From scan to conversation
A patient handed their own CT sees a stack of grey rectangles.
Praescia explores how scan-derived 3D models can make that anatomy easier to explore and discuss. Open a research CT in the browser to try the interaction.
Open a real CT in your browserNo signup or personal scan needed. The demo downloads a research case to your browser. A return link brings you back here.
Research prototype. Not for diagnosis or treatment decisions.
The problem is understanding, not another image
A scan is detailed. Its meaning is not self-evident.
A stack of slices asks the viewer to reconstruct spatial relationships in their head. A 3D view offers another way to explore those relationships, alongside the original images.
I built Praescia to investigate that interaction. The question is whether exploring a scan can support a clearer conversation, not whether a convincing image can replace clinical interpretation.
Try the interaction on a research case
Look through the slices. Then explore in 3D.
- Open the demo. A CT case loads automatically; no file selection is required.
- Explore the slices and respond to the structure-identification prompt.
- Explore the revealed 3D view, or show all structures to see their spatial relationships.
The browser demo uses NiiVue. It is separate from the native iPad prototype, which processes scans on the device. Neither is a clinically validated product.
The demo stores interaction history in your browser and offers an optional log export. Loading the demo makes network requests for its code and research cases. Site hosting currently enables Cloudflare Web Analytics.
Demo data: TotalSegmentator CT, Wasserthal and colleagues (2023), licensed under CC BY 4.0. Cases are packaged for this research demonstration; each case displays its source. The research annotations have not been reviewed for teaching suitability in this product.
Separate implementations, one spatial question
The browser is the invitation. The iPad is a separate prototype.
The NiiVue demonstration makes the slice-to-3D interaction accessible through a link. The native iPad app explores on-device scan processing and an interactive RealityKit view. The browser demo does not run RealityKit or the native Metal renderer.
The engineering work includes keeping orientation, physical units, and image geometry consistent. A plausible-looking model is not enough to establish that its coordinates are correct.
The implementation behind the native work
Geometry, loading, and rendering stay separate.
Praescia is a Swift package for volumetric image geometry, image loading, and Metal rendering on Apple platforms. Its geometry layer carries the relationship between voxel indices and physical coordinates; the host app owns the presentation.
The package, the iPad application, and the browser demonstration are separate projects. Work on one is not evidence that a feature has been integrated or validated in the others.
A public engine repository link is not available here yet.