Developed by: Tony Tadic and Tirth Patel

Acquiring data, image processing, data federation, data cleaning, data exploration, model creation, model testing and model validation can all be time-consuming and challenging tasks, requiring clinicians to go into each source system, search for and acquire the data of interest, manually process images (often taking hours per image), federate the data with other sources (often spreadsheets), clean it, examine it, then often need help to create, test and validate models based on the data, with models being siloed on individual computers. The outputs of these models are what can be useful in clinic. This has to be repeated across an entire cohort of patients. With MIRA, this data science workflow can be automated, saving on many time-consuming and repetitive steps, especially related to data wrangling, image processing and creating models for prediction and machine learning. It also allows for centralized storage of validated models.

MIRA currently integrates with a number of source systems at UHN including RayStation, Pinnacle, Web Publishing, MOSAIQ, JDMI PACS and HPC4H.