Classification of Cancer Cells from Mammogram Images Based on GoogLeNet and ResNet-50 Deep Learning Models

Breast cancer remains one of the most common and deadly forms of cancer affecting women globally. Early detection through mammogram screening plays a vital role in improving survival rates by enabling timely diagnosis and treatment. This study explored the application of deep learning techniques, particularly GoogLeNet and ResNet-50 architectures, to classify mammogram images into benign and malignant categories. The dataset, sourced from Kaggle, consisted of 479 mammogram images. These images were preprocessed and divided into training and testing sets, with 90% allocated for training and 10% for testing. Both GoogLeNet and ResNet-50 models were trained and tested, yielding different performance metrics. GoogLeNet achieved the highest testing accuracy of 70.83%, while ResNet-50 attained 68.75% accuracy. GoogLeNet demonstrated better sensitivity at 92.85%, but ResNet-50 outperformed in terms of precision at 69.69%. Although both models displayed competitive results, GoogLeNet exhibited superior performance in classifying cancer cells based on sensitivity, making it more effective in identifying true positives. These findings demonstrated the potential of deep learning models to enhance breast cancer detection and classification using mammogram images, contributing to improved diagnostic accuracy and patient outcomes.