Skin cancer is one of the most common—and treatable—forms of cancer, but only when caught early. Unfortunately, many existing tools still fall short for people with darker skin tones. SkinCheck is working to change that.
With an AI-powered mobile app designed to identify skin lesions across all skin tones, SkinCheck is bringing more equity to early detection and helping reduce diagnostic bias in dermatology.
Through our Reshape Health Grants, we’re supporting their efforts on two key fronts: improving model performance in real-world conditions, including variable lighting and image quality, and strengthening accuracy across a broader spectrum of skin tones, ensuring no one is left out of effective early diagnosis.
We sat down with Julian Abhari, CTO, and Daniel Marques, CEO of SkinCheck, to learn more about the company’s mission, the technology behind their app, and how our collaboration is helping make inclusive skin cancer detection a reality.
Q: Thank you, Julian and Daniel for your time with us. Please tell us more about you.
JA: Hello. I’m Julian Abhari. I’m the cofounder and CTO at Skincheck.
DM: I am Dani Marques. I am the cofounder and CEO here at SkinCheck.
JA: I actually started building SkinCheck after my mom battled multiple cases of skin cancer growing up during my childhood. And it was in high school that initially, I was really interested in machine learning and app development, and I created a computer vision prototype of an algorithm that would be able to detect skin cancer patterns. And most importantly, it was able to run natively, on a mobile smartphone.
And then it was after an initial success where I had shared this beta application with one of my English teachers in high school. He was able to, he noticed that he had some red spots on his arm. He used it, and he had never been to a dermatologist before. And when the AI reported that these spots could be indicative of a basal cell carcinoma skin cancer, He went to a dermatologist, got it checked out, and it turns out it actually was a skin cancer. It was basal cell, and he got it removed.
And as soon as that happened, I knew that this was a really special technology. So I went to university, and I continued to research and develop this technology alongside professors and experts, to the point where we were able to achieve accuracy on par with the dermatologists. It was there that I knew that there was a really exciting path forward in order to make this technology actually come to life, which was to make it work for all skin tones as machine learning always echoes the data that it’s trained on and particularly in skin cancer. There was a lot of racial biases. So I applied for a National Science Foundation grant, where I went to Atlanta and I worked with top researchers at one of the top computer vision institutes in Atlanta to mitigate these racial biases.
As soon as that technology, as soon as I was able to successfully invent a new machine learning architecture to mitigate these racial biases, that was when I was able to reconnect with one of my friends from high school who had actually gone on to create a lot of exciting ventures, and we decided to build SkinCheck as a start up company.
Q: What is SkinCheck’s mission?
DM: SkinCheck was built initially to solve a problem that we are extremely familiar with, a problem that we both handled inside our own household. So we wanted to build a solution that was able to help individuals, but also to be able to provide a solution to the actual health care system, a solution that would be an end-to-end solution to the actual health care provider.
So our product and solution nowadays is that we have a skin health management platform, which is the SkinCheck app. And on the SkinCheck app, users, patients, and frontline health care workers are able to monitor their skin health or their patients’ skin health. They’re able to track the lesion growth or the condition’s growth over time, and they are also able to monitor their skin lesions and the skin conditions of their whole body over time. And this upcoming April, they will actually be able to connect directly to a medical provider directly through the SkinCheck app. So in this product, we’re able to help individuals expedite their access to their dermatological treatment, being able to get a medical recommendation from a board-certified dermatologist within twenty-four hours instead of having to wait three to four months.
Additionally, what we have, what we are building, is a clinical decision support tool for skin cancer detection for primary care physicians. So we’re developing, we have built the world’s first algorithm that is able to detect skin cancer in a computer vision-based way for all skin tones. And now we are pursuing clinical studies and clinical trials to actually be able to bring this clinical decision support tool to aid primary care physicians on determining if skin lesions are concerning or not. So we also have the algorithm in a provider-facing platform that allows providers to prioritize patients that might have a bigger risk of skin cancer.
Q: What sets SkinCheck apart from other solutions?
JA: SkinCheck is fundamentally different than other computer vision based skin cancer detection technologies. As one of our biggest innovations is the fact that the patent-pending, technology that I invented was specifically for mitigating racial biases as this is currently what has prevented other technologies in this space from succeeding in their pivotal studies. What makes SkinCheck so unique is that, by providing this incredibly unique technology, we’re able to show that it is on par with the dermatologist’s level of accuracy across all skin tones and demographics that enables us to perform these clinical studies and succeed in our pivotal studies, seeking to achieve FDA approval. So, essentially, where machine learning in the past has, especially in this space, suffered from a loss of racial biases that have not only made it to where these clinical studies didn’t work out but we have surpassed that and overcome that challenge. And this enables us to provide our end goal of hopefully becoming the first computer vision-based AI skin cancer detector to achieve FDA approval.
DM: And, additionally, the focus that we have with the SkinCheck app and SkinCheck connect is being able to aid users and patients in their whole skin health journey and not only as a point solution, which allows individuals to detect concerns earlier on and therefore, be treated earlier. And we focus specifically on a market of employers that offer incentives for preventative care. So we focus really in helping the employees of those employers that offer incentives for preventative care.
Q: What is the core problem that SkinCheck solves?
JA: So at least one in every five Americans will be diagnosed with skin cancer before the age of 70. But what makes matters worse is, contrary to misconception, skin cancer is far more lethal for people of color. As in fact, there’s a thirty percent higher death rate for black men from skin cancers like melanoma more than any other demographic. But melanoma is a type of skin cancer that be can become as lethal and it’s just as short as six weeks, where the average wait time to get a skin cancer screening is of fourteen weeks. SkinCheck is a solution that enables patients to expedite their access to treatment and empowers health care providers to intervene when necessary.
One of the reasons why I know that it takes an incredibly long amount of time to get a skin cancer screening is because, unfortunately, when my mom was 27, she was diagnosed with multiple cases of skin cancer, and she battled these cancers all throughout my childhood. But fortunately for my family, about ninety-nine percent of all skin cancer cases are curable if detected early enough. The question is, is it possible for you to detect your skin cancer early?
Q: Who are SkinCheck’s primary users?
DM: Our primary users at SkinCheck are individuals that are concerned about their skin health and would like to expedite their access to treatment, monitor their skin health, and take control of their skin health directly through a platform on their phone. In that platform, those individuals, patients, are able to get educated about their skin health, but also are able to learn more about conditions and potential symptoms. On the SkinCheck AI side, our primary customers are primary care physicians and health systems that have a built-in specialty clinics, and they want to help those physicians make best, better decisions in regards of skin health assessments.
Q: How does SkinCheck leverage AI for skin cancer detection?
JA: So for medical professionals like primary care providers, we enable them to use our AI for skin cancer detection, directly in our provider-facing skin check platform. And the way that we compute and make predictions on the AI is by having it run natively, on device, which enables secure, privacy, as well as expedited accessibility, and access to where, regardless of where you are, as long as you have the SkinCheck platform, this AI will function as an AI’s medical device and be able to run directly on the native hardware.
Our visual skin cancer detection AI enables us to allow our medical professionals to intervene when necessary, which is one of the core components of being able to enable early detection for skin cancer.
Our AI technology for skin cancer detection is what we’re currently building as a medical device. One that our clinical studies are seeking to prove the validation of primary care providers being able to intervene when necessary and detect concerning skin lesions from their patients. And what makes our technology so unique is, of course, the fact that it was built to work for all skin tones, and now it’s become actually the first visual skin cancer detector that works for all skin tones. And this technology is one that throughout our SkinCheck platform, whether user or medical professional, is very much focused on optimizing and improving our technology, collecting data, in order to make sure that the accuracy, sensitivity, and specificity is on par, allowing us to achieve FDA approval.
Q: What were SkinCheck’s needs before winning the Grant?
DM: So our needs before winning the grant were many and far between. From a development standpoint, there has been a lot of UX and UI development that we’ve been needing to build and execute on.
There’s been a lot of different functionalities for the user side that we’ve been wanting to build as well as a a few needs we have had on the actual AI development side.
JA: Currently we face on our AI technology about three crucial issues that require far more research and development. One is in regards of stabilizing computer vision variability, basically, enabling us and our AI technology to standardize across the position, the angle, the distance, the lighting. And this will be built out entirely in the scanning functionality of our platform, which requires a lot of UX and UI research and development. And then at the same time, will require us to ensure that our models are extremely robust to adversarial attacks and, of course, any variability in the position angle distance lighting to ensure that our AI is as robust as possible. That’s one particular challenge that’s really important for us to go forward.
Another one is specifically in maximizing the sensitivity and specificity scores, for skin cancer overall, which requires us to improve and expand our classifications to multiple different types of skin lesions. And then finally, enabling us to mitigate racial biases, it’s a really core challenge that we’re able to effectively create and maximize our usage of the small minority of datasets that include darker skin tones for us to ensure that representation from communities of Pacific Islander, African, African American are sufficiently represented and that our models are able to check and correct the learned differences.
Q: What are your expectations after the grant process and how would you measure success?
DM: We expect to have developed and learned a lot about how to develop extremely comprehensive, extremely intuitive, and educational UX and UI in our platform. We want to increase our retention and engagement with our user base, as well as being able to, understand how we can best guide individuals in meeting them where they are in their skin health journey. So, expect to improve and learn a lot about how we can better serve and actually, have significant developments on the UX and UI side specifically for the app.
JA: A a really important success metric for us would be in ensuring that after this grant that the research and development that has come from improving our AI models has directly impacted our metrics of sensitivity and specificity across all skin tones for detecting and classifying different skin lesions from benign and cancerous lesions. And it would be also at the same time really important for us to evaluate the metric of reliability in the platform, across the different variability that can come from cameras and computer vision in general.
DM: On the UX/UI and on the app side, measurements of success would be an increase in number of users, increase in monthly active users and engagement and retention, as well as improving some of our feedback metrics on the existing feedback loops that we have with users. But, yeah, that’s one of a few of the areas that we would be able to measure success before and after the grant.
We’re extremely excited to participate on the grant. We’re extremely honored to have been selected and work alongside the Arionkoder team to improve our app and usability for the over 50,000 users that we currently have as well as having the expertise of the machine learning and AI experts from the Arionkoder team to improve also our machine learning and AI models.