|
Original research OPTIMIZING BREAST CANCER CLASSIFICATION THROUGH INNOVATIVE 2-STEP TRANSFER LEARNING APPROACHPages 49-56
Abstract
According to the American Cancer Society’s statistics of breast cancer for 2023, an estimated 297,790 new cases of invasive breast cancer were expected to be diagnosed in U.S. women. As the seriousness can be seen from the stats, the early detection of breast cancer is crucial for effective treatment, exploring the computer aided detection system as alternatives to labor-intensive manual histopathological analysis. This research aims to develop a deep learning model for breast cancer classification. An Innovative 2-step training technique was employed to optimize the transfer learning process, enhancing the effectiveness of the models. Leveraging a modified PatchCamelyon dataset, consisting of 220,025 image samples, the study rigorously evaluated four prominent models: DenseNet121, VGG19, InceptionResNetV2, and Xception. Notably, VGG19 showcases the exceptional performance by achieving the highest accuracy of 96.53% on the test set and 96.58% on the validation set, aiming to refine the model’s performance for accurate breast cancer classification.
Keywords: Breast Cancer Classification, Histopathological Images, Computer Aided Diagnosis (CAD), Medical Image Analysis, Transfer Learning, Deep Learning.
|
Shaurya Jain, 
PDF 