Microscope Conflux CxTissueSeg

Conflux CxTissueSeg ONNX inference with the pinned 40 µm/px normalization, tiling, blending, thresholding and lazy checksum-verified local weights.

This is an ordinary microscope-node.toml extension. The companion loads it through the public node-pack discovery and package-isolated execution path; no official-only runtime is used. Node IDs are namespaced under microscope_conflux. Historical microscope.* IDs remain supported by schema-declared legacy_ids when this pack is installed.

Install locally

Review the source, install the scientific dependencies into a package-specific interpreter, then register the local checkout explicitly:

python3 -m venv .venv
.venv/bin/python -m pip install -e .
microscope nodes register /absolute/path/to/microscope-nodes-conflux --python /absolute/path/to/microscope-nodes-conflux/.venv/bin/python --trust

This installs NumPy and CPU ONNX Runtime from this pack's declared dependencies. Use a user-managed environment if you need a CUDA-specific ONNX Runtime distribution. Registration is explicit local trust. The SDK injects the framework into the isolated node process; installed package environments do not share Python site-packages.

Development and tests

python -m pip install -e .
python -m unittest discover -s tests -v

No environment or model weights are committed. Model weights must be obtained by the user from the official source, stored locally, and pass the pinned SHA-256 check.

Attribution

This package's implementation is Microscope.io source code. It depends on NumPy (BSD-3-Clause) and OpenCV (Apache-2.0 Python distribution; OpenCV project components retain their upstream BSD notices). See NumPy licenses and OpenCV licenses. Model-specific attribution and source licenses are documented in the respective model README.

Model: Conflux CxTissueSeg, Apache-2.0, revision 14d60366074bbd02a6adace63718cd4bf16da116; SHA-256 3a30565a64ebcee8fbac11eb794a534283965236d59961d08351329ed6061893. Model weights are not included.

Use the node

The stable node type is microscope_conflux.cxtissueseg. It accepts an RGB image and returns both a binary mask and a probability raster aligned to that image. It is an overview node with a required working spacing of 40 µm/px on both axes. Connect WSI Region in overview scope to the node and set the source working MPP to 40. The node normalizes RGB by 255, runs overlapping 512-pixel patches at 448-pixel stride, blends probabilities with a Gaussian weight map, then thresholds the result.

Parameters are weights (local file path, private/non-portable), threshold (default 0.5; range 0.01–0.99), and device (cpu or cuda). The pinned ONNX model file is not bundled: obtain the official artifact from the linked source at the exact revision above, then set its local path. The node checks the SHA-256 before inference. For CPU use, the manifest installs onnxruntime; CUDA requires a compatible user-managed onnxruntime-gpu/CUDA environment and a CUDA provider. The selector cannot install GPU drivers or replace your environment.

In Workflow, register this pack, start/pair the companion, add WSI Region and Conflux CxTissueSeg, choose overview at 40 µm/px, set the local weights path, then run. Add either output to Layers: the binary mask for a thresholded view or the probability for a continuous score map. A 40 µm/px overview is coarse; it is not a full-resolution tissue boundary and should not be used to claim pixel-level localization.

If weights are absent or their digest differs, inference stops with an explicit error. If MPP is absent or not within 1% of 40 on both axes, the node refuses execution; the viewer does not guess it. If CUDA is selected but unavailable, choose CPU or install a matching CUDA stack. The optional ONNX model/environment are not part of the dependency-free browser viewer.

Conflux output is research model output, not a clinically validated result. Review masks/probabilities and downstream use; this package makes no diagnostic accuracy claim. Use the generated node schema for the exact schema. Package import tests do not run inference or require the large model file:

python -m pip install -e .
python -m unittest discover -s tests -v