Effective BCDNet-based breast cancer classification model using hybrid deep learning with VGG16-based optimal feature...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Breast cancer is a leading cause of death in women, necessitating early detection for improved survival.
• Manual diagnosis is time-consuming and subjective; existing CAD models struggle with ultrasound image variability.
• Ultrasound is a cost-effective, portable, and non-invasive imaging modality for breast cancer screening.

Purpose of the Study:

• To develop an effective deep learning model (BCDNet) for accurate breast tumor classification (benign vs. malignant) using ultrasound images.
• To optimize model performance using the RPAOSM-ESO algorithm for enhanced accuracy and reduced false negatives.
• To facilitate early-stage breast cancer detection and reduce mortality rates.

Main Methods:

• The BCDNet model employs transfer learning from VGG16 for feature extraction.
• It incorporates an AHDNAM classification approach: ASPP, DTCN, 1DCNN, and an attention mechanism.
• The RPAOSM-ESO algorithm fine-tunes model weights and parameters.

Main Results:

• The RPAOSM-ESO-BCDNet model achieved an accuracy of 94.5%.
• This performance surpasses previous models like DTCN (88.2%), 1DCNN (89.6%), MobileNet (91.3%), and ASPP-DTC-1DCNN-AM (93.8%).
• The model demonstrates superior accuracy in classifying breast tumors from ultrasound images.

Conclusions:

• The BCDNet model, optimized by RPAOSM-ESO, shows significant promise for accurate breast tumor classification using ultrasound.
• It can serve as a valuable tool for early breast cancer detection.
• The model has the potential to improve patient outcomes and reduce healthcare burdens.