LightIR: Enabling The Cancer Surgeries Of The Future
About half of humanity will develop cancer in their lives. For another half of these people, or a quarter of the world’s entire population, cancer is going to be what takes their lives, too.
While cancer is the second leading cause of death worldwide, I’d argue that cancer isn’t the only problem we face. The more glaring issue how we approach cancer.
For example: the leading treatment for solid cancers today is a surgery known as a tumour resection. Over 8M cancer patients this year will need at least one as part of their treatment. And yet, for such a straightforward operation, close to half of everyone who goes through one won’t get cured the first time:
In a situations like these, it’s hard not to wonder how our healthcare system became so slow and disappointing. How could almost 5M people go through our leading cancer treatment and receive the same effectiveness as a coin flip?
How, in the 21st century, when modern computing can let us drive cars and send rockets to the moon, did surgery get left behind?
We wondered what was stopping us from radically automating cancer treatment, and it led us to create LightIR. LightIR is the world’s first device that can detect cancer during surgeries, at the cellular level.
Today, resections come with about just as much room for error as they do success. They’re so precise that even a single cancer cell left behind after the procedure could grow back into a new tumour. And yet, surgeons have no way of seeing cancer cells.
While some parts of tumours can look obvious to anyone, others can look and feel like they healthy tissue that surrounds them, so surgeons can’t tell where the cancer begins or ends.
Tumours aren’t always black-and-white, so our only way of knowing for sure if we’ve gotten every single cell is with what’s called pathology test.
During a resection, surgeons cut out (excise) a patient’s tumour, along with a thin rim of healthy tissue around it to increase the odds they removed it completely. They then ship those samples to a facility called a pathology lab, where professionals analyze it in detail.
While the process eventually gets pretty complex, pathologists are really looking for one thing: margin status.
Think of margin status as an indicator that lets surgeons know if there could be any cancer left in their patient. In an ideal scenario, when pathologists observe a cross-section of a specimen under a microscope, they should see a dense cluster of cancer cells (the tumour), encased by a border of healthy cells going all the way around it.
That border’s called a margin, and when it contains the entire cancer inside of it, it’s a clear margin:
But often times, those cancer cells won’t form a well-defined circle. Instead, they can get dangerously close to the edge, or breach it entirely.
That’s called a positive margin. It means the sample only contained part of the entire cancer, and the rest of it is still inside the patient:
Pathology labs might be accurate, but after processing the tissue and painstakingly analyzing every cell, getting a result back can take weeks.
By then, the surgery is already over, and if there was ever a time to remove more of the cancer, it’s long gone. If the margin comes back positive, the only option left is to do the surgery all over again.
Except, a new surgery is a lot more than just an inconvenience.
For people suffering from cancer, every resection comes with a serious mortality rate, and even a chance of spreading further. On top of that, the least expensive resections start at around $5,000, but others can be as expensive as $40,000. And more often than not, families are forced to bear those bills alone.
But worst of all, this all happens because a test wasn’t fast enough to get a diagnosis. It’s completely preventable.
That’s what LightIR was designed to solve. LightIR combines microscopy with computer vision to analyze tissue in real time, letting it pack down the power of a pathology lab into a probe that can fit into a surgeon’s hand:
We developed LightIR to come in two main parts — a device and a software. Let’s understand things from the very basics.
When a surgeon places LightIR on a patient’s tissue for a diagnosis, a cascade of events gets triggered into motion.
Immediately after starting, the device begins using its built-in camera sensors to capture magnified images of its surroundings in real-time. Depending on its settings, these images can be anywhere from 50–100X larger the life, letting them conserve the slightest details in every fibre of tissue:
Then, after converting its vision into a digital format, the probe instantly beams its data to any USB-compatible device its connected to. Think of the process the same way you would a relay race. The probe’s goal is to pass its batons (or images) to an electronic device, where they can be processed and used to create a diagnosis.
From here, the software takes over. Built into LightIR’s interface is an AI called a Convolutional Neural Network (CNN) that’s waiting to get called to action behind the scenes. A CNN is a special kind of AI that can detect complex patterns in images and classify them into buckets, like “Cancer”, “Healthy Tissue”, or “Muscle” at lightning speed:
LightIR’s AI was trained on thousands images of cancer cells (independently sourced from a lab), and it could detect micron-sized clusters of them with >99% accuracy.
From start to finish, LightIR displays its live video feed for surgeons to inspect and measure. And when they’re ready, they can click a button to get an accurate diagnosis — all in less than a tenth of a second.
With the power of automation, LightIR lets surgeons make superhuman decisions at superhuman speeds — time and time again. But more importantly, those decisions translate to healthier, happier patients.
But let’s quantify just how much of a difference a faster test can make.
As we mentioned earlier, resections can cost about $5,000 each on the low-end. And with the snail-like turnaround times of today’s pathology labs, over 4M resections fail each year:
If LightIR drops that failure rate by just 50%, it could save 2M unneeded surgeries, or $10B / year in surgical costs. And considering how hospital stays and medicine can cost thousands more, its full impact could be even bigger.
But more importantly, about 150,000 people will die from failed resections this year. LightIR could save over 70,000 of them. That’s almost 20 billion extra hours of people living, just because LightIR existed.
A Way Ahead
Our vision at LightIR is a lot larger than just creating medical devices. We want to build on humanity’s most cutting-edge technologies and use them to solve the world’s most pressing diseases. We want our products to accelerate medicine into the 21st century and beyond.
That’s why LightIR is just the first step in our journey.
It’s still quite the ambitious first step, though. While LightIR is functional today, it’s an investigational device. To take things to the next level, we’ll need a massive windfall of evidence to prove LightIR can measure up to pathology labs — no matter what the situation.
And yes, even without pathology labs, every single cancer cell still matters. 99% accuracy is exponentially better than the 50–50 odds we have today, but it’s still not 100%. However, studies on multiple types of cancer have proven smaller volumes of residual cancer in patients are 5-6X less likely to return after resections than larger ones.
Of course, that’s an improvement the vast majority of people could benefit from. Statistically speaking though, our solution is still fallible. Even so, full accuracy is something we committed to from the very start, and we believe the path to get there exists:
That’s why we’re planning on running our first data collection trial from cancer centres starting in late 2021. After scanning dozens of breast cancer specimens at our partnering facilities, we expect to gather thousands of new images to re-train our AI with. And if things go as planned, it could get LightIR years closer to the operating theatre.
In the short term, this means introducing the device into the most common resections for the most sensitive cancers. When it comes to vital areas like the brain or lungs, every millimetre of healthy tissue counts, and our early strategy with LightIR is to enable landslide improvements where we need them most.
But moving forward, we plan on setting our sights miles further:
We see day where LightIR grows into its full potential — far beyond the walls of hospital and clinics. We believe in a future where LightIR’s AI doesn’t just identify every type of cancer, but any disease anyone could ever have.
Till then, there’s a lot to be done, but we aren’t going to hesitate to bridge the gaps we need to get there. So, even though it might take a while until you hear about the first LightIR-powered surgery, remember that when you do, you’ll be living in the future.
Thanks for reading,
Aaryan Harshith / Founder @ LightIR
— — — A Note From Me — — —
**Hi there! To anyone who’s gotten here, this is Aaryan (the person who wrote this article). I hope reading about what I’ve been building was an interesting experience, and that it gave you a more optimistic view of our future. That being said, LightIR is a constant work in progress.
If you believe in what this device is trying to accomplish, don’t hesitate to contact me (firstname.lastname@example.org) if you’re a surgeon, researcher or investor in this space, or if you happen to know any. Oh, and as always, questions + feature requests are open 24/7.
Once again, thanks for reading. I’ll see you sometime soon.**