Pittsburgh innovation history of medical AI
- John Kalafut
- Jun 8
- 5 min read
John F. Kalafut, PhD | Chief Technology & AI Officer, Asher Informatics | June 2026
SIIM's Back in Pittsburgh
Before I get into this, a quick note for folks (‘yinz’) in my network who aren't in the imaging informatics world. SIIM is the Society for Imaging Informatics in Medicine, the professional community of clinicians, researchers, engineers, and informaticists working across the full spectrum of medical imaging, from foundational information science and clinical validation research to the practical deployment and governance of the technology that makes it all run. If you work in health AI, clinical AI governance, or enterprise health IT, there's more overlap with your world than you might think.
And look, when I attended my first SCAR/SIIM years ago as a wide-eyed graduate student and early-career engineer, I definitely did not think I'd someday be writing a post reflecting on the history of the field. That was something those ‘old guys’ did. Well. Here I am. Still geeky. Just with more context and considerably less hair.
SIIM is in Pittsburgh this week, and I've been thinking about why that matters beyond local pride, the inevitable Primanti Brothers pilgrimage, and trying to find the corner of the North Side where Dr. Robby and the rest of the crew from The Pitt do their thing (Hint – tell Uber to find Allegheny General Hospital (AHN)). Because the history of this region is foundational to what we all do in imaging informatics and health AI. Not in a chest-puffing way but in a genuinely consequential, workman-like way that doesn't get talked about enough.
The medical AI origin story that lives here
You can't get through five minutes anywhere today without someone saying "AI." And Carnegie Mellon is always name-dropped. But do we actually know what happened there?
Allen Newell and Herbert Simon demonstrated their Logic Theorist at the 1956 Dartmouth workshop, one of the foundational moments for AI as an academic discipline. Newell eventually landed at CMU, and Simon, who would go on to win the Nobel Prize in Economics, was already there. These two, and numerous faculty, grad students etc. were architects of AI becoming a field. Topics that grew from these early efforts included symbolic reasoning, cognitive modeling, and how machines might replicate human problem-solving. Born, refined, and institutionalized in Pittsburgh.
And even back then, the tension existed. Symbolic and logic-driven approaches on one side, and the "connectionists," the somewhat dismissive label for people working on neural network ideas, on the other. Sound familiar? That argument is still going, just with a much bigger compute budget. And a lot more charlatans preaching “just build more compute!!”
CMU, being CMU, took a pragmatic engineering posture. Data-driven methods mattered if they worked. Robotics flourished there (and yes, "physical AI" is the cringe term du jour, but the work was real). And in 2006, CMU founded the world's first Machine Learning Department, formally establishing what had been building since 1997 through the Center for Automated Learning and Discovery. First. In the world. That's a Pittsburgh factoid worth knowing.
First. In the world. CMU's Machine Learning Department, founded in 2006, was the first of its kind anywhere.
PACS, Pittsburgh, and the people who built modern radiology informatics innovations
The first truly "filmless" radiology department is still debated among historians in the field and is a worthy bar conversation. University of Kansas, St. Mary's Hospital and/or Hammersmith in London, the Baltimore VA all have rightful ‘dibs’ and honestly, credit to all of them. But the era of enterprise-scale, web-based, clinically viable PACS? That story runs through Pittsburgh and through the work of Paul Chang and his team at Pitt and UPMC.
Working at UPMC in the 1990s, Chang was tasked with helping the organization go digital, and what came out of those efforts was Stentor. Co-inventing a lossless wavelet-based image distribution mechanism called dynamic transfer syntax, the technology enabled one of the first viable enterprise image distribution solutions, later commercialized as Stentor PACS and acquired by Philips Medical Systems. Server-side compression, web-based viewing, actual UX that clinicians could use. Revolutionary for its time.
For those working in imaging AI today, there's also a debt owed to the statistical methods work done here. Dr. David Gur at Pitt was a central figure in CAD validation methodology, ROC analysis, and multi-reader multi-case study design. The fingerprints of that work are still visible in FDA guidance documents for AI/ML device evaluation today and every 510k that gets cleared by FDA if you are a CADe or CADx device.
Pittsburgh's Digital pathology and the enterprise evolution
Pittsburgh has long been a gathering place for the pathology informatics community. Pittcon (not to be confused with Anthrocon – that’s something else BUT a community Pittsburgh also loves), the city's major annual laboratory science conference, has for years featured early diagnostic digitization work in its clinical and diagnostics tracks. And Pitt's own AI PathConnect symposium is now a dedicated annual forum for computational pathology and AI, right here in town.
The institutional weight behind all of it traces back to UPMC. In 2008, UPMC and GE Healthcare formed Omnyx, a joint venture aimed at bringing digital pathology into clinical use, tackling a diagnostic science that had relied on glass slides and microscopes for over 125 years. Hard work. Not all of it went as planned. But it moved the field forward.
Which brings me to something worth noticing about SIIM itself this week. SIIM started life in 1989 as SCAR, the Society for Computer Applications in Radiology, and changed its name in 2006 to better reflect the broadened role of imaging across the medical enterprise. That name change was a statement of intent. And the fact that SIIM and the Digital Pathology Association now offer a bridge pass for DPA members attending SIIM is a concrete signal that the "enterprise" framing wasn't just marketing. Pathology, dermatology, and other specialties aren't adjacent to this community anymore. They're in it.
Pittsburgh pioneers whose innovations made digital medicine work. For real.
It's worth stepping back for a second and giving credit where it's really due. The people in this community, clinicians, physicists, informaticists, and engineers, many of them shaped by or connected to institutions right here, were the ones who made digital medicine work in the real world. Not just conceptually. Operationally. They drove the adoption and governance of DICOM, the foundational standard that finally allowed imaging devices from different manufacturers to actually talk to each other. They transformed radiology into what is, by a wide margin, the most IT and software-enabled specialty in all of medicine. They built the workflows, the standards bodies, the validation frameworks, and the clinical evidence base that made any of this trustworthy at scale.
It's no surprise, then, that this same community is also ground zero for the serious, rigorous evolution of AI in medicine. Not the hype cycle version. The real one.
The ecosystem today builds on its history
Pittsburgh isn't resting on any of this history. The Pittsburgh Supercomputing Center, UPMC, CMU, and Pitt are collaborating on a Health Sciences and Medical Innovation Hub with real compute infrastructure behind it. UPMC Enterprises recently launched Ahavi, a virtual environment to evaluate and validate AI models against UPMC's patient population before clinical deployment. The Computational Pathology and AI Center at Pitt opened recently. The depth here is real and it keeps compounding. And.. of course, yours truly and “Asher Informatics” is growing our startup here (some visit us! Station #8 in the startup aisle).
So, welcome to the 'Burgh, yinz all. Grab an 'Ahrn' City, enjoy the conference, and maybe take a second to appreciate that you're standing on some genuinely important ground.
See yinz soon.
John F. Kalafut, PhD Chief Technology & AI Officer Asher Informatics | Health AI. Governed.
Comments