Texture Feature Analysis of Breast Lesions in Automated 3D Breast Ultrasound
Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
This thesis investigated a variety of texture features performances on classifying and detecting breast lesions in automated 3D breast ultrasound (ABUS) images with computer-aided diagnosis and detection system. Regions detected by the computer-aided detection system could be categorized into benign and malignant classes, which are supposed to have different texture features.
After normalization and segmentation on the original 3D ultrasound breast images automatically, we implemented four texture feature extraction algorithms on the detected targets. The proposed four algorithms are based on 3-dimensional gray level co-occurrence matrix (3-D GLCM), local binary pattern (LBP), Haar-Like and regional zernike moment (RZM) separately. Three major experiments were carried out on a set of ABUS images. In experiment one, we focused on distinguishing malignant lesions (165 samples) from benign lesions (258 samples). In experiment two, we added a number of normal cases (150 samples) to the dataset, by grouping them with benign lesions against malignant lesions and by isolating them from benign and malignant lesions. In experiment three, we tested texture features ability on reducing false positives in the existing computer-aided detection system. In this step, only normal cases (5263 samples) and malignant lesions (165 samples) were examined.
To estimate the discrimination power of different texture features, Support VectorMachine (SVM) and AdaBoost classifiers were adopted in corporation withleave-one-patient-out and 10-fold cross validation schemes respectively. The areaunder the receiver operator characteristic (ROC) curve (AUC, also known as Az)values were analyzed corresponding to each texture feature extraction method. TheAz values computed in experiment one are compared as follows: Haar-Like feature'sperformance outweighs others' with the Az value of 0.86, followed by LBP (0.84),RZM(0.81) and 3-D GLCM (0.75). With respect to the results from experiment two,the Az value of grouping normal cases with benign lesions against malignant lesions isbetter than separating them from benign and malignant lesions, in general. Regardingthe outcome from experiment three, the Az value was increased from 0.79 to 0.82after adding LBP and Haralick features to the existing computer-aided detectionsystem.
Based on the overall results, we concluded that texture features are useful on classifying benign and malignant lesions in ABUS images and they can improve the performance of the existing computer-aided detection system on detecting breast cancers.
Place, publisher, year, edition, pages
IT, 13 052
Engineering and Technology
IdentifiersURN: urn:nbn:se:uu:diva-211923OAI: oai:DiVA.org:uu-211923DiVA: diva2:669196
Master Programme in Computer Science
Tan, TaoPlatel, BramKarssemeijer, Nico
Bengtsson, EwertChristoff, Ivan