DIAL focuses on translating diabetes data into information, knowledge, and action. This involves research in informatics, data mining, visualization, modeling, and good old-fashioned stats. Some of our projects:


Toward a holistic view of the diabetes data ecosystem
The Diabetes Management Integrated Technology Research Initiative (DMITRI) pilot study was an intensive real-world physiologic monitoring effort. Seventeen people with type 1 diabetes used insulin pumps, CGMs, and other sensors and monitors for heart rate, physical activity, sleep, and more, for three or four days. They also photographed their meals and snacks for annotation by our dietician collaborators. The resulting datasets paint detailed personalized pictures of diabetes physiology under real-world conditions including a variety of different kinds of exercise. In partnership with the Predictive Analytics Center of Excellence (PACE) at UCSD, we are evaluating these datasets for training blood glucose forecasting algorithms.


Squeezing more information out of CGM data
A “motif” is a recurring pattern. For example, an electrocardiogram (ECG or EKG) displays data from the beating of the heart, and these data contain motifs like the QRS complex, which are informative to physicians examining the data. We think that motifs with meaningful information content also exist in the continuous data collected by a CGM. Identifying these motifs will enable us to develop an “alphabet” of the basic units of CGM data, as each motif represents multiple datapoints with a single “letter”. We can then relate the prevalence and time-positioning of these motifs in a PWD’s CGM data to other diabetes health factors and management strategies.


Exploring diabetes in space
Tools for integrating data from GPS, fitness, and diabetes devices, with interactive visualizations. A prototype can be found at www.glucomap.org


Using patterns identified by GlucoMotif, we are attempting to define distinct “glucotypes” for PWDs. Like a genotype characterizes an individual’s DNA sequence, we think a glucotype can characterize an individual’s glycemia in a manner much more personal than the traditional HbA1c test. More personalized characterization can lead to more personalized treatment – in a data-driven way.