Performance Analysis of Convolutional Neural Network Models for Breast Cancer Diagnosis.

Moctar Tache Mouhamed, Andronicus A. Akinyelu, Dinna Ranirina, Berthine N. Mpinda

Keywords: Breast Cancer, Convolutional Neural Network, DenseNet121, ResNet50, Transfer learning, Mammography images.

TL;DR: This study develops a CNN-based model for breast cancer diagnosis, achieving 96.20% accuracy.

Abstract: Breast cancer remains one of the leading causes of mortality among women. This study introduces a CNN-based model for breast cancer diagnosis and provides a comprehensive performance analysis, comparing it with two state-of-the-art pre-trained models: DenseNet121 and ResNet50. The models were trained on a dataset comprising three classes normal, benign, and malignant breast cancer images. To enhance the dataset and improve model generalization, data augmentation techniques were applied, increasing the total number of images to over 9,000. The study evaluates the models’ effectiveness in a multi-class classification setting, with results showing that the CNN model, ResNet50, and DenseNet121 achieved classification accuracies of 96.20%, 97.74%, and 93.94%, respectively. These findings highlight the proposed CNN model’s potential for medical applications, offering an optimal balance between computational efficiency, high accuracy, and minimal false positive and false negative rates.

Submission Number: 8