A Novel Breast Cancer Diagnosis Scheme With Intelligent Feature and Parameter Selections

Area of Science:

• Computational intelligence
• Biomedical engineering
• Machine learning for healthcare

Background:

• Breast cancer diagnosis relies on early detection for improved patient survival rates.
• Manual diagnosis methods are prone to human error, inaccuracies, and time constraints.
• Artificial Neural Networks (ANNs) offer powerful automated diagnosis capabilities.

Purpose of the Study:

• To propose two novel hybrid Artificial Bee Colony (ABC) optimization algorithms, HABC and Hybrid ABC.
• To enhance the exploration and exploitation capabilities of standard ABC algorithms.
• To improve automated breast cancer diagnosis through optimized ANN models.

Main Methods:

• Hybridizing ABC with Artificial Immune System (HABC) and Bacterial Foraging Optimization (Hybrid ABC).
• Applying HABC and Hybrid ABC for concurrent feature selection and ANN parameter optimization.
• Utilizing back-propagation algorithms (Resilient Back-propagation, Levenberg-Marquardt, Gradient Descent) for ANN tuning.

Main Results:

• HABC-RP achieved 99.14% accuracy; Hybrid ABC-RP reached 99.54% accuracy.
• Hybrid ABC-RP demonstrated high sensitivity (99.08%), specificity (99.81%), and precision (99.38%).
• The proposed methods resulted in high accuracy with low complexity ANN structures.

Conclusions:

• The Hybrid ABC-RP algorithm offers superior performance in breast cancer diagnosis with high accuracy and low complexity.
• Concurrent feature selection and ANN parameter tuning are crucial for accurate breast cancer diagnosis.
• The developed hybrid algorithms outperform existing breast cancer diagnosis systems and show future potential for tumor segmentation and classifier tuning.