Correlated-Weighted Statistically Modeled Contourlet and Curvelet Coefficient Image-Based Breast Tumor Classification...

Area of Science:

• Medical imaging
• Artificial intelligence in medicine
• Biomedical signal processing

Background:

• Accurate breast tumor classification from ultrasound (US) B-mode images is crucial for diagnosis.
• Statistical modeling plays a key role in enhancing image analysis for medical applications.
• Deep learning offers promising avenues for automated classification tasks in medical imaging.

Purpose of the Study:

• To propose a novel deep convolutional neural network (CNN) architecture for breast tumor classification.
• To introduce a new image processing technique combining correlated-weighted transformations with the Rician inverse Gaussian (RiIG) distribution.
• To compare the performance of the proposed method against other statistical models and existing strategies.

Main Methods:

• Development of correlated-weighted contourlet-transformed RiIG (CWCtr-RiIG) and curvelet-transformed RiIG (CWCrv-RiIG) image-based CNN architectures.
• Utilizing the Rician inverse Gaussian (RiIG) distribution for statistical modeling of ultrasound images.
• Weighting contourlet and curvelet sub-band coefficients based on their correlation with RiIG modeled images.
• Comparative analysis with Nakagami and normal inverse Gaussian (NIG) distributions.

Main Results:

• The proposed CWCtr-RiIG approach achieved over 98% accuracy, sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV).
• The method demonstrated superior classification performance compared to other statistical models and existing strategies on three public datasets (Mendeley, UDIAT, BUSI).
• CWCtr-RiIG images were particularly effective in achieving high performance metrics.

Conclusions:

• The proposed CWCtr-RiIG deep learning architecture offers a highly accurate and effective method for breast tumor classification from ultrasound images.
• This novel approach, integrating advanced statistical modeling and image transformation techniques, shows significant potential for improving diagnostic capabilities in breast cancer detection.
• The findings suggest that the RiIG distribution combined with correlated weighting strategies can enhance the performance of deep learning models in medical image analysis.