Patients with acute respiratory illnesses commonly present in emergency departments and ambulatory assessment units. Diagnosing prevalent conditions such as respiratory infections or pulmonary embolism, which often cause these symptoms, can be challenging for doctors.
Frequently, CT scans are necessary for this patient group to reach a diagnosis. This is problematic as CT scans not only pose a cancer risk but are also costly for the NHS.
The goal of this project is to employ machine learning methods to diagnose these conditions more accurately, potentially eliminating the need for scans. We plan to utilize routinely collected healthcare data, including observations, laboratory test results, and ICD-10 coded co-morbidities that are commonly collected before the scan, to predict the scan’s outcomes.
If successful, our model could potentially prevent some future patients from requiring a scan. This would not only reduce the individual patient’s cancer risk but also allow them to start treatment sooner. Moreover, this could help alleviate the strain on emergency healthcare services by reducing wait times for scans.