An improved breast cancer classification with hybrid chaotic sand cat and Remora Optimization feature selection...

Area of Science:

• Oncology
• Medical Imaging
• Computational Pathology

Background:

• Breast cancer (BrC) is a leading cancer diagnosis in women, necessitating efficient automated pathology analysis.
• Accurate and rapid identification of breast cancer histological images is crucial for timely prediction and treatment.
• Deep learning models offer enhanced speed and precision over traditional machine learning for breast cancer detection.

Purpose of the Study:

• To demonstrate the viability of an automated system for identifying and classifying breast cancer using histological images.
• To develop and evaluate a novel hybrid optimization and deep learning approach for breast cancer detection.
• To achieve high accuracy in discriminating between cancerous and benign breast tumors.

Main Methods:

• Image pre-processing using Adaptive Switching Modified Decision Based Unsymmetrical Trimmed Median Filter (ASMDBUTMF) for noise reduction.
• Image segmentation via Thresholding Level set approach.
• Hybrid feature selection using chaotic sand cat optimization and Remora Optimization Algorithm (ROA).
• Classification using a Conditional Variation Autoencoder (DL classifier).

Main Results:

• Achieved classification accuracy of 99.4% on the BreakHis dataset.
• Obtained high performance metrics including Precision (99.2%), Recall (99.1%), F-score (99%), and Specificity (99.14%).
• Demonstrated superior performance compared to existing research on the BreakHis dataset.

Conclusions:

• The proposed hybrid optimization and deep learning approach is highly effective for automated breast cancer identification and classification.
• The method offers a promising solution for improving the accuracy and efficiency of breast cancer diagnosis in digital pathology.
• The achieved results indicate a significant advancement in automated breast cancer detection using histological imaging.