One of our partners has developed an innovative enhancement of a radiology imaging modality.
They would like to create a library of studies that demonstrate the appearance of normal and
pathological exams with their new technique. They needed thousands of examples of a particular
pathology to help train an algorithm and to generate these new sample cases. They had numerous
inclusion and exclusion criteria based on study technique, pathology restrictions, and demographic
characteristics. Our partner also needed geographically and demographically matched control
studies, further complicating the request.
HOPPR delivered! We provided studies meeting all of their criteria in a rapid and cost effective
manner. They did not have to seek out and negotiate with dozens of facilities, nor did they have
to interface with the many IT systems and data engineering steps needed to get their data to be
research ready. Our partner is now able to jump start their model training with a large and highly
specific data set.
This partner is interested in generating new Indications For Use (IFUs) to further monetize one of
their implantables. One approach is to identify indications in prospective patients that suggest
better outcomes, fewer complications, or more suitability for their particular device. Much of that
information is embedded in medical imaging data and difficult to access, unlock, and search for. An
additional challenge is to be able to assemble the volume of specific studies that would support these
types of analysis.
HOPPR was able to find the relevant studies and to help extract the specific information that they
needed to power their analysis and model building. This saved countless hours sifting through thousands
of studies to first find criteria matches, and secondly having to manually extract the measurements and
information from each study. Our partner is able to drive this research forward powered by the data and
feature extraction supplied by HOPPR.
Our partner is creating a highly specialized workflow which can help predict the etiology of incidental
plain film findings, using cross sectional imaging as the ground truth. Their data requirements cut
across time, findings from other data types, and features on the images themselves. The challenge was
to sift through thousands of patients to get the match needed for the training set.
HOPPR's advanced Cohort Builder, powered by our proprietary NLP and data engineering was able to reduce
this to a simple set of queries. Cohorts of studies are quickly generated and our partner is enabled to
focus on their core competency and train and build their models.