Artificial Intelligence-Enhanced Echocardiography for Systolic Function Assessment

Figures

• 

  Figure 1. Logical diagram of AI, ML, and DL and the main characteristics of ML and DL. (AI: artificial intelligence; ML: machine learning; DL: deep learning).

• 

  Figure 2. The workflow of studies on AI in echocardiography, including four main steps: (1) clinical problem-oriented data collection; (2) data preprocessing operations based on task characteristics (classification tasks require explicit sample labels; segmentation tasks require the marking of regions of interest) and data splitting (training, validation, and testing datasets are mutually independent); (3) based on the type of tasks (regression, classification, or clustering), appropriate AI algorithms are selected for model development on the training datasets, and the performance of the model is validated on the validation datasets; (4) the reliability and generalization of the model are tested on the internal and external independent testing datasets.

• 

  Figure 3. Overview diagram of AI’s application in echocardiography. Current applications focus on view classification and image quality control (classification), cardiac phase detection and cardiac function assessment (regression), and disease diagnosis and prognosis analysis (clustering). Mainstream AI algorithms consist of the convolutional neural network (CNN), recurrent neural network (RNN), transformer, and traditional machine learning algorithms (RF and SVM).