Deep Learning Based Capsule Neural Network Model for Breast Cancer Diagnosis Using Mammogram Images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Vision

Background:

• Breast cancer is a prevalent disease globally, necessitating accurate and early detection methods.
• Mammography is a key screening tool, but interpretation accuracy relies heavily on radiologist expertise and image quality.
• Advancements in Deep Learning (DL) and Computer Vision offer promising solutions for automated breast cancer diagnosis.

Purpose of the Study:

• To develop and evaluate a novel Optimal Multi-Level Thresholding-based Segmentation with DL enabled Capsule Network (OMLTS-DLCN) model for breast cancer diagnosis using digital mammograms.
• To enhance the accuracy and efficiency of breast cancer detection through an integrated segmentation and classification approach.

Main Methods:

• The OMLTS-DLCN model incorporates Adaptive Fuzzy based median filtering (AFF) for noise reduction in mammograms.
• Optimal Kapur's based Multilevel Thresholding with Shell Game Optimization (OKMT-SGO) is utilized for precise breast cancer segmentation.
• A CapsNet (Capsule Network) acts as a feature extractor, coupled with a Back-Propagation Neural Network (BPNN) for classification.

Main Results:

• The OMLTS-DLCN model demonstrated superior performance on benchmark Mini-MIAS and DDSM datasets.
• Achieved high diagnostic accuracy rates of 98.50% on the Mini-MIAS dataset and 97.55% on the DDSM dataset.
• The proposed method effectively segments and classifies breast cancer abnormalities.

Conclusions:

• The OMLTS-DLCN model presents a highly effective approach for automated breast cancer diagnosis from mammograms.
• The integration of advanced segmentation and DL techniques significantly improves diagnostic accuracy and aids in early detection.
• This model holds potential for enhancing breast cancer screening programs and improving patient outcomes.