Multi-Process Remora Enhanced Hyperparameters of Convolutional Neural Network for Lung Cancer Prediction

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Lung cancer prediction is critical for reducing mortality rates.
• Existing methods often suffer from reduced accuracy during prediction.
• The need for advanced techniques in medical image analysis is evident.

Purpose of the Study:

• To develop an optimized deep learning model for accurate lung cancer prediction using CT images.
• To enhance the classification efficiency of lung cancer detection.
• To introduce the Multi-Process Remora Optimized Hyperparameters of Convolutional Neural Network (MPROH-CNN) for this purpose.

Main Methods:

• Utilized open-source CT image databases for lung cancer detection.
• Applied pre-processing techniques (filtering, contrast enhancement) to remove noise.
• Employed feature extraction methods including histogram, texture, and wavelet analysis.
• Developed a hybrid classifier combining Convolutional Neural Network (CNN) with Remora Optimization Algorithm (ROA) for multi-process optimization (structure and hyperparameter).

Main Results:

• The MPROH-CNN model demonstrated superior performance in lung cancer prediction.
• Achieved a high accuracy level of 0.98 in the prediction task.
• Outperformed traditional methods like CNN, CNN-Particle Swarm Optimization (PSO), and CNN-Firefly Algorithm (FA).

Conclusions:

• The MPROH-CNN approach offers a robust and accurate solution for lung cancer prediction from CT scans.
• The integration of ROA with CNN significantly improves prediction accuracy.
• This method holds potential for improving early detection and patient outcomes in lung cancer diagnosis.