Noise Removal for Mammogram Images Using Gaussian based Unsharp Masking

One major health issue that still affects women is breast cancer, and timely detection of breast masses dramatically improves life expectancy. (CAD)Computer-Aided Diagnosis help radiologists in achieving precise and accurate diagnoses, with mammograms being a key tool for early detection. However, noise in mammogram images can mislead diagnoses, making image preprocessing essential to remove noise while preserving critical information. Traditional noise reduction algorithms often deteriorate edge details, leading to false predictions and inappropriate treatments. In order to reduce noise effectively without sacrificing important information, this paper presents a cascade technique that combines unsharp masking with a Gaussian filter. The effectiveness of the suggested approach was evaluated against commonly used denoising filters such as Gaussian, Mean, Median, and Conservative Smoothing filters. Peak Signal-to-Noise Ratio (PSNR) and Mean Squared Error (MSE) on mammograms from the MIAS (Mammographic Image Analysis Society), InBreast, and DDSM (Digital Database for Screening Mammography) datasets were used to assess the effectiveness of the suggested approach. The suggested method showed considerable gains in noise reduction. It achieved PSNR values of 45.48 dB, 45.10 dB, and 50.26 dB, and corresponding MSE values of 1.64 dB, 0.37 dB, and 0.66 dB for the InBreast, MIAS, and DDSM datasets, respectively. Additionally, the impact of noise reduction on classification accuracy was examined using gradient boosting, SVM, and random forest classifiers. The analysis revealed a notable improvement in accuracy, with an average increase of 6.06% when using noise-free mammogram images. This enhanced accuracy indicates that removing noise from mammograms contributes to clearer and more reliable feature representation, enabling machine learning models to make more accurate predictions.