Enhancement of Mammographic Images Using Histogram-Based Techniques for Their Classification Using CNN

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Breast cancer affects 1 in 8 women and can be detected via mammograms.
• Mammograms identify abnormalities like masses and microcalcifications, indicative of disease.
• Early detection significantly improves treatment success rates.

Purpose of the Study:

• To evaluate histogram-based image enhancement methods for mammogram classification.
• To improve the accuracy of computer-aided detection (CAD) systems for breast cancer.
• To enhance the identification and categorization of malignant and benign breast lesions.

Main Methods:

• Utilized Contrast-limited Adaptive Histogram Equalization (CLAHE) and Histogram Intensity Windowing (HIW) for image enhancement.
• Applied Deep Convolutional Neural Networks for automatic mammogram classification.
• Trained and tested the model on the mini-MIAS dataset.

Main Results:

• The developed model achieved 62% accuracy in classifying mammograms.
• Image enhancement improved contrast, aiding neural network learning and tissue differentiation.
• The study demonstrated potential for increased correct classification rates.

Conclusions:

• Histogram-based enhancement techniques can improve mammogram classification accuracy.
• The developed algorithm can enhance CAD systems for earlier breast cancer detection.
• Improved early detection through enhanced mammography classification increases cure rates.