Light AI trial a culmination of many passions for Peter Whitehead
The Western Hockey League is the latest organization to benefit from Whitehead’s curiosity and drive.
Peter Whitehead has always been an inventor. His first invention, at 14 years old, was a part for a day-timer prepared and sent via his local library to Casio, the reputable dealer of stopwatches and keyboards. “They gave me 50 bucks,” he recalled. The revenue was immediately reinvested in a pair of hockey skates.
These two worlds—inventing and hockey—have collided for Whitehead recently. His company, Light AI, has been trialing a smartphone-based COVID-19 test with a host of Western Hockey League teams, including the Vancouver Giants.
Picture yourself as a junior hockey player. You arrive at the rink hopeful that you are healthy enough to play in that night’s game. Before you enter, you pull out your phone and open an app. You are greeted by a questionnaire that seeks the answers to questions many of us have asked ourselves over the pandemic:
“How are you feeling?”
“Have you been in contact with anyone with COVID-19?”
“Are you experiencing any symptoms?”
The final tap prompts you to open your phone’s camera. You take perhaps your first-ever uvula selfie. Within 10 minutes, you get the results.
Whitehead likens it to what flashlight and popsicle stick wielding doctors have seen for generations. Say ahhhh: “doctors have been looking at throats with their little flashlight and no one's really quantifying what's going on back there.”
So, that’s exactly what Light AI did. They spent the better part of a decade looking at throats, comparing healthy ones to throats with strep or influenza. Collaborating with UCLA and Labcorp, “basically the largest lab company in the world,” according to Whitehead’s estimation, they built a database of tens of thousands of images. These images are stored in the cloud.
Essentially, this research was a way to combat wrongfully prescribed medication. “80% of the people that show up in the emergency rooms with strep throat actually have a viral infection, not strep which is a bacterial infection. Because it's a 24 to 48 hour turnaround to get a result, a big percentage of patients are given antibiotics and told to go home. So, there's an enormous misuse of antibiotics.”
COVID-19 was an opportunity to add to the database they are building. “We're just adding to a platform,” Whitehead clarifies. UCLA and LabCorp “do close to 200,000 COVID samples a day. We are in these big pipelines of data collection. As someone is starting to line up [for the sample] or they’re in the emergency room, the researchers will take a picture of their throat.”
Whitehead and his team are also breaking new ground on COVID-19 symptoms. “It’s unbelievable how much we've seen in our data set and how it's becoming a really interesting research project. For example, we're seeing a mass amount of edema in the back of people’s throats and they didn’t even know they have COVID.”
From here, the AI takes over. Whitehead and his team have trained an algorithm, based on this database, to differentiate between covid positive and covid negative throats. “We're seeing fascinating differences in patients with strep or influenza, and we built several buckets to start quantifying what you're seeing and how to diagnose it.”
One such bucket is variants.
You don’t need to leave Vancouver hockey circles to see the grave reality of mutations of COVID-19. The Brazilian variant is hypothesized to be the culprit in what placed almost the entirety of the Vancouver Canucks on the National Hockey League’s COVID-19 Protocol. “You know, that new variant, which I hear is the primary cause of the first patient, is really evil and highly contagious,” Whitehead acknowledged.
Light AI has leaned on the support of Labcorp and the PCR tests running alongside their trial to understand the strains and variants of COVID-19. “We get to see which strain it is and Washington state's doing a good job of collecting that kind of data for us and we're finding that the variant is not impacting our data. Sure, symptoms are more aggressive but the upper respiratory diagnostics are very similar.”
While working with hockey players is an opportunity—high level sporting endeavours are covid testing and data goldmines—it also represents a weakness. The entirety of the WHL are males and league rules stipulate that their age range from 15 to 21 years old. Also, hockey is an overwhelmingly white sport. For a company that aims to create a robust database for AI to communicate with, this is far from a perfect data set.
Whitehead acknowledges this and understands that to appeal to larger regulatory bodies, a more diverse data set would be required. “Our trial with the Giants does not offer the stratified data sets that you need to have FDA approval because you need to have different ages, different sexes, demographics of that nature.”
What this trial does offer Whitehead and Light AI, ultimately, is an opportunity to showcase their robust ability to translate ideas and data into accessible technology. Thinking back to his entrepreneurial debut, Whitehead doesn’t know if his invention made it into a Casio product. Regardless, he remembers the moment fondly. “Whether it was used or not I have no idea but I got my skates.” Light AI is destined for more than just skates.