Use of T1-weighted MRI in the Quantitative Feature Analysis of Breast Lesions

Breast cancer is a severe threat to women's health worldwide. Diagnosis using MRI with an appropriate automatic segmentation algorithm can help in the early detection of breast cancer. Automated image analysis aims to extract relevant information from breast MRI images and improve accuracy.The main aim of this thesis is to utilize a computer-aided diagnosis scheme that combines quantitative features from T1-weighted MRI using a database of 72 benign and 49 malignant breast lesions in analyzing breast lesions. The algorithm is separated into three major sections: preprocessing, feature extraction, and classification. Firstly, automatic breast region segmentation was performed using a straight-line fitting method, and two categories of features were extracted. Then, the features were analyzed using the unilateral, bilateral, and asymmetry feature analysis methods. Thereafter, the sequential floating forward method (SFFS) was used to select the most effective features.Finally, the extracted features are used in a classification task with a support vector machine classifier. Using the 3-fold crossvalidation, the unilateral, bilateral, and asymmetry feature analysis method based on T1 weighted gave an AUC value of 0.8074, 0.8300, and 0.8353, respectively. When all feature analysis methods were observed, the asymmetry feature analysis method gave the highest AUC value in T1 weighted modalities.