Journal: Neurocritical Care
Article Title: Unsupervised Clustering in Neurocritical Care: A Systematic Review
doi: 10.1007/s12028-024-02140-w
Figure Lengend Snippet: An overview of the information flow between studies ( N = 18), the corresponding unsupervised clustering methods, and the clustering data. In the alluvial plot’s final column, functional outcome results are shown assessed by the Glasgow Outcome Scale-Extended, Disability Rating Score, or mortality. The publications are listed alphabetically (first author), the unsupervised clustering methods according to the frequency of occurrence, the clustering data by data type. Three methods shown are variations of probabilistic graph model approaches: Bayesian models, GBTM and unsupervised hidden Markov model. The Bayesian models include both a standard model and an extension incorporating Markov chain extension. The monitoring data (and derived measures) included: microdialysis, audio data, intracranial pressure, cerebral perfusion pressure, pressure reactivity index, the correlation coefficient ( R ) between mean pulse amplitude and mean intracranial pressure, arterial blood pressure, heart rate, cerebral/body temperature, brain tissue oxygen partial pressure, cerebral blood flow velocity, peripheral oxygen saturation, electroencephalography. Functional outcomes are listed sequentially, the studies that did not assess functional outcome (not assessed), the presence (relationship) or absence (no relationship) of a significant difference between the clustering results. The size of the individual bars corresponds to the number of in/output flows and may not correlate with the number of patients/methods included in the individual studies; GBTM Group-based trajectory modeling, MH Medical history, MOCAIP Morphological clustering and analysis of intracranial pressure, NCC Neurocritical care, SOM Self-organizing maps. A brief description of each method is available in the Supplementary Tables and and Supplementary Fig.
Article Snippet: The size of the individual bars corresponds to the number of in/output flows and may not correlate with the number of patients/methods included in the individual studies; GBTM Group-based trajectory modeling, MH Medical history, MOCAIP Morphological clustering and analysis of intracranial pressure, NCC Neurocritical care, SOM Self-organizing maps.
Techniques: Functional Assay, Derivative Assay
SAS institute is a verified supplier 
