A New GLLD Operator for Mass Detection in Digital Mammograms.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Machine Learning

Background:

• Computer-aided diagnosis (CAD) systems are crucial for analyzing digital mammograms.
• Accurate detection of masses in mammograms remains a significant challenge in breast cancer screening.
• Existing texture analysis methods like Local Binary Patterns (LBP) have limitations in modeling complex mass textures.

Purpose of the Study:

• To develop an efficient methodology for enhanced mass detection in digital mammograms.
• To introduce a new local feature extraction model, Gray Level and Local Difference (GLLD), for improved texture analysis.
• To evaluate the performance of machine learning classifiers, including Artificial Neural Networks (ANNs), Support Vector Machines (SVM), and k-Nearest Neighbors (kNN), for mass classification.

Main Methods:

• Proposed a novel Gray Level and Local Difference (GLLD) feature extraction technique, incorporating absolute gray level values and local differences.
• Utilized ANNs, SVM, and kNN classifiers to distinguish between malignant masses and non-masses.
• Evaluated the methodology using 1000 regions of interest (ROIs) from the Digital Database for Screening Mammography (DDSM).

Main Results:

• The GLLD method demonstrated superior performance in mass detection compared to traditional texture operators.
• Artificial Neural Networks (ANNs) exhibited the best classification performance among the tested algorithms.
• Achieved a high area under the curve (A(z)) of 0.95 for the mass detection step.

Conclusions:

• The proposed GLLD feature extraction model significantly improves mass detection accuracy in digital mammograms.
• The GLLD approach, particularly when combined with ANNs, offers a robust solution for computer-aided diagnosis.
• This methodology represents a promising advancement for enhancing the early detection of breast cancer.