From: Visualization of medical concepts represented using word embeddings: a scoping review
Study | Visualization objective | Visualization method |
---|---|---|
Wang et al. [24] | Exploration of «visual» clusters | t-SNE with focus on manually-selected areas |
Shah et al. [31] | Synthesis of patients' medical history + soft validation of clusters, hence embeddings | ad hoc chronology of symptoms and diseases, organized visually based on embeddings’ cluster assignments |
Beaulieu et al. [32] | Illustration of the embeddings’ consistency with the ICD-9 hierarchy | Plot of 2-d embeddings and Hierarchy Tree [44] |
Dynomant et al. [33] | Exploration of « visual» clusters presented as an evaluation of the embeddings | t-SNE with focus on manually-selected areas |
De Freitas et al. [22] | Illustration of the embeddings and of some identified disease phenotypes | UMAP with highlight of neighborhoods around selected diagnostic codes |
Chen et al. [34] | Illustration of the embeddings and some relations between terms used as part of the evaluation tasks | PCA with highlight of selected words, some with their neighborhoods (used for word-retrieval evaluation) |
El-Assady et al. [37] | Topic models’ exploration & user feedback gathering (part of the modeling process) | ad hoc t-SNE-based visualization of topics and related words grouped into concepts + interactive system to gather feedback and trigger model evolutions |