CEO SUMMARY: Australian artificial intelligence (AI) company Harrison.ai got AU$129 million from multiple investors, including both Sonic Healthcare—Australia’s largest pathology group—and I-MED Radiology Network. Pathology’s growing interest in AI tools, along with the growing relationship between radiology and pathology business interests, indicate more opportunities in integrated diagnostics that use AI.
Technological advancements in diagnostics—and a blooming symbiotic relationship— have moved the professions of anatomic pathology and radiology into closer partnerships in recent years, a business shift pathologists should be keen to follow.
The latest sign that pathology and radiology are poised for integration is the recent investment in a little-known artificial intelligence (AI) company in Sydney, Australia. On Dec. 1, 2021, Harrison.ai announced that it had raised AU$129 million (USD$97 million) from multiple sources, including from Australia’s largest clinical pathology company, Sonic Healthcare.
“These companies could create an imaging powerhouse using AI,” said Ajit Singh, PhD, Partner at Artiman Ventures, a venture capital fund in Palo Alto, Calif., and former CEO at BioImagene, an oncology diagnostics startup that Roche acquired.
While the deal itself is not a blockbuster in scope, it marks the second time in five months an AI company has been involved in a partnership with a clinical laboratory, commented Singh, who is also Adjunct Professor at Stanford Medicine’s Rad/Molecular Imaging Program and serves as an adviser to Ibex Medical Analytics in Tel Aviv, which develops AI-based cancer detection software for pathology. In July, PathAI, a technology company in Boston that develops image analysis software, acquired Poplar Healthcare Management of Memphis, an anatomic pathology group with a large national base of clients. (See TDR, Sept. 7, 2021.)
AI Tools for Diagnostic Use
The new funds mean Harrison.ai has raised AU$158 million (USD$118 million) over the past two years, which the company is using to develop AI tools for clinical use.
The most significant part of the deal may be new equity investments from two Australian companies:
- Sonic Healthcare, which runs clinical and pathology labs and performs radiology imaging.
- I-MED Radiology Network, which handles the largest number of radiology exams in Australia.
Also funding the deal were venture capital companies Horizons Ventures, Blackbird Ventures, and Skip Capital.
The equity investments are important because Harrison.ai also announced a partnership with Sonic to co-develop new clinical tools in pathology that will use AI to improve the efficiency of pathology diagnostics.
Bridging Two Professions
In an interview with The Dark Report, Singh explained the significance of this announcement.
“For the past two decades, pathologists and radiologists have talked about the importance of integrating the two professions,” he said. “This funding announcement from Harrison.ai seems to have moved the two sides even closer to the altar.
“The implications of this deal are far and deep,” he added. “In Australia, three reasonably significant players in radiology and pathology are involved in this deal. Keep in mind that, as a country of about 26 million people, Australia is about the size of Canada, making it an important market in the English-speaking world.”
Harrison.ai straddles the radiology and pathology markets, although it specializes in diagnostic imaging. Singh said the company’s new investments from I-MED and Sonic are newsworthy as the pathology industry looks to the future.
Integration of Diagnostics
“Just imagine what could develop now that these three companies are working together,” Singh noted. “I-MED, the largest radiology company in Australia, is now working with Sonic, which is the largest pathology company in Australia and the third largest pathology company in the world after only Labcorp and Quest Diagnostics. I-MED and Sonic are now working with Harrison.ai to develop AI tools for use in both worlds.
“In one way or another, imaging involving radiology and pathology encompasses almost all of diagnostics,” he added. “I would estimate that when you combine all the imaging that’s done in radiology and anatomical pathology, you get about 70% or so of all diagnostics work worldwide. What’s left after imaging is routine diagnostics, which makes up the remaining 30% or so. That’s why I say that the strategic implications of this deal are significant.”
The equity investment that Sonic is making in Harrison.ai is important because AI companies working in diagnostics need patient cases so that they can demonstrate the potential value of applying AI tools to patient care, Singh explained.
“After all, the goal of AI in diagnostics is to improve the consistency and efficiency of pathologists and radiologists by supplementing what diagnosticians do,” he said. “Any company developing artificial intelligence tools for radiology or for pathology needs to find customers to buy that software. Therefore, companies developing AI tools for diagnostics need to have the backbone that comes in the form of a pathology network or a radiology network.”
From that perspective, the new Australian partnership solves challenges on multiple levels: Harrison.ai can now weave itself into a network with patients, while Sonic and I-MED gain greater access to AI technology.
“The pathology and radiology companies have recognized that if they don’t have AI, they will have a problem in the future because they need AI to increase efficiency and throughput,” Singh said. “And the AI software companies have recognized that without a big source of patients and a constant flow of images, they have a problem.”
The Dark Report reached out to Sonic Healthcare and Harrison.ai for comment. A spokesperson said Sonic had no further comment beyond the joint announcement about the deal that the parties issued on Dec. 1. Harrison.ai did not respond back.
Observers in the anatomic pathology world watching these and similar deals should reach two conclusions.
First, pathology and radiology groups may pack a one-two punch for future investors, a new status that is worth exploring.
Second, more AI vendors are seeking partnerships with pathology groups to meet their pathology business needs, which opens doors for opportunities and new services that pathology groups can offer their clients.
Market Transactions Show How AI Companies and Diagnostics Providers Need Each Other
IN JULY, A BOSTON TECHNOLOGY COMPANY called PathAI that develops artificial intelligence (AI) software for pathologists bought Poplar Healthcare, a clinical laboratory in Memphis, Tenn.
In that deal, PathAI acquired access to patients served by Poplar, a national pathology provider, as well as Poplar’s lab facilities, management team, and 350 employees, including 25 pathologists. PathAI also got Poplar’s library of 50,000 patient samples, dating back to the company’s founding in 1995. The move was significant because it was the first time an AI developer acquired a pathology lab.
AI in Digital Pathology
In more recent news from Australia, Harrison.ai is teaming up with Sonic Healthcare to pursue AI-related work, including in digital pathology. For Ajit Singh, PhD, the latter development dwarfs the deal involving PathAI and Poplar.
“The lab that the Boston company bought was a deal worth several million dollars,” Singh commented. “In the big picture, that deal is relatively small. The scale of the Sonic deal is orders of magnitude bigger, and it’s significant because it shows that both pathology and radiology companies have invested in AI.”
This melding demonstrates that AI has come of age, meaning that the technology is useful in helping pathologists and radiologists work more efficiently to improve patient care. Does that mean AI is as good at identifying tumors or any malignancies as human physicians?
“That’s a very good question,” Singh said. “AI is not going to replace what pathologists or radiologists do. But over time, AI has become more useful in clinical settings. However, it still has limitations.”
AI Solves Inconsistencies
One of the major limitations is AI doesn’t work well for all healthcare applications. “It’s good for those applications where the number of variables is small,” Singh said. “If you apply AI to breast cancer, for example, the number of free variables is very large. That’s a difficult problem for AI to solve.” (See TDR, “Artificial Intelligence Ready for First Use in Anatomic Pathology,” Apr. 12, 2021.)
“On the other hand, if you apply AI to prostate cancer or melanoma, the number of free variables is small,” he added. “Therefore, AI is much more effective. Similarly, in radiology, if you apply AI to brain imaging, the number of variables is much higher than if you apply AI to identifying fractures, polyps, or lung nodules.”
Also, despite common misconceptions, the purpose of AI is not to “beat” the best pathologists or radiologists. Instead, AI helps physicians make more consistent diagnoses.
“If I run the same slide five times, I should get the same results, I should not get five different results,” Singh said. “When pathologists or radiologists are inconsistent, that creates a lack of confidence even among the best clinicians. AI solves the consistency problem.”
Contact Ajit Singh, PhD, at 650-845-2020 or email@example.com.