Classification Of Breast Cancer Stages Utilizing Polarization Images – Mueller Matrix Transformation And A Deep Learning Approach

Abstract

Early identification of breast cancer is critical for successful treatment globally. Mammography and ultrasound are used to diagnose and screen breast cancer. These imaging methods have limited specificity and sensitivity, especially in distinguishing benign from malignant tumors. Thus, breast cancer identification and grading need novel imaging tools. Polarization images with Mueller matrix transformation may correctly grade breast cancer. A laser source 633 nm combines polarized glass system illuminate the tissue and measures polarization changes. With 589 Mueller matrix images (generated from 21,204 raw images captured via CCD camera) collected from four different types of breast tissue that countributed a great dataset for building artificial intelligence model to classify the condition of samples. A deep learning model for breast cancer grading also has promise. Applying ResNet-18 and Random Forest network can evaluate medical images and uncover patterns, values, and characteristics that may greatly support doctors in diagnosing medical condition of patients. Training this algorithm on a large dataset of Mueller matrix images may provide reliable and objective breast cancer detection and classification models. These models might increase diagnosis accuracy and patient outcomes in clinical practice. Polarization images with Mueller matrix transformation and artificial intelligence models for breast cancer grading may enable more accurate and suitable treatment options for breast cancer patients. Healthcare workers may improve therapy and reduce long-term problems by accurately diagnosing and evaluating breast cancer. This revolutionary technology may improve breast cancer diagnosis and help patients choose the best therapy.