Breast Cancer Masses Detection from Mammography Images based on Advanced Reinforcement Learning

Breast cancer is one of the most dreaded diseases that affects womenworldwide and has led to many deaths. Early detection of breast massesprolongs life expectancy in women and hence the development of anautomated system for breast masses supports radiologists for accuratediagnosis. In fact, providing an optimal approach with the highest speedand more accuracy is an approach provided by computer-aided designtechniques to determine the exact area of breast tumors to use a decisionsupport management system as an assistant to physicians. This studyproposes an optimal approach to noise reduction in mammographicimages and to identify salt and pepper, Gaussian, Poisson and impactnoises to determine the exact mass detection operation after these noisereduction. It therefore offers a method for noise reduction operationscalled Quantum Wavelet Transform Filtering and a method for precisionmass segmentation called the image morphology operators inmammographic images based on classification with ReinforcementLearning (RL). The hybrid approach called QWT-RL is evaluated interms of criteria compared to previous methods such as peak Signal-to-Noise Ratio (PSNR) and Mean-Squared Error (MSE) in noise reductionand accuracy of detection for mass area recognition. The proposedmethod presents more performance of noise reduction and segmentationin comparison to state-of-arts methods. Obtained results presented thatproposed approach have better performance in comparison to othersbased on some evaluation criteria such as accuracy with ۹۸.۵۷%,sensitivity with ۹۰%, specificity with ۸۵% and also ROC and AUC withthe rate of ۸۶.۷۷.

کلیدواژه ها:

Breast Cancer ، Noise Reduction ، Image Segmentation ، Mammography ، Quantum Wavelet Transform (QWT) ، Reinforcement Learning(RL)