Susan Conover of Piction Health, is building an AI to improve results and reduce costs in the treatment of skin conditions. Piction has amassed 500,000 diagnosed images and is already helping primary care physicians interpret skin problems in patients. A great chat with an inspiring founder.
Highlights:
Sal Daher Thanks Sponsors Purdue University Entrepreneurship and Peter Fasse, Patent Attorney at Fish & Richardson
Sal Daher Introduces Susan Conover of Piction Health
Susan Conover’s History with Melanoma and the Scarcity of Dermatologists
"What are opportunities for identifying skin diseases and how can technology help?"
“...we help elevate every doctor to have that visual expertise of a dermatologist...”
Piction Health’s Data Set Includes 500,000 Diagnosed Images with a Variety of Skin Tones
“We're, right now, focused on rashes.”
“...what makes the most sense to start with is a buyer who's realizing the value of our product now.”
“We estimate we can save these organizations, risk-bearing organizations, $160,000 per doctor per year.”
Piction Health Is Looking for Providers Who Have at least 50% of the Downside Risk of Patient Costs
Looking for a Product Manager and a Senior Machine Learning Engineer
Piction Health Is Partnering with a Big Provider to Develop Its Product Offering
Scaling Dermatology from 40 Patients per Day to 40,000 Patients per Day
The Value of Techstars and MassChallenge to Piction Health
Parting Thoughts – Advice to People Who Are on “Survival Mode” at Work
ANGEL INVEST BOSTON IS SPONSORED BY:
Transcript of, “AI for Skin 2”
Guest: Founder Susan Conover
Sal Daher: I'm really proud to say that the Angel Invest Boston Podcast is sponsored by Purdue University entrepreneurship and Peter Fasse, patent attorney at Fish & Richardson. Purdue is exceptional in its support of its faculty, faculty that's top five engineering school, in helping them get their technology from the lab out to the market, out to industry, out to the clinic.
Peter Fasse is also a great support to entrepreneurs. He is a patent attorney specializing in microfluidics and has been tremendously helpful in some of the startups which I'm involved, including a startup that came out of Purdue, Savran Technologies. I'm proud to have these two sponsors for my podcast.
Welcome to Angel Invest Boston, conversations with Boston's most interesting and dynamic founders and angels. Today, we are privileged to have an alumna of the podcast. Welcome back, Susan Conover.
Susan Conover: Thank you, Sal, so much for having me.
Sal Daher: Susan Conover is the founder of Piction Health, which used to be called LuminDX, so the previous episode is titled LuminDX. They're doing a remarkable thing. They're applying machine learning to the problem of figuring out if a skin lesion is really serious or not, to make that really, really easy for providers, and to improve the level of care and reduce costs. Anyway, Susan, tell us exactly the problem that you're solving with Piction Health, so that people who haven't listened to the podcast before know exactly what Piction Health is doing.
Susan Conover: Of course. In my own experience, I found a concerning lesion on my back and tried to go to a dermatologist who had previously said, "If you see anything new or changing, just call us. We'll make sure to get you in." I heard, "We can see you in three months. Everyone thinks they have melanoma." I went to my primary care physician who biopsied my mole and it ended up being a stage two melanoma. About a month and a half later, I had a second one, and then about eight months later, I had a third one. It taught me some of the challenges around accessing specialty care, especially dermatology, which is the hardest specialty to get into, especially now.
When I was at MIT, I focused on my thesis and zoomed out of it and said, "What are opportunities for identifying skin diseases and how can technology help?" Learned that 2.3 billion people around the world have skin problems, so a huge portion of the overall population. Every year, two-thirds are managed in primary care and because those doctors are doing their best but get limited training in medical school, misdiagnose about half of cases leading to about 23 billion in excess spend related to misdiagnosis every year just in the US.
What we do is we help elevate every doctor to have that visual expertise of a dermatologist in an encounter, save time in evaluating that case, because they're only focused on the most relevant history. Save their organizations substantial spend by being able to get the right disease identified the first time a patient is evaluated.
Sal Daher: Phenomenal. Machines solving a problem, helping humans. They're not going to replace the human element, but it's just going to make it used more efficiently. Tell us, how is it that you are able to now have an AI that is as good at detecting a serious skin condition as a dermatologist? How did you achieve this?
Susan Conover: Certainly. I think that's been one of the big puzzles. Also, how do you get people to pay for it is another one of the big puzzles.
Sal Daher: We'll get to that.
Susan Conover: We'll get to it. Obviously, I came to this problem with my melanoma lens and spoke to the market, talked to over 800 different stakeholders in healthcare, and learned that the bigger and low-hanging fruit opportunity was in rashes like psoriasis or eczema or shingles or rosacea. These are also conditions that get misdiagnosed, and so decided to say, "Let's focus on this area of care that's 80% of problems in primary care and 75% of spend related to skin disease." Often, those cases are kept in primary care rather than referred out, so an opportunity for intervention at that time. We're like, "Okay, let's build this rash detector. Where's the data?" Were really surprised to learn that there wasn't a pre-existing data set for a lot of reasons. One is identifiability for body parts or birthmarks, and a whole host of things.
Sal Daher: I'm sorry, what does identifiability mean?
Susan Conover: Sure. With a CT scan, as long as you don't have the patient name or record on there.
Sal Daher: It has to be de-identified.
Susan Conover: Exactly, de-identification. The definition of de-identification is continually changing. Someone can build an ear detector.
Sal Daher: Oh yes. Recognize people by ears.
Susan Conover: Yes. Privacy and confidentiality is a critical thing. Also, hospitals are really careful about how they manage their patient data, because it can be really expensive and a breach of privacy to have data leaked. They're very careful who they work with, and cybersecurity and everything related to that. We decided to get creative in the pandemic, especially when all these hospitals we could partner with, where COVID response was their focus.
We partnered with over 200 dermatologists in more than 15 different countries, including South Africa, India, Spain, and Bolivia, in order to build the largest and most actionable database in the world of skin diseases that's equitable, that's representative across all different skin tones. Then we use that to train our AI, which now is on par the accuracy level of a dermatologist, which is one of our biggest milestones that we just accomplished.
Sal Daher: Are you willing to state publicly the size of your dataset?
Susan Conover: It's over 500,000 photos.
Sal Daher: These have been diagnosed?
Susan Conover: Yes. Almost all of the data we bring in has at least one dermatologist that's evaluated it and determined the diagnosis.
Sal Daher: The coverage of your AI covers all major skin conditions?
Susan Conover: It's a great question. There are two primary scenarios in that primary care setting of, is this lesion likely to be cancerous? Also, what is this rash? We're really addressing that second one. We still have people who use our product on moles. We just tell them, "Hey, it looks like you took a photo of a mole. We're, right now, focused on rashes. We'll let you know when we're addressing moles in the future."
The definition of a rash is basically just something that's not in a single spot, but may take up an area. It can be raised, and it could be flat, but it's a lot easier to image with a smartphone and get actionable feedback on what it could be. Then we're addressing over 97% of rashes that appear in primary care and, of course, working to increase that.
Sal Daher: How about the melanoma problem that you had?
Susan Conover: We just let doctors know, who are using our product, "Hey, it looks like you're taking a photo of a mole. Right now, we can't address this, but we can potentially address that in the future." There are a few different things related to a product to identify moles that are technical limitations. For now, focused on rashes, but can address it in the future.
Sal Daher: A lesion that could be cancerous, how do you deal with that?
Susan Conover: The vast, vast, vast majority of rashes are not cancerous. There are cases like cutaneous T-cell lymphoma that is often misdiagnosed for eczema or other things. Our intended use case is really for providers to use it in a clinic on a first pass evaluation. If there's something suspicious of concern, or needs to be biopsied or checked out by a dermatologist, we just recommend they refer that case out. Trying to just stay in our lane with our first product, and then can expand to identify rarer and more concerning conditions as we grow.
Sal Daher: Primary care physician uses the app on a patient and that will say, "This should be biopsied." Is it a binary thing?
Susan Conover: Yes, great question. We really do visual search. How it works, or the simplest way to describe it is they use an iPhone in a clinical encounter, take a photo of the rash, answer three simple questions, and instantly see the ranked order of the most visually similar conditions that are both common conditions and also urgent ones they should rule out before moving on. Then we point them toward, what's the additional history to gather in order to differentiate disease number two versus disease number four? Something like that.
Sal Daher: Okay. It's almost like you do a Google search, then you get it back, except you don't have any sponsored responses. Organic responses. [chuckles]
Susan Conover: Yes, exactly. We focus on the most relevant information for them. It's visual search, exactly.
Sal Daher: Okay. Let's talk about the revenue model and where you are now and how that has evolved over time.
Susan Conover: Of course. That's probably the biggest puzzle in digital health, when you're making something new that you can't predicate on something that's been done before. That's why we needed to speak to over 800 different stakeholders and continually iterate. Basically, I guess the last round we raised, we were like-- We could go in many different directions, real-world data, identifying patients earlier for trials, risk bearing organizations. In this last round, we identified that all of those are different options, but what makes the most sense to start with is a buyer who's realizing the value of our product now. We estimate we can save these organizations, risk-bearing organizations, $160,000 per doctor per year. Our next six to nine months is really showing our product helps doctors make better choices. That we're hitting our cost savings numbers that we've estimated, and so we're identifying the right partners for us at this stage. We have a few different ones in our pipeline. We already started we're with the Air Force. Basically, our next phase is showing those numbers and then aimed at charging about 15% of the money we can save, which is a very similar model to patient paying another Boston health tech company that just exited for 21 times [unintelligible 00:12:00] last summer.
Sal Daher: Excellent. Would you review again, so basically, payers, are you talking about insurers or ACOs?
Susan Conover: Of course. We're focused on organizations that have at least 20,000 Medicare Advantage lives covered. That's publicly available information where we can know their patients are at least 50% downside risk or more at risk. Those are the organizations that can save the most on the long-term downstream cost, so we can reduce-
Sal Daher: I don't understand the 50% downside. What does that mean?
Susan Conover: Sure. Almost every healthcare provider in the US says, "We're under risk-based contracts." The trickiest thing about that is they'll take very scoped, limited risk. We will get a bundled payment every time we do a hip transplant or a hip replacement, and we'll get paid a certain amount of money, then if something goes wrong, we're on the hook for it. A lot of these current risk agreements are not fully capitated. They're like, "We're only going to do the hip replacement. We're not going to take on that patient's mental health needs or other areas of medical spend." Or like, "When we refer out, a dermatologist or specialist has a different method of payment." That's 85% of the US healthcare market is more unstructured and fee for service, where a doctor gets paid to do a thing. We are focused on the 15% of the market that is like, "No, we're actually going to take on that full patient risk and be paid by CMS X dollars a month and just figure it out." That market is growing 20% year over year and has large parts of the international market in the UK, Canada, other regions-
Sal Daher: 50% downside means that their downside is limited?
Susan Conover: At least 50% downside, but yes, 50% to 100% downside risk because there are a lot of providers that take on upside risk of, "We'll get paid if we do well, but we're not exposed if we have very expensive patients." It's really amazing for making sure standards happen. A1C measurements are recorded for all the patients needed and stuff like that, but organizations-- Because we don't fit into a perfect quality bucket, we're looking for that organization who's like, "We've already identified the most expensive patients who are frequent flyers. Now we're focused on empowering primary care to do more and deliver the best care possible for patient. We're not just thinking of primary care as a loss leader service that's intended to refer out to specialists, and that's where we make our money." Which is a lot of medicine today.
Sal Daher: In a way, you're targeting not the most expensive patient population, but a moderately expensive population where any cost savings will be significant.
Susan Conover: Exactly. We're putting ourselves on the hook for making sure our product works and making sure patients get better faster.
Sal Daher: Very good. What talent do you need in this effort?
Susan Conover: We are in the process of closing around, which we expect to close in the next month, month and a half. We do have ambitious goals for bringing on and expanding the team. One person that we're bringing on is a product manager. I love and am passionate about product, but I just don't have the bandwidth to own it. Looking for someone to bring on that primary care physician's perspective, but also make sure our product is tracking against the metrics we say it will for the buyers within these agreements.
Also, a senior machine learning engineer. We've done a lot, obviously, but we're always looking to push that boundary, be able to identify rare diseases better, be able to make sure we give quality of care in all subsegments of the population our product can be used on. Looking for someone who can bring some research methods to improving clinical care.
Sal Daher: If you're a brilliant product person that wants to join up to a really exciting team, give a shout-out to Piction Health, and also senior machine learning engineer. This is a fun group to work with. How fun? Susan Conover met her co-founder at improv. That's how fun. [unintelligible 00:16:52] a great place to work. That is tremendous. I understand that you're negotiating various types of partnerships to increase your dataset?
Susan Conover: Yes. We're very excited to announce a partnership with a large academic medical center that will help us grow and scale in the ways that we need to grow and scale from a product and product development perspective. Also, run that validation, because they have, of course, a primary care organization and really excited dermatologists who want to be part of the future of what dermatology looks like. The writing's on the wall of computer vision can really be on par with an expert at just simple tasks of bucketing and identifying visual patterns within different images.
Sal Daher: This is happening in the space of pathology. It's happening in a lot of areas. These machines are becoming extremely helpful to human beings. I'm even invested in a company that uses natural language processing to grade the AP exams. They're giving humans superpowers, figuring stuff out, which is very mundane stuff. They're not doing higher function stuff. It's very mundane stuff in eliminating errors, so it's really interesting. Susan, is there anything else that you'd like to touch on in terms of the update that you've given?
We've talked about the number of diagnosed images you have, the progress you've made, the staffing that you're looking for. You're looking for more pilot partnerships with players, so consider this an invitation to organizations. If you know an organization that could benefit from the significant reduction of costs in the patient population that Susan mentioned, you should get in touch and connect. I think there's a potential for improving patient care and also reducing costs.
Susan Conover: I think one of the coolest things about what we're building is that we bring the expertise that a dermatologist can use to help 40 patients a day and extend that to 40,000. We can start to make healthcare scalable in a way that it hasn't been able to scale up and tailored, very customized and tailored to every application. I think that's one of the really cool things that AI can help doctors with. Then they can focus on the relationship with the patient and counseling them.
They can focus on the really important human elements rather than focusing on the logistics of what's most common and flipping through a textbook and Google Image searching and the other ways that they solve this problem now. I guess one thing that I've learned in my startup journey is it takes a community Is that we need phenomenal people to work across different areas of deep learning, of deep market expertise. Phenomenal people to bring on the team to build a product or extend the product into different markets or track how it works. We're always looking for really excited people who want to make an impact. If that's on the hiring side, that's great. If that's on investing side, that's great too, or just getting the word out there of what we're doing.
I am lucky enough to follow in the footsteps of quite a few different alumni from MIT and from Techstars and other institutions in Boston that have already built phenomenal healthcare AI companies. There's at least a little bit of a playbook, but I think what's most important is having the right community, including people like us, Sal, to help us grow.
Sal Daher: I'm such a scatterbrain these days. It's interesting you bring up Techstars. In concrete terms, how did the business change for you before Techstars and after Techstars? What did Techstars help you do?
Susan Conover: Certainly, before Techstars, I was like, "We don't need Techstars. We'll do it ourselves," and like any founder, was like, "I'll figure it out." [laughs] I think what we learned is, certainly, fundraising, that's the obvious thing, is how to connect with investors in a way that you're meeting them where they're at for how they analyze companies. That's really important, and running that as a process.
There's a lot of institutional knowledge about how to fundraise that I had no idea about. There's also identifying the top 5 or 10 people in Boston that if they care about you, your company can grow and not be the same company a month from now. That was really instrumental. Then some basics of running a company, like running up a KPI or OKR process where you know how you're tracking, you know how you're measuring progress and you know you're only going to focus on this section of the funnel this week and not be overwhelmed by everything.
Stuff like that, or how to run a board meeting. Just basics that you just don't even think about but are important, because these are multiple tools in your toolkit to help you grow and scale. Also, just the community that we built at Techstars is certainly a challenging program that pushes you. You're like, "I have to run a business, and also I have to learn all these new [laughs] things." Having looked back, it's been one of the best decisions we've ever made.
Sal Daher: That is tremendous. I have yet to run into anyone on and off the record who says anything less than, "Techstars was just a huge plus for me. It made a huge difference." Not just Techstars, I must also say that MassChallenge also gets excellent rates in my experience. Each one has their own particular special thing to offer.
Susan Conover: Yes. We were also in MassChallenge the year prior. The chief or the VP of marketing from Vistaprint ended up-- I've been working with her. She helped our rebrand from LuminDx to Piction Health. She's been such a great supporter and also, eventually, angel investor, and our regulatory expert committed to our round. He was the first person to commit to our round that we have been recently raising on email. [laughs]
It's partially because we've been able to build relationships with those folks. MassChallenge is an amazing community to be part of. Just plug you into the other weirdos in Startupland who are like, "I believe in a dream." [laughs]
Sal Daher: Yes. Those people, there's so few of them in the larger population that it's important to go to a place where they're likely to gather.
Susan Conover: Exactly, so you don't, every day, at the end of the day, say, "Am I crazy?" They're like, "You have at least other people who are crazy around you." [laughjs]
Sal Daher: It’s a completely abnormal thing to do, to try to do something completely new that has never been done before. Just abnormal, it can be lonely. That support is really important. Anyway, Susan, take a moment and think, is there any parting thoughts that you want to leave our audience, our founders, and people who are supporting startups by investing in them or advising them that you'd like to communicate about Piction Health or a bunch of experience in general?
Susan Conover: Yes. I had jobs before this one. I had various levels of satisfaction with those jobs, but then with this job, it's very hard, obviously, but also is incredibly rewarding. I get to be able to help potentially millions to billions of people who've been in the same shoes I have. I know it's a wild time for the hiring and people changing jobs with the great resignation and everything like that. There's so many people I know who are in survival mode and like, "How do I get through the day? How do I get through the month?" If you could flip the script, instead of saying, "How do I survive? How do I thrive?" and you make sure you're making conscious decisions to support that vision you have for your future, that's my best recommendation. It's just, don't let the urgent fires get in the way of what you-- When you are on your deathbed, what do you want to look back and be able to say you've done? Focus on that.
Sal Daher: That is very inspiring. That is very inspiring. Tremendous. Susan Conover, extraordinary founder of Piction Health, which is creating an AI to help better diagnose skin conditions and to improve the care and reduce the cost of health care. Thanks for being on the Angel Invest Boston Podcast.
Susan Conover: Thank you so much, Sal, for having me.
Sal Daher: Thank you. I'm Sal Daher.
[music]
Sal Daher: I'm glad you were able to join us. Our engineer is Raul Rosa. Our theme is composed by John McKusick. Our graphic design is by Katharine Woodman-Maynard. Our host is coached by Grace Daher.