At the University of Pennsylvania, engineers have quietly solved a problem that has frustrated radiologists for decades: medical image analysis that actually fits into a clinical schedule. Meet FireANTs, an open-source algorithm that combines artificial intelligence with mathematical geometry to compare complex medical images in minutes instead of weeks—speeding up the work 100 to 1,000 times over compared to previous methods.
The challenge sounds simple but has long been intractable. When a radiologist needs to compare two brain scans or CT images from the same patient, they're not just asking whether something looks different—they need to know exactly which tissue regions have changed, by how much, and whether those shifts signal something serious. This is dense correspondence matching, the computational equivalent of finding and measuring the precise differences between two nearly identical paintings. Subtle changes matter enormously. Unusually rapid brain shrinkage, for instance, can be an early sign of cognitive decline, but only if clinicians can spot and quantify it.
For more than a decade, Gee's lab at Penn had developed ANTs—Advanced Normalization Tools—an open-source software that became the gold standard for this work. Yet as medical imaging datasets exploded in size, ANTs, which predates the modern AI era, simply couldn't keep pace. When doctoral student Rohit Jena, mentored by assistant professor Pratik Chaudhari and radiologic science professor James C. Gee, began experimenting with the software, he spotted the bottleneck. "ANTs worked much better than existing methods," Jena recalls, "but ANTs simply wasn't built for the kind of massive datasets that have become standard in medical and biological imaging research today."
Rather than merely moving the old software onto faster computer hardware—a solution people kept requesting online—Jena fundamentally rethought the approach. FireANTs borrows optimization techniques from modern AI but solves the matching problem mathematically, determining how one image actually corresponds to another without relying primarily on pattern-matching guesses learned from training data. It's a hybrid: AI's speed married to geometry's precision.
The results speak loudly. Testing across more than 15,000 image pairs spanning multiple organ systems, different imaging modalities, and even various animal species, FireANTs not only matched ANTs' accuracy—it ran hundreds to thousands of times faster depending on the problem, with zero loss in precision. In some cases, analysis that previously consumed an entire week now completes in minutes.
This speed transforms the technology from a research curiosity into something genuinely useful in patient care. As Gee explains, "In radiology, a large fraction of reads involve follow-up imaging, to see what changes have occurred between scans. Image registration can help automatically pinpoint any differences, but if the processing takes too long, it simply doesn't fit into the clinical workflow. The speed makes a huge difference in making this practical for patient care."
The algorithm is open-source, meaning radiologists, researchers, and developers worldwide can access it freely. That combination—radical speed, maintained accuracy, and community-driven availability—positions FireANTs to ripple outward across medical imaging far beyond radiology. In hospitals and research labs already drowning in imaging data, faster analysis could mean subtler diagnoses, caught earlier, for more patients.