learning for robust classification.
• To evaluate the model's performance against medical diagnoses.

Main Methods:

• A two-stage hybrid model utilizing Self-Organizing Map (SOM) for patient clustering.
• Complex-Valued Neural Network (CVNN) for classifying clustered patient data into benign or malignant.
• Dataset comprising 822 patients with features including mass shape, margin, density, age, and BI-RADS assessment.
• Performance evaluation using Receiver Operating Characteristic (ROC) analyses and confusion matrix.

Main Results:

• The hybrid SOM-CVNN model demonstrated high detection accuracy.
• Health detection ratio reached 94%, and disease detection ratio reached 95% in the testing phase.
• The model proved superior to existing methods for reliable breast cancer detection.

Conclusions:

• The proposed hybrid computational intelligence model offers a reliable and robust approach for breast cancer detection.
• Integration of SOM and CVNN provides an effective framework for classifying breast cancer severity.
• This model has the potential to significantly aid in clinical breast cancer diagnosis.