Enhancing breast cancer classification using a deep sparse wavelet autoencoder approach.

As digital imaging technology advances, accurate classification of 2D breast cancer images becomes increasingly crucial for early detection and staging. This paper introduces a novel classification approach that integrates deep learning, sparse coding, and wavelet networks through a unique architecture we call the Deep Sparse Wavelet Autoencoder (DSWAE). The key innovation of our method lies in its construction: DSWAE combines stacked wavelet autoencoders to create a robust model specifically designed for differentiating between distinct categories in 2D breast cancer image datasets. This architecture not only enhances classification accuracy but also optimizes computational efficiency by utilizing deep networks with minimal parameters, which significantly reduces processing time and costs. Our experimental results demonstrate the superior performance of the DSWAE model, achieving precision rates of 94.5% for benign and 93.8% for malignant cases, with recall rates of 93.65% for benign and 96.2% for malignant cases. Remarkably, our method attained a perfect precision rate of 100% for normal cases. These results highlight the effectiveness of our approach, which outperforms current state-of-the-art methods in 2D breast cancer image classification.