Deep Learning for Breast Cancer Detection: Comparative Analysis of ConvNeXT and EfficientNet

TL;DR

This study compares ConvNeXT and EfficientNet convolutional neural networks for breast cancer detection in mammograms, finding ConvNeXT outperforms EfficientNet in accuracy, AUC, and F-score on a large screening dataset.

Contribution

It provides a comparative analysis of ConvNeXT and EfficientNet models specifically for mammogram-based breast cancer detection, highlighting ConvNeXT's superior performance.

Findings

ConvNeXT achieves 94.33% AUC, 93.36% accuracy, 95.13% F-score.

EfficientNet achieves 92.34% AUC, 91.47% accuracy, 93.06% F-score.

ConvNeXT outperforms EfficientNet in all evaluated metrics.

Abstract

Breast cancer is the most commonly occurring cancer worldwide. This cancer caused 670,000 deaths globally in 2022, as reported by the WHO. Yet since health officials began routine mammography screening in age groups deemed at risk in the 1980s, breast cancer mortality has decreased by 40% in high-income nations. Every day, a greater and greater number of people are receiving a breast cancer diagnosis. Reducing cancer-related deaths requires early detection and treatment. This paper compares two convolutional neural networks called ConvNeXT and EfficientNet to predict the likelihood of cancer in mammograms from screening exams. Preprocessing of the images, classification, and performance evaluation are main parts of the whole procedure. Several evaluation metrics were used to compare and evaluate the performance of the models. The result shows that ConvNeXT generates better results with a 94.33% AUC score, 93.36% accuracy, and 95.13% F-score compared to EfficientNet with a 92.34% AUC score, 91.47% accuracy, and 93.06% F-score on RSNA screening mammography breast cancer dataset.