بهبود کیفیت تصاویر ماموگرافی با استفاده از تکنیک‌های پردازش تصویر

با توجه به گسترش ابتلا به سرطان سینه، تشخیص زود هنگام آن از روی تصاویر ماموگرام با استفاده از روشهای کامپیوتری، بعنوان روشی موثر در جهت کاهش میزان مرگ و میر بیماران مورد توجه قرار گرفته است. از این رو در این پژوهش روشی مبتنی بر تکنیکهای پردازش تصویر جهت بهبود کیفیت تصاویر ماموگرافی ارائه شده است. از این رو در این پژوهش در جهت ایجاد یک سیستم پزشکیار، سعی در بهبود کیفیت تصاویر ماموگرافی با تکنیک های پردازش تصویر داریم. این پژوهش دارای دو مرحله پیش پردازش شامل یکسان سازی ابعاد و تعدیل هیستوگرام تصاویر و مرحله‌ی استخراج ویژگی با استفاده از تبدیل‌های کانتورلت و کرولت از تصاویر ماموگرافی است که سه دسته ویژگی ریخت‌شناسی و بافت شناسی، آماری و فرکانسی را جهت بهبود تشخیص ارائه می‌نماید و سبب افزایش صحت تشخیص می‌شود. روش بهبود پیشنهادی روی مجموعه داده mias اجرا شد و زیر مجموعه ای از‌ ویژگی‌های استخراج شده برای ورودی طبقه‌بند انتخاب گردید. بامقایسه عملکرد روش پیشنهادی روی طبقه‌بندهای مختلف، این روش میزان صحت 86.3 را نشان می‌دهد که نسبت به بقیه روش ها نتیجه بهتری می‌باشد.

improve the quality of mammogram images by image processing techniques

introduction: unfortunately, breast cancer has become the most common disease among women, leading to death. breast cancer can be diagnosed using the classification of breast masses. among the various methods used to diagnose breast cancer, mammography is considered the most promising method and has been used many times by doctors. mammogram images are usually noisy and have low contrast. therefore, the creation of an automatic or semi-automatic diagnostic aid system can be of great help to the doctor in the disease diagnosis process. in this field, machine learning and image processing algorithms are used. therefore, in this research, we have tried to improve the quality of mammogram images and increase segmentation accuracy to increase the accuracy of diagnosis and segmentation of breast cancer tumor masses. to achieve this goal, we used image processing and data mining techniques.method: this research is looking for a method that can improve the accuracy of the final diagnosis. therefore, after the pre-processing stage, which includes rescaling and adjusting the texture of the image, highlighter transforms in the frequency domain such as curvelet and contourlet are used to highlight and increase the differentiation of areas with masses in the image for decision-making. traditional enhancement methods are fixed neighborhood methods used to enhance images based on overall features. however, local features based on image zoning are used for image segmentation, and these methods are also used to increase the contrast of mammogram images concerning their surroundings. the improvement methods used in this research use features based on the wavelet domain.results: the input image to the system is subjected to the feature extraction process and three main categories of frequency, morphology, and histology features are extracted from it. this process is done through the cycle, size equalization, histogram adjustment, contourlet transform, and curvelet transform. due to the sensitivity of the systems, it has been tried to extract the features with various levels and known matrices such as the matrix of events and gray groups.the classification results evaluated the said method. the best results on the data set were the proposed method, which reached an accuracy rate of 86.3 and showed a good improvement on the displayed data set