Imagene taps Oracle Cloud Infrastructure in fight against cancer

Dean Bitan, co-founder and CEO of Imagene AI, a pioneer in AI-based precision oncology, is a software engineer by training. But he was inspired to create this Tel Aviv-based startup partly as the result of personal trauma.

Typical cancer diagnoses can take weeks, leaving patients in a frightening limbo before learning of their treatment options and prognoses. And even then, this diagnostic odyssey can lead to imprecise treatments and negative outcomes. Bitan, who lived through this hellish uncertainty with his mother’s cancer, understands viscerally how much it can weigh on patients and their families.

“I decided to get into this field because of that experience,” he says. “Sometimes, when we only look at numbers like survival rates, we forget that there are other parameters as well, like quality of life. You think about a patient waiting one month for test results, not knowing whether they’re going to live or die. It’s crazy. And we don't measure that enough, I think, today in medicine.”

Rapid molecular profiling

Imagene AI, an Oracle Cloud Infrastructure (OCI) customer, is Bitan’s response to that state of affairs. The company’s technology uses AI to capture complex features and patterns within biopsy images. The technology is being validated and used by 20 healthcare facilities worldwide, including Johns Hopkins Hospital, Northwestern, and Hoag in the United States, as well as Oncoclínicas, the largest cancer treatment organization in Latin America, and two of the leading oncology centers in Israel.

Imagene’s first AI-based test, called LungOI, uses AI to help doctors identify which specific form of lung cancer a patient has, to better navigate diagnostic and treatment options by analyzing within minutes images of tissue samples, including samples too small for standard biopsies to accurately analyze. (The OI stands for oncology intelligence.) Bitan admits that when he first started the company, together with Jonathan Zalach and Shahar Porat, it felt more like the realm of science fiction.

Today, patients can receive a precise cancer diagnosis in hours rather than weeks, and they can get treatments for the specific mutation and their specific genotype, rather than having to submit to general, “one-size-fits-all” treatments, such as radiation or chemotherapy.

The US Centers for Medicare & Medicaid Services recently granted reimbursement codes and payment rates for LungOI, “another important milestone for us,” Bitan says.

An AI boost to research

Imagene AI recently unveiled CanvOI, a state-of-the-art Oncology Intelligence foundation model developed in collaboration with Oracle and the Ellison Institute of Technology. Imagene built its cloud-based OI Suite platform using the CanvOI model, thus enabling pharmaceutical and health research organizations to build AI models that respond to a wide range of oncology questions so that they can advance their research and support drug development and clinical trial processes.

Using Oracle’s OCI Supercluster and OCI AI infrastructure, CanvOI analyzes complex features and patterns within biopsy images of suspected cancers, discovers biomarkers, assesses disease prognosis, identifies pathological features, and provides predictions for treatment responses.

CanvOI is a foundation model of 1.1 billion parameters that has been pretrained on a diverse data set that includes more than 630,000 tissue images from over 100 healthcare provider sites, covering more than 40 major body organs and tissue types.

Researchers can use Imagene’s OI Suite to create specific AI-based applications—without needing data scientists or AI specialists—to get answers to detailed questions about specific forms of gene alterations, whether a patient will respond to a given treatment, and whether the disease will recur. 

The AI service also lets researchers use even extremely small data sets to identify candidates for clinical trials who are most likely to respond to an experimental drug, and which ones are more likely to develop resistance or toxicity to a given treatment. “There are many different things you can ask. It’s very versatile,” Bitan says.

CanvOI is intended to help pharma companies develop treatments that target specific biological mechanisms that caused the cancer, potentially producing better response rates with fewer side effects. And it helps tests come to a conclusion more rapidly, with less patient churn, helping bring more treatments to market. “It’s an evolution, a scaling up, of precision medicine,” Bitan says. “It’s more precise precision medicine, allowing you to do precision medicine at scale.”

“It’s an evolution, a scaling up, of precision medicine.”

'Democratize oncology intelligence'

According to Bitan, CanvOI is to cancer research what ChatGPT is to internet search. “We wanted to democratize oncology intelligence so more diagnostics, developers, and researchers, and even pharmaceutical companies, can create cutting-edge AI models for unmet needs in cancer without having the AI expertise or the data,” he says.

Imagene had been running CanvOI on several cloud services but switched to OCI because of Oracle’s greater access to high performing GPUs, as well as its expertise in managing data and its experience in health sciences. “This partnership with Oracle allows us to do more with our resources and the data,” Bitan says.

The AI platform runs on OCI using Kubernetes Engine and HeatWave MySQL, leveraging the Zero Trust security paradigm.

Imagene, with offices in Miami and Tel Aviv and a lab in Phoenix, has raised $21.5 million in private funding and has about 30 employees, including physicians, data scientists, biologists, and software engineers.

View more Oracle Connect articles