A model exposed to the medical community as a web application that allows users to upload ECHO reports and extract a standardized representation of their input.
Next to Chest X-ray, Echocardiogram (also known as cardiac ultrasound) is the single highest-volume cardiac imaging test practiced in millions of hospitals worldwide. There are numerous guidelines documents on what and how to measure heart anatomy and function in an echocardiogram. This has led to different hospitals, medical professionals and patients, speaking a different language when it comes to communicating ECHO results. Harmonization of reports is needed not just to directly improve clinical communications, but also to leverage the ontology labels in pursuit of advanced machine learning on cardiac ultrasound imaging. We propose a hierarchical ontology mapping model that can map free-text in ECHO reports to a standardized ontology. This model is exposed to the medical community as a web application that allows users to upload ECHO reports and extract a standardized representation of their input.
A user can upload an echo report in the CSV format and the web application runs preprocessing pipeline and model predictions in the backend to generate the three-level ontology. Users can explore all the processed reports and individually view each by selecting them.
Hospitals can upload a data dictionary that is a compilation of possible ECHO report free-text sentences used within their facility. This option enables hospitals to test the tool their vernacular style of writing reports. The tool will read the sentences in the data dictionary and generate the ontology for each sentence in the data dictionary. The hospitals can then review the resulting ontology on the web tool.