Wearable device data
This web-based visualizer uses the Fitbit API to query data directly from the Fitbit website. We use the Fitbit Charge HR to collect data from patients recovering from critical illness. (Noam Jacobson, Torrey Firth, Anthony Persico)
Pulsus physiology tracker
In this project, pulsus paradoxus is derived from arterial line pressure waveforms and tracked across a cohort of ICU patients. Each row (x-axis) represents a single patient's pulsus values over a 24-hour period. Thresholding can be added to identify critical physiologic events. (Phil Laird)
Vital signs outliers
We analyzed more than 20 million vital signs measurements from the MIMIC database and identified outliers in each of the vital signs represented. The greyscale pie charts show the percentage of measurements deemed outliers for each vital sign type, while the coloured histograms show the prevalence of outlier recordings by time of day.
Sparse vs. dense vital sign monitoring
These data (simulated) show the potential for higher-frequency sampling to identify important differences between patients, a necessary condition for precision approaches to critical care.
Graph-based visualization of the relationship between scientific publications based on semantic analysis of their abstracts. In this example, the relationship between terms in an abstract are shown. Concepts are represented as nodes, with relationships between concepts represented as edges (Amal Khalil and Shadi Khalifa)
Time-based visualization of trauma resuscitation
In this project, archival data was used to explore different visualizations of clinical data derived from patients admitted to the ICU following trauma.
- Django Web Server (Python 3.4.3)
- MySQL Database with Django Tables + ICU Data Tables
- Bokeh Graphing Interface
Above, the first 2 days of patient resuscitation showing pressors, fluids, labs and vitals. Colour intensity reflects numeric value for each field. Below, individual patient Entire Stay Graph with procedures, labs, vitals, nutrition, ventilation using APACHE II scores to shade concerning values more intensively (Killian Newman)
Schematic representation of the diversity of ICU data
A variety of data types are represented in the ICU including continuous, ordinal, categorical, structured, and unstructured, all collected at different time scales.