Analysis on Machine Learning-Based Early Breast Cancer Detection
Dipali Ghadge, Shrutik Hon, Tushar Saraf, Tejas Wagh, Abhishek Tambe, Yogesh Suresh Deshmukh · 2024
Our study introduces an innovative method for diagnosing breast abnormalities, focusing on differentiating between cancerous and non-cancerous instances using feature extraction and dimensionality reduction from pre-trained Convolutional Neural Network (CNN) models. By concatenating relevant features, we employ machine learning algorithms (SVM, RF, KNN, NN) for classification. The NN-based classifier achieves a remarkable 92% accuracy on the RSNA dataset, attributed to the incorporation of age and multiple views in the updated dataset. Notably, our approach outperforms advanced techniques, demonstrating a 94.5% accuracy on the MIAS dataset and an exceptional 96% accuracy on the DDSM dataset. These findings highlight the method’s efficacy in precisely identifying breast lesions, surpassing existing approaches in terms of sensitivity and accuracy.