MGWO-CNN: hyperparameter optimization of CNN classifier for cervical cancer detection using Modified Grey Wolf...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Cervical cancer is a significant global health issue with increasing mortality rates.
• Current detection methods using Convolutional Neural Network (CNN) models often suffer from overfitting and gradient vanishing.
• Efficient and accurate diagnostic tools are crucial for reducing cervical cancer deaths.

Purpose of the Study:

• To propose a Modified Grey Wolf Optimizer-based Convolutional Neural Network (MGWO-CNN) for improved cervical cancer detection.
• To address limitations of traditional CNN models like overfitting and poor convergence.
• To enhance the accuracy and reliability of cervical cancer diagnosis using AI.

Main Methods:

• Developed a Modified Grey Wolf Optimizer-based Convolutional Neural Network (MGWO-CNN) integrating chaos theory and differential evolution mutation.
• Utilized metaheuristic techniques to optimize CNN hyperparameters for effective model training.
• Extracted key features from cervical Pap smear images for outcome prediction.

Main Results:

• The MGWO-CNN model demonstrated remarkable effectiveness in detecting cervical cancer.
• Achieved high performance metrics on the Herlev and SIPaKMeD datasets.
• Outperformed existing methods with 99.45% accuracy, 100% precision, 97.96% sensitivity, and 100% specificity.

Conclusions:

• The proposed MGWO-CNN model offers a superior approach for cervical cancer detection.
• The optimized model provides accurate and reliable predictions from Pap smear images.
• This AI-driven method has the potential to significantly reduce cervical cancer mortality rates.