Traditionally Professional Courtesy is something that physicians gave each other - but we had to be careful with it when…
Healthcare AI News 7/9/25
News
Rock Health reports that US digital health startups raised $6.4 billion in the first half of 2025, which represents a modest increase over the same period last year. AI-enabled companies captured 62% of that total, raising average round sizes that were nearly double those of non-AI startups. Nine of the 11 funding rounds of greater than $100 million went to companies that offer AI-driven products, including two mega-rounds for Abridge within just four months.
An AI cybersecurity company says that its average health system audit uncovers 70 active AI applications, many of them embedded in tools from Microsoft, Salesforce, Google, and LinkedIn. It notes that while healthcare organizations often believe that they have limited AI use by blocking tools like ChatGPT and Gemini, they often overlook AI features that are contained in vendor-provided technology.
Cleveland Clinic Abu Dhabi appoints Peng Xiao, the CEO of Emirates-based AI development company G42, as its board chair.
Business
John Snow Labs spins off Martlet.ai, which will apply AI to HCC coding.
FDA grants Breakthrough Device Designation to Artera’s precision medicine tool for prostate cancer.
Research
UCLA researchers create an AI tool that turns structured EHR data into “pseudo-notes” that can be used by clinical decision support systems without EHR integration.
Mayo Clinic develops an AI tool that diagnoses surgical site infections by analyzing patient-taken photos of wounds after surgery.
University of South Florida researchers develop an AI system that assesses pain in NICU babies in real time by analyzing data from cameras and sensors.
Other
Huntsville Hospital (AL) upgrades its campus security system with 1,800 AI-powered cameras that employ facial recognition and license plate detection.
Contacts
Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Follow on X, Bluesky, and LinkedIn.
Sponsorship information.
Contact us.
It is great to hear about health systems building and deploying models. As with all software built by health systems in the past, this will prove to be unmaintainable in the long term. Nobody in the health system will know how many models have been built and deployed – none of them will be maintained. Ultimately, the winning strategy is to still work with a handful of trusted software vendors and use AI built and integrated with commercially released software.
Are the training models more like code or data? The answer to that could dictate how AI is deployed.
I can imagine that if the models are viewed as code-like, then AI systems could be versioned. However even then, if the hosting entity is a vendor accessed through the cloud, you might end up with a continuous integration (CI) model.
If the models are viewed as data-like, then AI systems might get high frequency or even perpetual training. The result is very similar to the CI model. Another way to think about this is analogous to an internet search engine.
Then I wonder, why limit ourselves here? If AI is the subject, why not model a Real Intelligence (RI)? Perhaps we should aim to emulate the hiring of an employee, who can be trained at any time?