Physicians searching body tissue for cancer is like analysts' looking at satellite images for missiles, requiring similar technology.
An Israeli startup is looking to use computer vision and artificial intelligence to help pathologists process and analyze biopsies for cancer more accurately and quickly.
Today, most diagnoses of diseases including cancer are performed by pathologists, who examine tissue taken during a biopsy and determine whether it is normal, benign or malignant. This painstaking process, using slides and a microscope, has remained stagnant over the past century, requiring years of training. Even a small cluster of unnoticed cancer cells can lead to a possibly fatal misdiagnosis.
Now, a Tel Aviv based startup is developing an AI pathology system to help pathologists in the job.
“My father was suspected to have prostate cancer and it was a very stressful month till he received negative biopsy results,” said Avi, the firm’s co-founder and CEO and a graduate of an elite technological unit of the Israeli Military Intelligence corps specializing in computer vision.
“Pathologists examine the biopsies and look for malignant cells based on their abnormal shape and size, similar to the way image analysts interpret a satellite image looking for missiles or other concealed objects,” Avi said. “Both processes require time and experienced professionals. Automating the pathologist’s and the image analyst’s tasks are similar in nature.”
There is a growing need for pathologists’ diagnostic skills as the prevalence of malignant disease increases, said Avi. But the number of pathologists in the world is declining, he said. “In Israel, there are just some 100 pathologists in the whole of the country,” he added. “You need to bridge that gap with AI and computer vision.”
The company has developed software that uses computer vision, algorithms and artificial intelligence to scan biopsies. The software then processes the information — comparing it to thousands of other scans it has been fed previously — to identify problematic ones and flag them for the pathologists overseeing the process.
In a couple of months the technology, which is still under development, will be deployed in hospitals and research centers in Israel and the US, Avi said. The company on Monday announced it has raised some $5 million in financing from Vertex Ventures and Grove Ventures, and from private investors including Brian Cooper, the founder of software firm Retalix and entrepreneur Nir Kalkstein.
The artificial intelligence market for healthcare applications is expected to expand rapidly globally, with revenues growing at 40 percent annually through 2021 and reaching $6.7 billion by that year, a report by the Frost & Sullivan research and consulting firm says.
AI has the potential to “improve outcomes by 30 to 40 percent” while reducing the costs of treatment by as much as 50%, the report says. “Already playing a critical role in other industries, AI systems are poised to transform how we think about disease diagnosis and treatment,” said Frost & Sullivan transformational health industry analyst Harpreet Singh Buttar in a statement. “Augmenting the expertise of trained clinicians, AI systems will provide an added layer of decision support capable of helping mitigate oversights or errors in care administration.”
Israel is considered a global leader in AI, high-performance computing, and computer vision, as well as in healthcare. According to data compiled by the nonprofit Startup Nation Central, Israeli venture capital funding in digital health startups jumped by 30 percent last year compared with 2016 to almost $340 million, with a total of 470 digital health startups operating in seven subsectors, including AI.
The company was founded in 2017 by Avi and two friends who served with him in the IDF and have years of experience in developing advanced systems in the areas of computer vision and data science. “Our solutions must meet the strictest standards of accuracy due to the significance of any error in diagnosis,” said Avi, in a statement.
“Diagnostics performed by pathologists combine the analysis of visual information and textual clinical information, which creates a complex technological challenge.” To address this challenge, the company recruited a team of experienced researchers and partnered with leading hardware suppliers, laboratories, and hospitals to gain access to vast amounts of clinical information.
At the moment the company’s solutions address prostate, breast, and gastrointestinal-related diseases. In next few months, the firm plans to will double its team by recruiting additionally experienced machine learning researchers in order to speed up development.
“The company will dramatically improve the accuracy and efficiency of pathologists,” said Emanuel, general partner at Vertex, in a statement.
“The introduction of artificial intelligence into digital pathology will enable pathologists to cope with growing needs in the field of cancer diagnostics,” said Sigalit, a partner at Grove Ventures.