BI-TLM: Bilinear Interpolation with Transfer Learning Model for Breast Cancer Classification

Breast cancer is another fastest growing reason for death in women, trailing only lung cancer. Breast cancer mortality can be reduced if it is discovered early and treated. Because conventional breast cancer assessment needs much longer, an automated method needs to be invented for early cancer detection. This research provides a network for detecting breast cancer from ultrasound images which uses deep neural networks and a trainable sub-layer network to select the best features. Because ultrasound images contain noise and artifacts, a bilinear filtering method is considered in the preprocessing step to reduce the noise and artifacts. The DenseNet201 model was used in our proposed network as a deep learning-based technique since It has the potential to disperse feature strength, stimulate feature reuse, and significantly decrease the parameters. The proposed system along with other models were trained and evaluated on a balanced ultrasound dataset divided into benign, malignant, and normal classes. Each model’s performance was evaluated utilizing several accuracy metrics. Among the models examined, the suggested model outperformed the others, suggesting its potential for accurately diagnosing breast cancer from ultrasound images. The discoveries of the work contribute to medical imaging and emphasize the efficacy of the DenseNet201 model.