Automated Detection and Classification of Microcalcification Clusters with Enhanced Preprocessing and Fractal...

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computational Pathology

Background:

• Automated detection of microcalcification clusters in mammograms is crucial for early breast cancer diagnosis.
• Preprocessing digital mammograms is essential for accurate analysis, involving tissue extraction and artifact removal.

Purpose of the Study:

• To develop an automated system for detecting and classifying microcalcification clusters in mammograms.
• To enhance preprocessing techniques for improved breast tissue extraction and artifact reduction.
• To classify suspicious microcalcification clusters as benign or malignant using fractal and texture analysis.

Main Methods:

• Enhanced preprocessing operations including breast identification, realignment, and pectoral muscle separation.
• Fractal analysis of suspicious regions to extract texture features.
• Principal Component Analysis (PCA) for optimal feature reduction, identifying 10 key features.
• Classification using a Scaled Conjugate Gradient Backpropagation neural network with 15 hidden layer neurons.

Main Results:

• Optimal feature reduction achieved with 10 fractal features using PCA without compromising accuracy.
• Maximum classification accuracy of 96.3% was obtained with the proposed method.
• The neural network architecture was finalized with 15 hidden layer neurons for maximized accuracy.

Conclusions:

• The developed automated system effectively detects and classifies microcalcification clusters with high accuracy.
• The combination of enhanced preprocessing, fractal analysis, PCA, and neural networks provides a robust approach for mammogram analysis.
• This method holds potential for improving the efficiency and accuracy of breast cancer diagnosis through automated analysis of mammograms.