While brainstorming and writing a proposal for a device to detect pediatric tuberculosis has been extremely valuable, we recognize the challenge of developing our devices as undergraduate/graduate students. This acknowledgement led us to try to identify a healthcare problem in Ghana and to come up with a solution that undergraduates could potentially pursue. The process began after we arrived in Ghana, with each student independently identifying a problem and brainstorming a solution. Next, we played an entrepreneurial game, in which each student gave a pitch for an idea, and everyone gave hypothetical money to his or her favorite idea. The ideas with the most hypothetical monetary investments would move on to the next round. After two rounds of pitches, we narrowed our list down to two ideas: Big Data and the Multi-Cot. Splitting up our group between the two ideas, we then prepared a presentation to give to Kumasi Center for Collaborative Research in Tropical Medicine (KCCR) researchers. Today and Friday we present the summaries of our ideas.
Big Data: Deciphering Acoustic Trends in Tuberculosis, Pneumonia and Healthy Coughs
David Pontoriero (gave first-round pitch) ’18, Kathleen Givan ’20, Jason Grosz ’19, Danielle Tsougarakis ’20, Ethan Zhao ’19
Our goal was to think of a project that a team of undergraduates at Penn could complete in one year to produce something of value to KCCR in the scope of Ghanaian healthcare. We turned our attention toward big data science and the difficulties in tuberculosis diagnosis. One of the difficulties identified was the lack of diagnostic tools in more remote arms of the healthcare system. This lack leads to unnecessary and numerous referrals to larger care centers, inconveniencing the patient and placing a burden on the efficiency of the healthcare system.
Specifically, the only standard-of-care diagnostic ubiquitous throughout all clinics was patient-reported symptoms — the most notable of which is prolonged coughing. Moreover, this symptom can often be confused with asthma or pneumonia. However, asthma involves bronchial constriction, and TB and pneumonia have different sputum distribution profiles. We theorized that this difference would correlate with differentiated sound profiles for patient coughs or baseline breathing and, subsequently, measurable biomarkers. The idea proposed was that, if blind data could be collected from KCCR with sound recordings of patients coughing and breathing, along with their demographics and final diagnoses, then analyses could be run to produce an algorithm capable of differentiating between each cough or breath. This algorithm could then be extended to a phone app that could be used to more empirically diagnose patients in any setting and increase overall healthcare efficiency.