Recent studies employing Deep Convolutional Neural Network models for automated breast cancer detection have yielded promising results, yet two critical challenges persist. Firstly, in supervised contexts, Deep Learning-based models necessitate extensive datasets and comprehensive training processes to achieve accurate diagnostic outcomes. Secondly, the predictive performance of existing models for breast cancer diagnosis is significantly influenced by their complexity and depth, particularly the number of convolutional layers. As a result, deploying these models in resource-constrained environments, such as handheld screening devices for real-time medical analysis, becomes impractical. To address these challenges, a simplified Convolutional Neural Network model incorporating Depthwise Convolution Layers, conventional convolution layers, and edge detection filters for automated breast cancer diagnosis is proposed. The proposed model is efficient and compact, containing only 0.18 million parameters and occupying less than 2 MB in size. Experimental validation on the publicly available DMR-IR dataset demonstrated the superior performance of the proposed model compared to significantly heavier state-of-the-art approaches.