Early detection of breast lesion types is crucial for timely intervention and improved clinical outcomes. However, current diagnostic methods often lack sufficient accuracy and frequently necessitate invasive biopsies for definitive diagnosis. To address this limitation, we proposed a novel approach known as Radio Frequency Time Series Dynamic Profiling (RFTSDP) in 2020. This method analyzes radio frequency (RF) backscatter signals during controlled stimulations, revealing tissue mechanical properties and providing richer information about the dynamic microenvironment. In this study, we developed and implemented a device capable of generating vibrations at varying frequencies. Data were acquired from 11 ex-vivo human breast tissue samples embedded in a healthy breast tissue-mimicking phantom. Ultrafast beamformed data were collected under three conditions: no stimulation, constant force, and vibrational stimulation at various frequencies, using a Supersonic Imagine ultrasound system. Deep learning techniques were employed for automated feature extraction, classification, and grading of the tissue samples into five categories. The performance of the deep learning-based RFTSDP method was compared with conventional machine learning approaches, which involved spectral and nonlinear feature extraction from RF time series, followed by classification using a Support Vector Machine (SVM). Incorporating 65 Hz vibration into the RFTSDP alongside deep learning achieved an accuracy of 99.53% ± 0.47% in classifying and grading breast lesions. Using deep learning for automated feature extraction and breast lesion classification led to increased performance accuracy and a more robust model. Furthermore, RFTSDP combined with deep learning demonstrated a 28.67% improvement in classification accuracy compared to the non-stimulation condition and the analysis of focused raw data. These results confirm that enriching backscatter signals using RFTSDP, particularly when integrated with deep learning, enhances the determination of breast lesion types and facilitates more accurate cancer grading.