Figure 2.

(A) Characteristics of big data. (B) Common AI definitions. (C) Common model architectures used in AI depend on the purpose of modelling. With supervised learning, predictors are mapped to a known outcome. When the outcomes of interest are clinical, machine learning methods such as random forest and support vector machine are used. When the outcome of interest is imaging-based, then deep learning methods such as convolutional neural networks are used. (D) With unsupervised learning, the predictors are visualised on a plot to find natural clustering of the data. A typical use in valve disease studies has been in phenotyping to identify higher risk phenotypes. Methods used with unsupervised learning include topological data analysis, model-based clustering, agglomerative hierarchical clustering and clustering around medoids. AI, artificial intelligence.