Breast cancer in women is a foremost global public health issue, and numerous cases are not identified until the cancer is very advanced. Globally, breast cancer is the second most prominent cause of mortality for women. To save a life, appropriate treatment planning for breast cancer requires early and precise assessment. A non-invasive imaging technique called ultrasound (US) helps to identify breast abnormalities and track the health of cancer patients. Due to its high operator dependency, it exhibits more false negatives despite having the highest sensitivity for identifying breast masses. Breast lesion treatment holds up in remote regions because there is insufficient US competence to diagnose breast lesions. Deep learning neural networks could assist doctors in making early decisions by accurately and quickly identifying patients and tracking their prognosis. Authors have employed a variety of pre-trained deep-learning models to detect breast tumors; however, the effectiveness of these models is unclear. In this study, various deep-learning methods are employed to accurately classify breast cancer cells for further diagnoses. The adopted deep learning models used, include a basic custom CNN model, VGG16, ResNet50, EfficientNet-B0, and Vision Transformer. The models have been trained and tested on a benchmark dataset with varying complexities and characterizations. The simulation results show that when compared to other strategies, the recommended approach works better.