toxoplasma-from-hoechst-cpsam

Toxoplasma from Hoechst (cross-channel)

Segments Toxoplasma gondii parasitophorous vacuoles from the Hoechst (nuclear) channel alone — no parasite-specific stain required. A cross-channel model: it is given the nuclear image and predicts where the parasites are.

• Architecture: Cellpose-SAM (cpsam_v2)
• Model Zoo key: toxoplasma_from_hoechst_v1
• Checkpoint: toxoplasma_from_hoechst_pv
• Trained by: einarolafsson

Use it in spaCR

This model is distributed through the spaCR Model Zoo. spaCR is an open-source package for spatial phenotype analysis of CRISPR screens and microscopy images.

pip install spacr

Model Zoo (GUI)

Launch the GUI and open the Model Zoo:

spacr

Find Toxoplasma from Hoechst (cross-channel) in the model list and press Download. The Model Zoo verifies the checkpoint's SHA-256 after download, so a truncated or substituted file is rejected rather than silently used.

Model Zoo (Python)

from spacr import model_zoo

entry = next(e for e in model_zoo.catalogue() if e.key == "toxoplasma_from_hoechst_v1")
path  = model_zoo.install(entry, dest="~/spacr_models")
print(path)   # verified local checkpoint

Mask generation

Point spaCR's mask generation at the downloaded checkpoint:

from spacr.core import preprocess_generate_masks

settings = {
    "src": "/path/to/images",
    "pathogen": "cellpose",
    "pathogen_model": str(path),     # the checkpoint fetched above
    "pathogen_diameter": 12,
}
preprocess_generate_masks(settings)

In the GUI the same thing is under Make masks — choose the downloaded model in the Cellpose model field for the relevant object.

API: :func:spacr.core.preprocess_generate_masks, :func:spacr.spacr_cellpose.generate_masks_from_imgs

Performance

Scored on a well-grouped held-out split of 463 fields never seen in training (no well appears in both train and test), including fields with no parasites, so false positives are counted. That is a 252x improvement in F1 over stock.

Per host cell line (held-out):

Checkpoint scores: ep10 0.4962, ep20 0.4261, ep30 0.4342, ep40 0.3940, ep50 0.5689, ep60 0.5499, ep70 0.5609, ep80 0.5631, ep90 0.5593, ep100 0.5594.

Training curves

Loss is on a log scale. Train and validation tracking each other is the overfitting check: a validation curve that turns up while train keeps falling is the signature to look for.

Objects are reference (ground-truth) objects. Training-set object counts were not recorded at training time; the held-out counts come from the scoring bundle.

Predicting parasites from Hoechst is the harder cross-channel direction — stock Cellpose-SAM scores only 0.0023 here. Compare with Toxoplasma from Cell Mask, the same task driven from the host cell mask.

Training data

2567 training fields and 463 held-out fields, split by well (training/split_by_well.csv) so no well leaks across the split. Targets are PV-regenerated masks (masks_pv). Hosts: HFF, HeLa and THP1.

Trained for 100 epochs from stock cpsam_v2, AdamW, lr 1e-05, weight decay 0.1, batch 1.

Files in this repository

Limitations

• The held-out split is used for checkpoint selection, so it is validation data rather than a fully independent test set.
• Targets are automatic reference labels (PV-regenerated masks), not hand-drawn ground truth.
• Hoechst-driven prediction is harder than the cell-mask route; expect lower scores than the cell-mask model.
• Accuracy falls above IoU 0.8 — suited to counting, occupancy and area rather than precise morphometry.

Links

• spaCR on GitHub: https://github.com/EinarOlafsson/spacr
• Model Zoo API: spacr.model_zoo — catalogue(), install(), fetch(), verify()
• Mask generation API: spacr.core.preprocess_generate_masks
• Issues and questions: https://github.com/EinarOlafsson/spacr/issues