This Israeli Startup Is Using AI To Revolutionize Pathology

Interview
Share on linkedin
Share
Share on facebook
Share
Share on twitter
Tweet

Today, pathology based analysis is the gold standard for diagnosing a wide array of complex diseases, like cancerous tumors, and the variations accordingly. While it is estimated that 70% of all healthcare decisions affecting diagnosis or treatment involve some pathology based analysis, digital decision support systems have yet to receive formal FDA and CE approval. This is an opportunity and reformation Israeli AI startup DeePathology.ai anticipates to disrupt by enhancing and augmenting the pathological evaluation process.

“We see the benefit in saving Pathologists time in their workflow and enhancing their performance, especially as the supply for the profession is in a shortage today, and more so in the future” explained co-CEO and co-founder Chen Sagiv. “Pathologists can benefit from information presented to them in better way, with the support for decision making from AI.”

The Helicobacter Pylori (H.Pylori), a germ that’s known to cause inflammation, gastritis, and ulcers, is lurking in the stomachs of a staggering 50% of the global population. When people go through gastric biopsies, part of the protocol is to look for the H.Pylori germ because it can lead to cancerous outcomes and diseases in the upper intestinal tract if left untreated – otherwise, it’s eradicated with antibiotic treatments. Unfortunately for Pathologists, the identification process is manual: a Pathologist scans through magnified digital gastric slides with Haemotoxylin and Eosin (H&E) and Giemsa stains to identify the germ. DeePathology.ai is setting out alleviate the Pathologist’s workload albeit with the highest likelihood of widespread hospital adoption. Such a goal requires the unique expertise of AI research and hardware optimization.

Chen Sagiv, a PhD in Applied Mathematics, began her career as an algorithms developer specializing in image processing, and co-founding computer vision technology and projects company SagivTech. Nizan Sagiv, the business savvy counterpart, is a former executive in high-tech companies.

SagivTech in itself is major success. The company’s know-how has already been tapped by industry leaders, like Google’s ATAP group – Advanced Technology and Products – for their  ‘Project Tango‘, a computer-vision technology that allows mobile devices to detect their position relative to the world around them, without requiring GPS. SagivTech was a collaborator in the project. SagivTech also co-founded together with Dr. Koby Cohen the largest technological event in Israel for machine vision and deep learning, IMVC, that will celebrate on March 18th its 10th anniversary.

In 2017, SagivTech began to collaborate with Roche, a pharmaceutical and diagnostics leader with headquarters in Switzerland to develop deep learning based solutions for Pathology. Their work with Eldad Klaiman from the Roche Pharmaceutical Research and Early Development (pRED) group and others was published in two recent papers (Generalizing multistain immunohistochemistry tissue segmentation using one-shot color deconvolution deep neural networks; and Virtualization of tissue staining in digital pathology using an unsupervised deep learning approach). Based on the success of the collaboration, Roche had invited SagivTech to participate in the first batch of the Digital Health Accelerator of Roche in Munich, Germany.

Inspired by the participation in the accelerator, the Sagivs partnered with Jacob Gildenblat to form DeePathology.ai. Gildenblat, CTO and co-founder of DeePathology.ai, began his career as a software engineer at the Israeli Air Force, and later computer vision algorithm development for several startups until leading deep learning development at SagivTech.

“Relying on our expertise in deep learning and computer vision, we are confident we can bring value to the pathology market for a variety of diseases and ailments” explained Sagiv.

The vision of DeePathology.ai is to develop algorithms to handle several problems in digital Pathology diagnostics and for pharmaceutical research. For medical diagnostics, they have already released their baseline product of the automated H.Pylori detection software for gastric biopsies (available for a 90-day free trial).  It was the synergistic teamwork with Sebastian Klein, M.D. and Prof. Reinhard Buttner, M.D. from the Institute for Pathology Cologne, Germany that led to the successful development of the first ever automated H.Pylori detector software on gastric biopsies.

“We are using deep learning algorithms to support the Pathologist” explained Jacob. “In the case of the H.Pylori detection, our algorithm scans the slides and marks each region with a score of presence of H.Pylori. Then the highest scoring regions can be reviewed by a Pathologist and can be quickly confirmed. For difficult and ambiguous cases, slides can be sent to further expensive tests like IHC staining or PCR. Using our algorithm execute this selection makes it very economical since the hospitals don’t have to send all the slides to these expensive tests.”

The plan is to integrate the H.Pylori detector into slide management systems at hospitals, albeit without forcing the hospitals to acquire the necessary hardware to execute the algorithm intensive commands. Combining the rich expertise of the DeePathology.ai team in deep learning algorithm development and optimization they are able to deliver AI solution that’s optimized enough to run on a laptop of any Pathologist.

DeePathology.ai also plans to corner the pharmaceutical market with their technology. “The process of creating a drug takes average of 10 years and $2 billion. Anything you can do to reduce the cost associated with drug development or the time to market is meaningful” explained Chen.  

The young startup has validated their first market offering at length with the help of industry experts in the research and commercial domains. They’ve already submitted two provisional patents related to Pathology diagnostics and hinted at another breakthrough AI development for pharmaceutical development.

They’re currently bootstrapped and plan to get into strategic collaborations and funding in order to extend their range of products and offerings passed diagnostics and pharmaceutical. Their free trial offering has been downloaded hundreds of times, and they’re in discussions with medical groups to integrate their software into the larger flow of Pathologists. With incredible momentum behind them, it’s safe to say Pathology, the categorical incumbent practice for healthcare diagnoses and treatments, is scheduled to get a realistic AI enhancement.

  1. https://arxiv.org/pdf/1805.06958.pdf
  2.  https://arxiv.org/pdf/1810.06415.pdf

scroll down to read more