Hi everyone – I’ve been experimenting with browser-based medical volume rendering.
It’s a volume path tracer that runs entirely in the browser. Under the hood it’s all C++ compiled to WebAssembly – Diligent Core’s WebGPU backend does the GPU rendering, Dear ImGui is the interface, and GDCM loads the DICOM data (experimental for now).
Demo: MedicalViewer
Chrome or Edge only – it relies on WebGPU features other browsers don’t fully support yet.
Nice work. Did you see @pieper’s WebGPU examples: GitHub - pieper/SlicerWGPU: Use wgpu, a rust-python WebGPU implementation, in 3D Slicer · GitHub
Thanks! I haven’t seen those yet. On native platforms, I use the native graphics APIs through Diligent Core. In my project, WebGPU is used only for the Web version. WebGPU has quite a few drawbacks – from small but annoying limitations, like the lack of hardware trilinear filtering for fixed-point formats (R16_Unorm, etc), to more fundamental issues, such as the specification not really addressing behavior in a multithreaded environment