Prostate gold seed segmentation in the browser
SeedSeg is a browser-based tool for automatic segmentation of gold fiducial markers in T1-weighted prostate MRI.
It uses an nnU-Net (residual encoder) model, trained with 5-fold cross-validation, to detect the gold seed markers used in prostate radiotherapy treatment planning. Each marker is reported with its mean probability; cases with fewer candidates than the implant count are flagged for review.
All computation runs locally in your browser using ONNX Runtime Web. No data is uploaded to any server.
SeedSeg runs entirely in your web browser using WebAssembly and ONNX Runtime. All processing occurs locally on your machine.
SeedSeg uses ONNX Runtime Web to run deep learning models directly in the browser via WebGPU or WebAssembly. Your NIfTI or DICOM files are read into browser memory, processed locally, and results are displayed without any network transfer.
The website downloads NiiVue from unpkg, pinned model weights from Hugging Face and other runtime assets, and counts page views using shared site analytics. The optional dicompare report uses bundled controller, worker and schema assets. When requested on the website, it downloads Pyodide and Python packages from jsDelivr and PyPI. These requests contain no input images or DICOM metadata.