In 2017, Vinod Khosla, the founder of Khosla Ventures, made a bold statement to CNBC, saying, “The job of the radiologist will be obsolete in five years.” Although he later revised his timeline to as long as 15 years, Khosla maintained his stance that artificial intelligence (AI) image recognition could soon diagnose diseases on scans better than human doctors.
However, seven years later, radiologists are still required to interpret most scans, despite the assistance of AI software. The more pressing challenge lies in the shortage of these doctors in the United States and globally.
While Khosla Ventures has backed several imaging startups, including Vista.ai and Q Bio, the firm’s latest bet is on a company that aims to ease radiologists’ workload by reducing the time spent on report documentation, rather than attempting to replace the physician with a machine.
On Tuesday, Khosla led a $50 million Series B funding round for Rad AI, a company that developed a tool capable of generating reports for radiologists. Other participants in the round included World Innovation Lab and returning investors ARTIS Ventures, OCV Partners, Kickstart Fund, and Gradient Ventures (Google’s AI-focused fund). This financing brought the company’s total capital raised to over $80 million.
Rad AI was founded in 2018 by Dr. Jeff Chang, who completed his medical training as a radiologist at the age of 16 and later received an MBA from UCLA, and serial entrepreneur Doktor Gurson. Recognizing that most of a radiologist’s time is spent documenting findings rather than analyzing images, the pair decided to develop a proprietary large language model (LLM) trained on radiology report datasets to automate doctors’ findings and impressions documentation.
While generative AI was not widely used by tech companies until OpenAI’s ChatGPT burst onto the scene in 2022, Rad AI prides itself on being an early adopter of this technology. “I’m confident we’re the first company in radiology to start using LLMs,” said Gurson, Rad AI’s CEO. “We started doing that work in 2018, around the same time that open AI was creating their [first] models.”
Six years later, Rad AI’s products are used by about a third of U.S. health systems and nine of the 10 largest radiology groups in the country, according to Gurson.
The fresh capital will be used to build a team that deploys Rad AI’s latest product: a standalone radiology reporting solution. “We have a lot of interest, but there’s only so much we can deploy at once,” Gurson said, adding that Rad AI is hiring people who can install and maintain the software.
While some incumbents have been trying to add generative AI functionality to their radiology reporting software over the past 18 months, Rad AI doesn’t consider these companies to be true competitors yet. “At this point, probably 99% to 100% of the market uses our products,” Gurson stated. “If it’s any indication, we’ve not lost a single customer since we started.”
