By Anna Püsök
On 19th of February, the Sewanee Molecular Diagnostics Lab announced a new study that could detect COVID-19 status based on the characteristics of a cough. Dr. Alyssa Summers and Dr. Clint Smith, professors of biology and directors of the lab, started to use cutting-edge machine learning techniques to investigate whether COVID-19 positivity can be predicted based on the sound of participants’ coughs.
Sewanee works together with Hyfe, a phone app that detects and tracks users’ coughs, co-founded by Sewanee alumna Joe Brew (C’08). Dr. Eric Keen, professor of biology, and Brew reached out to Dr. Summers and Dr. Smith with the opportunity.
The Hyfe app records voice, but specifically focuses on coughs, and analyzes them. Now, Sewanee’s collecting data for Hyfe. Smith explained: “When those recordings, the audio files, are sent to them [Brew’s company], they run them against their algorithm and pretty much ask the question based on their dataset, do they think this individual might be a positive for Sars-Coronavirus-2 or negative.”
When asked about the lab’s decision to participate in the study, Smith said, “It is a unique opportunity for many reasons: one, to get students involved in many different processes, and also to begin to ask the question of what is mean to be a clinical diagnostic lab at Sewanee, as we’re not like a huge for-profit clinical lab, and how do we provide services and support that help the local community.”
“Also, because Dr. Summers and I are both research scientists too, how do we make a molecular diagnostics lab work in the setting of a liberal arts college, how do we involve students in meaningful ways, and how do we keep asking scientific questions that most interest us,” Smith continued.
Sewanee’s big volume of testing is very beneficial for the study. “Sewanee’s participation is important for the study as their [Hyfe’s] algorithm needs to be trained. One question we’ve gotten is: if we’re doing our jobs public healthwise well, we shouldn’t have that many positive cases on Sewanee’s campus, so how useful is this? Well, one also needs a lot of negative data to train the model as well. Machine learning means to know what a negative individual sounds like and also what a positive individual sounds like,” Smith said.
This study is useful for COVID research and testing as “this particular cough A.I. tool could be used to change the frequency at which diagnostic testing needs to be performed.”
Smith said, “One of the cool features of the study in a sense is that because the sound is being recorded, the human ear has a finite capacity to distinguish between frequency and intensity and such, and it might be that there are differences that are perceptible to a computer that aren’t perceptible for humans.”
Emily True (C’23), research assistant for the study explained how Sewanee students can participate in the study: “New participants of the study first need to sign the consent form, which contains five points of check-in: first two emphasizes the consent that Sewanee pairs the COVID test result with the sound of the cough, the second two points are about the app, called Hyfe, and the last one emphasizes the opportunity of opting out of the study any time.”
True further explained, “If you have already signed up, you tell the person there your last name. They will instruct you to the microphone and the computer screen outside.You’ll go stand next to the microphone, you’ll see 3,2,1 on the computer,you’ll take your mask down, cough, and then you’re free to go.”
Participating in the study shouldn’t take longer than a minute every week.
True continued, “We would love it if you would cough every week. It’s just another part of the COVID test. It’s so beneficial to have as many different coughs as we can work with.”
According to True, the first week of the study was very successful. She said, “The first week, we had about 600 hundred people. The following week, we had about a 60% less participation, because people didn’t realize it was an every week thing, but we were able to get about a 100 new participants, which was really big and awesome.”
So far, about half of the students on campus participated in the study. True said, “Our goal is to boost the number of new participants, but really keep up the number of recurring participants.”
Now, the cough study has 16 student volunteers.
Sophia Higgs (C’23) volunteers at the station every Monday morning. She said, “At 10 am, Nellie [Bowers (C’23)] and I go and set up the cough station, so that means, putting the computer and microphone outside, turn on the app, and setting up tables so people know where to go.”
She continued, “ I check people in, so I put people’s names to the app, and make sure that their cough is associated with their test results.”
Volunteers also make sure to keep everything clean. Higgs explained, “We wipe down the set up and doors that people use, and the pens people use to sign their forms, and we’re constantly using hand sanitizer.”
As of now, Sewanee plans to participate in the study at least until the end of the semester. Future goals will be determined based on how the dataset looks like at the end of the semester, and what questions arise during the study.