sm-ai-team/simplemind-test-sets

SimpleMind test sets

Small test sets that the published SimpleMind workflows are stamped on. A workflow's example cases are the cases it was validated on, so there is no separate demo set. Each set is a CSV and a folder of case files with paths relative to the checkout's data/:

python -m simplemind.kg assets fetch <workflow>

Downloading needs no account. Each set has a manifest in SimpleMind's data_manifests/examples/ with the sha256 of the CSV and of every case file.

Licences

Set	Used by	Cases	Licence
cxr_portable_examples	xray_chest_binary_segmentation	5 chest X-rays	None stated by NIH; citation requested
kidney_ct_images_5	ct_abdomen_left_right_kidney_masks, ..._functional_kidney_masks, ..._functional_kidney_volume_measurements	5 abdominal CTs with labels	CC BY-NC-SA 4.0

cxr_portable_examples

Five portable (AP) frontal chest X-rays from the NIH ChestX-ray14 dataset (NIH Clinical Center): image indices 00000005_001, 00000050_000, 00000087_000, 00000133_002 and 00000179_000, all labelled "No Finding". cxr_portable_examples/provenance.csv lists age, sex, view and labels.

• The images are upsampled. NIH ships 1024 x 1024 pixel PNGs; here they are resampled (linear) to the original matrix size and NIH's pixel spacing, so the pixel size is larger than the NIH source. They add no resolution.

• They are illustrations, not a performance sample. The five were chosen from 40 films because the lung and trachea workflow segments them well.

• NIH states no licence or restrictions with the dataset; it asks only that its paper is cited:

  Xiaosong Wang, Yifan Peng, Le Lu, Zhiyong Lu, Mohammadhadi Bagheri, Ronald M. Summers. ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases. IEEE CVPR, pp. 3462-3471, 2017.

  Source: NIH Clinical Center, ChestX-ray14 (https://nihcc.app.box.com/v/ChestXray-NIHCC). We acknowledge the NIH Clinical Center.

kidney_ct_images_5

Five cases (case_00498, case_00258, case_00435, case_00259, case_00448) of the KiTS23 challenge data, each an image (image.nii.gz) and its segmentation (seg.nii.gz), unchanged from the official release. CC BY-NC-SA 4.0: non-commercial, with attribution, share-alike.

Attribution: Heller N, Isensee F, Tejpaul R, Wood A, Papanikolopoulos N, Weight C. The 2023 Kidney and Kidney Tumor Segmentation Challenge, MICCAI 2023. doi:10.5281/zenodo.7840134 (kits-challenge.org; github.com/neheller/kits23).

This is not legal advice.

Reference outputs

outputs/<workflow>/ holds the stamped run of each runnable published workflow: the outputs and recorded target results for every case, with a bundle.yaml naming the workflow, its subgraph_hash and the KG version. Compare a run of your own with

python -m simplemind.kg compare-runs <your run> --reference <workflow>

identical or equivalent means your installation reproduces the stamp. These outputs derive from the cases above and carry their licence.

Research use only

Not for clinical use.