Abstract

The accuracy of time-delay estimation (TDE) in ultrasound elastography is usually measured by calculating the value of the normalized cross correlation (NCC) at the estimated displacement. NCC value is usually high if the TDE is correct. However, it could be very high at a displacement estimate with large error, a well-known problem in TDE referred to as peak-hopping. Furthermore, NCC value could suffer from jitter error, which is due to electric noise and signal decorrelation. In this thesis, we propose a novel method to assess the accuracy of TDE by investigating the NCC profile around the estimated time-delay in a supervised approach. First, we extract seven features from the NCC profile, and utilize a linear support vector machine (SVM) to classify the peak-hopping and jitter error. The results on simulation, phantom and in-vivo data show the significant improvement in the classification accuracy realizing from the proposed algorithm compared to the obtained form the state of the art techniques. Second, we build on our model by utilizing the continuity features in the axial and lateral directions as a prior knowledge. We show that these features also improve the sensitivity and specificity of the classifier. After extracting the continuity features in addition to the seven features, we show the performance improvement of the proposed model on the available data sets. Furthermore, we show that our proposed model could be trained by other elastography methods in future, since we use a new elastography algorithm to train the model. Third, we compare the performance of the method developed using well-known classifiers in the literature and then study the importance of the proposed features using the mean decrease impurity method of the random forest classifier.