PROJECT TITLE :

Compressed Sensing Reconstruction of 3D Ultrasound Data Using Dictionary Learning and Line-Wise Subsampling

ABSTRACT:

In this paper we gift a compressed sensing (CS) technique adapted to 3D ultrasound imaging (US). In distinction to previous work, we tend to propose a new approach primarily based on the employment of learned overcomplete dictionaries that enable for abundant sparser representations of the signals since they are optimized for a particular class of pictures like US pictures. In this study, the dictionary was learned using the K-SVD algorithm and CS reconstruction was performed on the non-log envelope knowledge by removing 20percent to eightypercent of the first information. Using numerically simulated pictures, we have a tendency to evaluate the influence of the coaching parameters and of the sampling strategy. The latter is finished by comparing the two most typical sampling patterns, i.e., purpose-wise and line-wise random patterns. The results show in specific that line-wise sampling yields an accuracy love the standard purpose-wise sampling. This indicates that CS acquisition of 3D data is possible in a comparatively simple setting, and therefore offers the attitude of accelerating the frame rate by skipping the acquisition of RF lines. Next, we evaluated this approach on US volumes of many ex vivo and in vivo organs. We have a tendency to initial show that the learned dictionary approach yields higher performances than standard fastened transforms like Fourier or discrete cosine. Finally, we tend to investigate the generality of the learned dictionary approach and show that it's potential to build a general dictionary permitting to reliably reconstruct totally different volumes of different ex vivo or in vivo organs.