A Deep Hybrid System for Effective Diagnosis of Breast Cancer
Adyasha Sahu, Pradeep Kumar Das, Sukadev Meher · 2024
Breast cancer is still a major global health concern, which emphasizes the need for reliable and effective screening techniques. This work provides a deep learning–based breast cancer detection system that leverages the potent combination of ShuffleNet for feature extraction and Random Forest for classification. In the initial stage of the pipeline, ShuffleNet, a lightweight deep convolution neural network (CNN) architecture, is used to extract features. ShuffleNet is perfect for large-scale medical image analysis because of its effective design, which provides faster processing while maintaining discriminative information. The collected features are then fed into the widely used random forest classifier. The collected features are then sent into a random forest classifier, a well-liked ensemble learning technique that excels at handling high-dimensional data and robustness. The proposed method makes use of ultrasound (BUSI) and mammography (mini-DDSM) datasets for training and evaluation in order to improve the precision and dependability of breast cancer diagnosis. The efficiency of the proposed deep learning–based breast cancer detection system is shown by experimental findings. It displays the best performance among the comparing schemes with a malignancy identification accuracy of 97.50% and 92.31% in mini-DDSM and BUSI datasets.