Artificial intelligence - based ultrasound elastography for disease evaluation - a narrative review

Abstract

Ultrasound elastography (USE) provides complementary information of tissue stiffness and elasticity to conventional ultrasound imaging. It is noninvasive and free of radiation, and has become a valuable tool to improve diagnostic performance with conventional ultrasound imaging. However, the diagnostic accuracy will be reduced due to high operator-dependence and intra- and inter-observer variability in visual observations of radiologists. Artificial intelligence (AI) has great potential to perform automatic medical image analysis tasks to provide a more objective, accurate and intelligent diagnosis. More recently, the enhanced diagnostic performance of AI applied to USE have been demonstrated for various disease evaluations. This review provides an overview of the basic concepts of USE and AI techniques for clinical radiologists and then introduces the applications of AI in USE imaging that focus on the following anatomical sites: liver, breast, thyroid and other organs for lesion detection and segmentation, machine learning (ML) - assisted classification and prognosis prediction. In addition, the existing challenges and future trends of AI in USE are also discussed.

Keywords: artificial intelligence; deep learning; elastography; machine learning; radiomics; ultrasound.

Figure 1

The category of ultrasound elastography (USE) technique according to the excitation methods, including external compression or internal physiologic motion (green), acoustic radiation force impulse (blue), and mechanical vibrating (yellow).

Figure 2

The relationship between artificial intelligence (AI), machine learning (ML), deep learning (DL) and convolutional neural networks (CNNs).

Figure 3

The classification and computer vision task of machine learning (ML). ANNs, artificial neural networks; DT, deciosin tree; FCM clustering, fuzzy C-means clustering; KNN, K-nearest neighbor; LR, logiastic regression; MRF, Markov random fields; NBC, Naïve Bayes classifier; RF, random forest; SVM, support vector machine.

Figure 4

The architecture of convolutional neural network (CNN), which is formed of one input layer, multiple hidden layers and one output layer. Convolutional and max pooling layers can be stacked alternately until the network is deep enough to acquire optimal features of the images that are salient for classification task.

Figure 5

Conventional machine learning (ML)-based ultrasound elastography (USE) models vs deep learning (DL)-based USE models. Conventional ML-based USE models depend on carefully handcrafted features, while DL allows learning an end-to-end mapping from the input to the output.