Seen by many as the key to self-driving cars, computer vision is one of the tech industry’s fastest-developing fields. But can those advances also help with the detection and treatment of skin cancer?
The founders of SkinIO, a Chicago-based healthtech startup, certainly think so.
“For all skin cancers, if you catch it early enough, you will have a cure rate of 99 percent or higher,” said co-founder Dr. Jean Christophe Lapiere. “So the key is to check frequently.”
A dermatologist and skin cancer surgeon, Lapiere was an attending physician at Northwestern University before opening his own practice that focuses on screening, diagnosis and treatment. Concerned about the number of patients who came in with late-stage skin cancer, Lapiere came up with the idea for an app that could catch suspicious moles early.
Designed to be as easy to use as possible, SkinIO lets a patient capture a full-body photography in less than five minutes. The patient’s images are then run through a computer vision-powered mapping system that automatically flags moles and other skin aberrations. The flagged images are then passed on to a team of in-house dermatologists for review.
The SkinIO app notifies users to perform a scan once per month, comparing captured photos as a time series. If the team spots any new or changing moles, patients get a warning that they should schedule an in-person appointment.
In addition to its in-house dermatologists, SkinIO works with a network of external dermatologists. These doctors can be granted access to the full image time series with the patient’s permission, to review how moles and aberrations have developed over time.
Co-founder Kyoko Crawford said skin cancer detection is an interesting challenge from a computer vision standpoint, because detecting moles requires far more nuance than spotting oncoming cars or distinguishing bushes from traffic cones.
The startup’s initial results are promising. Crawford said the startup recently performed a clinical study with Northwestern University’s Department of Dermatology, in which its algorithm detected new or changed lesions with a sensitivity of 92 percent.
That accuracy, said Crawford, will only get better as the platform is exposed to more data. That said, Lapiere was careful to note the app is only an early warning tool — a real skin cancer diagnosis can only be performed through an in-person consultation.
In addition to helping patients receive treatment earlier, Crawford said SkinIO hopes to further the field of skin cancer research by expanding the data sets and tools available to researchers.
“To date, there hasn’t been a lot of data about how moles evolve,” she said. “We are constantly training our deep learning model with all of the new data points we get in.”
Founded in 2015, SkinIO currently has nine employees in Chicago, the bulk of which are on the tech side. The company has raised $1.5 million in external funding to date.
SkinIO just launched its app in the iOS App store, but its service is currently only available for patients across Illinois. Crawford said SkinIO has other states on its launch roadmap.
“We do have an in-house dermatology team, which has licensing requirements,” she said. “As we expand into other states, our goal is to expand that team to appropriately serve patients in other states, as well as to expand our dermatologist networks in those states.”
Images via SkinIO.