An Improved Fusion Model from GoogLeNet and AlexNet to Predict Breast Cancer using Deep Learning
R. Sathishkumar, M Nirmalraj, B Vinothini, N Rajasri, E Sivasakthi
Abstract
Despite the fact that breast cancer is one of the most common cancers to kill women due to its high mortality and ease of recurrence, an early and precise diagnosis can significantly improve the chances of a successful recovery. It is very important to diagnose the disease at beginning level to avoid any complications. Traditional early diagnosis, however, is inaccurate and depends on human experience. Since prediction accuracy and efficiency may be improved, numerous researchers have suggested several machine learning techniques. Deep convolutional neural networks like GoogLeNet and AlexNet models are chosen for the diagnosis of breast cancer because they are excellent at extracting deep characteristics from images and performing fantastic image categorization. Additionally, the GoogLeNet model and AlexNet model is integrated to form a fused model which is enhanced to make better accuracy, more specifically aimed at medically related digital images. During the diagnosis our aim is to achieve the higher level accuracy, with reduced computing weight. The enhanced model, which is more focused on breast cancer pathological sections and increases the performance, by combining the properties of two different types of network structures. The integrated proposed model achieved 97.5% Precision, 97.3% Recall, 98.3% Specificity, 98.3% Accuracy, and 98.1% F1-Score. The improved model may more precisely identify Breast cancer, reduce the likelihood of inaccurate diagnosis and delayed diagnosis due to physician personal reasons, and assist healthcare professionals in providing patient care and monitoring, making the entire evaluation and therapy process more sophisticated and effective.