close
close

Medtech startup uses Oracle AI for cancer drug research

In recent years, human ingenuity and scientific know-how have enabled the development of advanced medicines that promise relief and even cure for millions of people around the world who have been diagnosed with and are now living with previously incurable forms of cancer.

Historically, identifying the active ingredient for potential drugs is a simple process. Drug discovery relied on ancient knowledge from traditional folk medicine—drugs like aspirin came about this way, or through happy accidents, as in the case of penicillin, the first identified antibiotic, which was discovered by chance on a moldy petri dish in 1928.

A century after Alexander Fleming's chance encounter with Penicillium chrysogenum, modern drug discovery is big business: a large and ever-growing pharmaceutical industry pumps billions of dollars into the process, often with government support – as in the case of the first Covid-19 vaccines. But the process is still difficult and riddled with inefficiencies.

What if there was a better way? We may be on the verge of major changes with technology being developed and deployed at Imagene AI, a startup based in Israel. Imagene and its founder and CEO Dean Bitan hope to jumpstart the drug discovery process and find new ways to help oncologists deliver the best possible care by using Oracle artificial intelligence (AI) to fight the scourge of cancer.

Bitan is a lifelong technologist who graduated with a degree in computer science at the tender age of 15. However, it was not until a few years later that he became interested in the field of cancer research. Cancer is a disease that touches the lives of almost everyone on Earth in some way during their lifetime, and in Bitan's case, he has unfortunately lost a close relative to the disease.

Speaking to Computer Weekly on the sidelines of Oracle Cloud World in Las Vegas, Bitan said: “I was interested in technology and entrepreneurship, and that was my area of ​​expertise. Initially, I had no connection to cancer.

“But when it happened, I learned a lot about the disease, and I learned about the gaps and what we could do better… There aren't always enough options in terms of drugs – we see a lot of cancers where we can't offer patients enough drugs.

“So that's the story of Imagene. I decided to get into this field and I really wanted to help doctors make better diagnostic and treatment decisions,” says Bitan.

Is AI the answer?

Did Bitan always have a hunch that AI could offer a way forward? He says he considered this possibility very early on in the research that led to Imagene.

“The assumption was that we could potentially use the technology in this area. [But] “I'm an engineer. And when you talk to doctors, you see a different attitude,” says Bitan.

“We sat down and thought: We know the challenge and we know how we can do it better. Before we founded the company, some doctors told me that they have a strong intuition when looking at a biopsy image.

“They say they can identify patterns that are likely related to the presence of biomarkers that indicate whether or not a patient will respond to a particular drug. They build that up over many years of practice and observing their patients, but that's something that can be improved… Intuition and AI go together very well. So we knew we would try to see if we could get more information from these biopsy images.”

Bitan and his team used 630,000 anonymized biopsy images of different types of cancer at different locations in the body and developed a basic model to provide what he now calls “oncology intelligence.”

This 1.1 billion parameter base model is called CanvOI. At its core, it captures the complex features and patterns in biopsy images that a human unaided might never see, to improve a researcher's understanding of various pathological features and derive new insights.

The idea is to provide a “robust data backbone” for developing downstream applications in cancer research. This not only applies to identifying new drugs, but can also predict how people with different biomarkers might respond to them. This will ultimately enable frontline doctors to offer tailored cancer treatment based on their patients' individual physiology.

OCI supports CanvOI

Imagene's model runs on Oracle Cloud Infrastructure (OCI) and takes advantage of the OCI AI Infrastructure and OCI Supercluster, which can currently scale to tens of thousands of GPUs for AI inference and may exceed 130,000 in the near future.

Bitan says that by applying Oracle's computational capabilities and Imagene's new approach to digital pathology primitives, CanvOI is already achieving industry-leading performance on its various tasks, even when using minimally labeled data.

“Oracle is in a unique position to help AI companies with these challenges. So when we talk about compute power, the fact that they had this strategic agreement with Nvidia allows us to have higher availability of GPUs. And higher availability of GPUs means more compute power and that's better for us,” he says.

“We need companies like Oracle to join us on this long journey because the challenges are real and we want to keep going and show more milestones in this context. With ChatGPT and LLMs, we saw the technology go from elementary to high school level in less than two years and now you're talking about experts at PhD level. We want to see similar things in the world of oncology and we've done this with biopsy images, but over time we'll add more models.”

More broadly, CanvOI forms the cornerstone of the company's new OISuite, a platform designed to support researchers and developers of diagnostic systems, allowing them to find answers to a wide range of questions required for their research. Bitan says this reduces the need for AI expertise and data collection and enables new breakthroughs, while maintaining the highest possible standards of data privacy and security. With that in mind, Bitan says, all data used by Imagene's systems is anonymized in advance.

“And of course,” he continues, “we work based on the highest standards of GDPR, HIPAA compliance, etc. That much is clear, but in addition to that, we also work on zero-trust approaches, conduct vulnerability scanning, and encrypt data at rest and in transit, even though it is anonymized.”

Future goals

Imagene already works with medical institutions in several countries, including world-renowned US research facilities such as Johns Hopkins in Baltimore and Northwestern University in Chicago, as well as leading cancer centers in Brazil and Israel. Bitan wants to go even further and include not only academic medical centers, but also private and reference laboratories.

“In the field of cancer research, we do not have the privilege of not doing what we can, because every day counts,” he says.

Looking to the future, Bitan sees opportunities to apply the technology developed at Imagene to other areas of medical research, such as Covid-19 or HIV/AIDS.

“We will move closer to these areas over time. There will be more and more models that combine different modalities – so not just biopsy images, we could perhaps add radiology, MRI and X-rays. Or microbiomes or maybe even genome sequencing. Then we will be able to answer much more complex questions about different aspects of healthcare,” he concludes hopefully.

You may also like...