A Hybridized ELM for Automatic Micro Calcification Detection in Mammogram Images Based on Multi-Scale Features

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Biomedical Engineering

Background:

• Digital mammography is crucial for breast cancer screening.
• Detecting microcalcifications, an early sign of breast cancer, is challenging for radiologists.
• Computer-Aided Detection (CAD) systems assist in identifying breast lesions.

Purpose of the Study:

• To develop and evaluate an Extreme Learning Machine (ELM) algorithm for accurate microcalcification detection in digital mammograms.
• To enhance the classification of malignant versus benign microcalcifications.
• To improve upon existing CAD frameworks for mammographic analysis.

Main Methods:

• Pre-processing techniques were applied to remove image interference.
• Multi-scale features were extracted using a feature generation model.
• Nature-inspired optimization algorithms were employed for optimal feature selection.
• A hybridized ELM classifier utilized selected features to differentiate microcalcification types.

Main Results:

• The proposed ELM-based system achieved a high accuracy of 99.04%.
• The system demonstrated superior performance in training time, sensitivity, specificity, and accuracy compared to Support Vector Machine (SVM) and Naïve Bayes (NB) classifiers.
• Optimal feature selection and an efficient classifier significantly improved detection performance.

Conclusions:

• The hybridized ELM classifier with optimal feature selection offers a highly effective approach for microcalcification detection in mammograms.
• This method shows significant potential for improving the accuracy and efficiency of breast cancer screening.
• The proposed system outperforms existing classification methods, offering a promising advancement in CAD for mammography.