Global Ultrasound Elastography Using Convolutional Neural Network

TL;DR

This paper introduces GLUENet, a deep learning-based ultrasound elastography method that improves displacement estimation accuracy and robustness against decorrelation noise, leading to higher quality strain images.

Contribution

The study presents a novel CNN-based elastography technique that effectively addresses decorrelation noise and reduces parameter sensitivity compared to traditional methods.

Findings

CNR and SNR of strain images are comparable to GLUE.

The method is less sensitive to parameter tuning.

Achieves accurate displacement estimation in ultrasound elastography.

Abstract

Displacement estimation is very important in ultrasound elastography and failing to estimate displacement correctly results in failure in generating strain images. As conventional ultrasound elastography techniques suffer from decorrelation noise, they are prone to fail in estimating displacement between echo signals obtained during tissue distortions. This study proposes a novel elastography technique which addresses the decorrelation in estimating displacement field. We call our method GLUENet (GLobal Ultrasound Elastography Network) which uses deep Convolutional Neural Network (CNN) to get a coarse time-delay estimation between two ultrasound images. This displacement is later used for formulating a nonlinear cost function which incorporates similarity of RF data intensity and prior information of estimated displacement. By optimizing this cost function, we calculate the finer displacement by exploiting all the information of all the samples of RF data simultaneously. The Contrast to Noise Ratio (CNR) and Signal to Noise Ratio (SNR) of the strain images from our technique is very much close to that of strain images from GLUE. While most elastography algorithms are sensitive to parameter tuning, our robust algorithm is substantially less sensitive to parameter tuning.