A New Weighted Deep Learning Feature Using Particle Swarm and Ant Lion Optimization for Cervical Cancer Diagnosis on...

Area of Science:

• Biomedical Imaging
• Computational Pathology
• Machine Learning in Healthcare

Background:

• Cervical cancer is a major global health concern for women.
• Early detection through improved screening, like Pap smear image analysis, is crucial for reducing mortality.
• Existing methods often focus on whole images or nuclei, limiting diagnostic scope.

Purpose of the Study:

• To compare the efficacy of analyzing entire cells, cytoplasm, or nucleus regions for cervical cancer classification.
• To develop an efficient computer-aided diagnosis system for cervical cancer using advanced machine learning techniques.
• To enhance the accuracy and reliability of early cervical cancer detection.

Main Methods:

• Image augmentation was applied to address imbalanced data.
• Automated feature extraction was performed using AlexNet, DarkNet 19, and NasNet.
• Principal Component Analysis (PCA) reduced features, followed by feature weighting.
• Optimization was conducted using Ant Lion Optimization (ALO) and Particle Swarm Optimization (PSO).
• Support Vector Machine (SVM) and Random Forest (RF) classifiers were employed for classification.

Main Results:

• The SVM classifier achieved 99.5% accuracy for seven cervical cancer classes using PSO optimization.
• The RF classifier achieved 98.9% accuracy.
• Analysis focusing on tissue regions demonstrated superior performance compared to nucleus-only analysis.
• The developed system aids physicians in diagnosing precancerous and early-stage cervical cancer.

Conclusions:

• Analyzing tissue regions in Pap smear images offers a more effective approach for cervical cancer detection than nucleus-only analysis.
• The proposed computer-aided diagnosis system, optimized with PSO and SVM, shows high accuracy and potential for clinical application.
• Further improvements can be achieved with larger datasets for enhanced cervical cancer screening.