Telehealth tech is making a very positive contribution to healthcare and is currently being used in a variety of ways. But telemedicine does have certain limitations when it comes to treating patients.

The latest report on the Future of Healthcare published in The Times newspaper explores the use of AI in cancer care.

Due to the COVID-19 pandemic, routine cancer screenings, and urgent referrals and treatments, have been delayed or canceled, leading to a backlog of patients. In a worst-case scenario, this can lead to up to 35,000 additional cancer deaths within a year in the UK alone, reports Abby Young-Powell.

One challenge in oncology is early diagnosis. AI in cancer care could help find innovative ways to reach individuals faster and identify early signs of cancer. This is precisely where our portfolio company Ezra comes to the stage.

Emi Gal gave a keynote speech at LDV Vision Summit in 2019 © Robert Wright

At LDV Capital, we invest in people building visual technology businesses. We organize the annual LDV Vision Summit, a premier global gathering of people in the visual tech sector.

In his keynote, Emi Gal, co-founder, and CEO of Ezra, showcases the tech his company is utilizing to detect prostate cancer and how he believes computer vision will revolutionize MRI-based cancer screening.

Emi Gal at LDV Vision Summit © Robert Wright

Why are we not detecting cancer early enough?

Here’s an uncomfortable truth: one in three people will get cancer; over 50% of people are detected late; only two out of ten people who are detected late survive longer than five years

I learned it's because there is no way to screen for cancer everywhere in the body that is fast, accurate, and affordable, so I went on this quest to see whether we could build a better test.

Now let's say you're a responsible individual like Evan, and you want to get screened for cancer today. What are your options?

What you see on your left is five of the highest incidents of cancers. To screen for these cancers, you would have to do five different tests, most of which are invasive or intrusive. What you see on your right is the other seven highest incidents cancers, and the bad news for these is that they don't have any screening guidelines. You only get tested for these cancers if you have symptoms, and if you have symptoms, it's generally too late.

Our goal at Ezra was to take these tests and try to replace them with a single test that you can do every year. I have a computer science background, but I don't have a medical education, so I went on a research spree. I spent two years focusing on ideas ranging from digital microfluidic biochips to DNA-based nanobots, trying to find a way to screen for cancer.

Every Summer, we produce a proprietary LDV Insights research report, where we deep dive into a sector where visual technology will revolutionize businesses. Learn about the latest advances in visual technologies and the role of artificial intelligence in medical diagnostics in our LDV Insights on Healthcare. It’s interesting to note that we got to know Ezra when we were working on this slide:

And then in 2016, on my honeymoon, I was reading research papers (isn’t it what you do on your honeymoon?!), and I had this idea – what if you could do a full-body MRI and use AI to decrease the time that you spend scanning and the analysis to make it affordable? My wife loved the idea, and that's how Ezra was born.

To show that MRI and AI can be used for cancer screening, we decided to focus on one cancer, and we started with the prostate. We chose this type of cancer because one in nine men will get prostate cancer during their lifetime. One in five African American men will get this disease. The standard way to screen for prostate cancer is pretty invasive and not very accurate; men have to undergo a PSA blood test followed by a prostate biopsy. If you've never done a prostate biopsy, it's done by putting a needle through the rectum and poking the prostate at random in 10 different locations, which is not an experience that anyone would like to have. One million prostate biopsies are done every year in the US, and 800,000 of them are likely unnecessary.

At Ezra, we're replacing that with a 20-minute prostate MRI. You sign up on Ezra website, you get a prostate MRI. We send you a report generated by a radiologist assisted by our AI within three days. MRI hasn't been used for screening because this technology is an expensive imaging modality. A prostate MRI can cost up to $4,000. At Ezra, we're using AI to decrease the scanning time and the analysis time to make prostate MRI affordable.

Our AI algorithm helps radiologists in their analysis process, and it automates many things that they dislike, like measuring the size of the prostate, size of lesions, locations of lesions. We've fed our algorithm with MRI sequences that were annotated by expert radiologists, and then we trained it to output segmentations for lesions.

What you see on your left is a radiologist segmentation for prostate and lesion. In the middle is an AI segmentation for prostate and lesion, and on the right is a heat map of what the AI sees.

Here’s another example:

As you can see, the segmentation from the AI is almost indistinguishable from a radiologist. The accuracy of our AI is 93%, it’s more accurate than radiologists in identifying clinically significant prostate cancer.

We went live with our prostate cancer screening program in January of 2019. As our AI can generate segmentations, we can also create 3D models of prostates and lesions. This visualization can not only help radiologists to analyze the problem but also make it easier for patients to understand the problem.

After Ezra’s successful release of prostate cancer screening, we expanded to full-body cancer screenings covering 13 cancers in women and 11 cancers in men. The company recently announced COVID 360 – an opportunity to check lungs and blood antibody status for COVID-19 effects.

You may also want to hear experts from Harvard Medical School, Facebook AI Research, Pearl, and LDV Capital discuss trends of computer vision and machine learning impact on medical diagnostics to revolutionize precision medicine.

LDV Blog