Fast Strain Estimation and Frame Selection in Ultrasound Elastography Using Machine Learning

Abstract

Ultrasound Elastography aims to determine the mechanical properties of the tissue by monitoring tissue deformation due to internal or external forces. Tissue deformations are estimated from ultrasound radio frequency (RF) signals and are often referred to as time delay estimation (TDE). Given two RF frames I 1 and I 2 , we can compute a displacement image which shows the change in the position of each sample in I 1 to a new position in I 2 . Two important challenges in TDE include high computational complexity and the difficulty in choosing suitable RF frames. Selecting suitable frames is of high importance because many pairs of RF frames either do not have acceptable deformation for extracting informative strain images or are decorrelated and deformation cannot be reliably estimated. Herein, we introduce a method that learns 12 displacement modes in quasi-static elastography by performing Principal Component Analysis (PCA) on displacement fields of a large training database. In the inference stage, we use dynamic programming (DP) to compute an initial displacement estimate of around 1% of the samples, and then decompose this sparse displacement into a linear combination of the 12 displacement modes. Our method assumes that the displacement of the whole image could also be described by this linear combination of principal components. We then use the GLobal Ultrasound Elastography (GLUE) method to fine-tune the result yielding the exact displacement image. Our method, which we call PCA-GLUE, is more than 10 times faster than DP in calculating the initial displacement map while giving the same result. This is due to converting the problem of estimating millions of variables in DP into a much simpler problem of only 12 unknown weights of the principal components. Our second contribution in this paper is determining the suitability of the frame pair I 1 and I 2 for strain estimation, which we achieve by using the weight vector that we calculated for PCA-GLUE as an input to a multilayer perceptron (MLP) classifier. We validate PCA-GLUE using simulation, phantom, and in vivo data. Our classifier takes only 1.5 ms during the testing phase and has an F1-measure of more than 92% when tested on 1,430 instances collected from both phantom and in vivo datasets.

Comparative Study of Ultrasound Tissue Motion Tracking Techniques for Effective Breast Ultrasound Elastography

“…The greater the deviation from this situation, the lower the quality of the estimated displacement field. The recently developed method known as PCA-GLUE [18], which is a variant of the GLUE method optimized with machine learning, can be used for assessing RF frame pair quality. Given the high speed of the AM2D displacement estimator, it should also be possible to consider it a good candidate for real-time frame pair quality assessment before near imaging at near real-time rates.…”

“…One of the most significant issues plaguing quasi-static USE is the lack of high-quality data due to out-of-plane tissue motion, imperfect application of force, and other operator inconsistencies. PCA-GLUE was recently proposed for data quality assessment [18]. This is a machine-learning-based method developed using data pertaining to tissue mimicking phantoms and only 3 patients; thus, it may not be rigorous for clinical applications.…”

Best Frame Selection to Enhance Training Step Efficiency in Video-Based Human Action Recognition

“…Zhenxing et al [9] used the Laplacian operator to select the appropriate sequences for short periods as the input of a deep learning scheme to identify Hong Kong sign language. Zayed and Rivaz [10] performed elastomeric experiments using ultrasonic images taken from a pressurized mechanical object. The acquired images were first submitted to a multilayer perceptron (MLP) classifier, and any two consecutive frames that contain no relevant information are removed from the list of training frames.…”

An unsupervised learning approach to ultrasound strain elastography with spatio-temporal consistency

“…A common solution to this problem is to compute the strain between all image pairs and associate each resulting strain with a confidence score based on image similarity (Jiang et al 2006, Treece et al 2011 and/or tracking information (Foroughi et al 2013). In Zayed and Rivaz (2020), the frame selection is performed before displacement estimation by using a classifier that gives a binary decision on the suitability of the image pair for strain computation.…”