A manta ray-bayesian optimization approach for hyperparameter-tuned convolutional neural networks in lung cancer...

Area of Science:

• Medical Image Analysis
• Artificial Intelligence
• Computational Biology

Background:

• Lung cancer classification using deep learning models is challenged by inefficient hyperparameter selection.
• Conventional methods like grid or random search are computationally expensive in high-dimensional spaces.

Purpose of the Study:

• To develop an efficient hybrid Convolutional Neural Network (CNN) for lung cancer image classification.
• To address the computational inefficiency of traditional hyperparameter optimization techniques.

Main Methods:

• A hybrid CNN model integrating Bayesian Optimization (BO) and Manta Ray Foraging Optimization (MRFO) for hyperparameter tuning.
• BO models the objective function using Gaussian Process and Expected Improvement; MRFO refines parameters via foraging mechanisms.

Main Results:

• The proposed hybrid CNN achieved a testing accuracy of 98% in lung cancer image classification.
• This performance surpasses many state-of-the-art models, demonstrating the effectiveness of the optimization strategy.

Conclusions:

• Hybrid metaheuristic-based optimization significantly enhances deep learning model performance in medical image analysis.
• The dual-stage optimization approach balances exploration and exploitation for efficient hyperparameter tuning.