Intelligent Breast Cancer Image Classifier using Convolutional Networks and Ant Lion Optimizer

Abstract

Ultrasound images are key tools in the diagnosis of breast cancer, as they allow the assessment of the pathophysiological state of breast tissue. However, their interpretation requires the intervention of specialists and can present diagnostic challenges. This work proposes an intelligent system for the automatic classification of breast ultrasound images, using convolutional neural networks (CNNs) optimized by metaheuristic algorithms. Two architectures were implemented: a standard CNN and the pre-trained ResNet50 network. Hyperparameter optimization was performed using the Ant Lion Optimizer (ALO) metaheuristic algorithm, with the aim of improving classification accuracy. Experimental results show a significant improvement in the performance of both models: the CNN's accuracy increased from 56.57% to 81.57%, while ResNet50 improved from 86.84% to 92.00%. These results demonstrate the potential of combining deep learning architectures with evolutionary optimization techniques for image-assisted diagnostic tasks, especially in the context of breast cancer.