A SOLUTION MEMO
Lower respiratory illnesses (COPD, asthma, pneumonia, and pulmonary tuberculosis) are the second leading cause of death in rural Uganda. While up to 58% of patients in rural Ugandan clinics present with symptoms of respiratory illnesses, the shortage of trained medical staff and lack of diagnostic equipment causes the rate of under diagnosis to remain over 60%.
FULL PROBLEM BREAKDOWN + STATUS QUO OF CURRENT DIAGNOSTIC PROCESS
Our solution leverages machine learning to process cough recording through a smartphone device and determine if a patient has one of the four prevalent respiratory illnesses (COPD, asthma, pneumonia, & pulmonary tuberculosis). It takes advantage of supporting demographic and symptom metadata which is collected through a yes/no questionnaire to further the accuracy of the screening.
Ideally, this app will be used in the scenario where the patient is clearly presenting with respiratory illness symptoms, but it is a matter of discovering which disease they potentially may have. If the algorithm cannot detect any of the four common diseases, it will be able to produce an "undetermined" result which would indicate the potential possession of another disease.
Unless FDA approved, our app will ultimately be a screening tool, opposed to a concrete diagnostic procedure. While screening does not act as a diagnosis, it's an incredibly useful tool for areas where diagnostic efforts are close to absent, specifically in rural Uganda. It can allow those who present with symptoms to be given an indication of which disease they potentially may have, and allow for a more solid plan of action to be formulated into specific tests/care that may be required.
The incredibly easy screening procedure means that patient check ups do not have to only take place when the doctor is present at the clinic. We will be creating a smart-phone attachable 3D printed device which can standardize the cough recording, eliminating the need for medically trained support. Only a trained health care volunteer will be required for supervision and to guide the patient through the yes/no questionnaire. The rest of the procedure is fully automated by AI within the smartphone.
The overall increased accessibility of such product, could help to filter patients and lessen the burden that doctors face.
The screening will be made with the input of a coughing recordings, and would not require the use of spirometers, x-ray machines, and/or microscopes. Additionally, recording the sound of a patient coughing is a much simpler procedure compared to learning to work with complicated lab equipment and interpreting results, which requires professional training. While smartphones will have to be acquired, these devices are becoming increasingly accessible in low-resource areas.
Results are generated in seconds, allowing for hundreds of people to be rapidly screened per day. With this speed and simplicity, it enables the possibilities of on-site screening anywhere, even in low-resource areas, allowing for more people to have point-of-care respiratory check-ups and receive the direct care that they need.
In order to ensure high quality data input and optimize results, as well as minimize supervision, an initial filtration algorithm will determine if the cough recording is a valid input before the data is processed through the model. The algorithm will consider facts such as the volume of the coughing noises, background noise in the recording, and whether the recording even contained a cough. If the recording isn't optimal, it will automatically ask the user to record again.