Investigators have been working on creating a validation a dataset using MIDRC data from diverse locations to validate and compare available pneumonia classification algorithms on chest x-rays for clinical effectiveness. They have also been developing open source visual analytic tools that a) allow radiologists to perform subset analyses to look for mathematical biases in the classification model and b) can evaluate the effect of changing thresholds on model performance.

Current plans include

• Examine whether imaging findings anytime during the course of acute disease are associated with the likelihood of developing long-term morbidities. Examine whether findings from a single institution dataset (MGH) are genaralizable to the MIDRC dataset