| { |
| "_ilex": { |
| "architecture": "ilex.models.fsm_seg.model.FSMSeg", |
| "constructor_kwargs": { |
| "feat_mult": 2, |
| "nb_conv_per_level": 2, |
| "nb_features": 24, |
| "nb_labels": 11, |
| "nb_levels": 3, |
| "pool_size": 2, |
| "src_feats": 1 |
| }, |
| "format": "ilex", |
| "framework_version": { |
| "equinox": "0.13.8", |
| "ilex": "0.0.0.dev0", |
| "jax": "0.10.0", |
| "jaxlib": "0.10.0", |
| "numpy": "2.4.4", |
| "safetensors": "0.7.0" |
| }, |
| "has_state": true, |
| "origin": "ilex-native" |
| }, |
| "authors": "Greve D. N., Billot B., Cordero-Grande L., Van Leemput K., Fischl B., Iglesias J. E.; with substantial overlap across variants -- per-variant first authors are listed under references.", |
| "copyright": "Network architectures, training code, and pretrained weights: copyright (c) the respective upstream authors, distributed as part of the FreeSurfer software bundle under the FreeSurfer Software License (FSLA; permissive academic / non-commercial research offering). See https://surfer.nmr.mgh.harvard.edu/fswiki/FreeSurferSoftwareLicense for the binding terms. JAX / Equinox port code: copyright (c) the ilex authors, released under the Apache-2.0 / GPL-3.0 dual license used by ilex itself.", |
| "data_type": "nibabel", |
| "description": "A small 3-level neurite UNet shared by five FreeSurfer segmentation tools, ported to JAX / Equinox from the FreeSurfer bundled reference implementations. The architecture is a compact variant of the SynthSeg / SynthSR UNet body (nb_features=24, nb_levels=3, conv_size=3, feat_mult=2, nb_conv_per_level=2, per-level BatchNorm, ELU activation, 1-channel T1 input, softmax head). Five published variants share this architecture exactly; they differ only in the number of output labels and in the recommended inference field of view. The variant id is recorded in the bundle's card, so loading by repo id selects the appropriate weights without changing the JAX class.", |
| "equinox_version": "0.13.8", |
| "ilex_version": "0.0.0.dev0", |
| "image_classes": "Single-channel T1-weighted MRI volume conformed to the variant's recommended FoV (160 mm cube, 144 mm cube, or 72 mm cube).", |
| "intended_use": "Small-FOV brain structure segmentation from a single T1-weighted volume. Inputs are conformed to the recommended FoV per variant (sclimbic / hypothalamic_subunits / entowm at 160 mm cube, vsinus at 144 mm cube, mca-dura at 72 mm cube) and intensity- normalised per the upstream preprocessing recipe (mri_sclimbic_seg's normalisation path -- min/max + percentile clipping). The output is a per-voxel softmax over the variant- specific label set.", |
| "jax_version": "0.10.0", |
| "network_data_format": { |
| "inputs": {}, |
| "outputs": {} |
| }, |
| "numpy_version": "2.4.4", |
| "pred_classes": "Per-voxel softmax over nb_labels classes -- the variant-specific label set (sclimbic 14; hypothalamic_subunits 11; vsinus 6; entowm 5; mca-dura 2). The corresponding label IDs and names are recorded in the upstream .ctab colour table that FreeSurfer distributes alongside each model.", |
| "references": [ |
| "Greve D. N., Billot B., Cordero-Grande L., et al. (2021). A deep learning toolbox for automatic segmentation of subcortical limbic structures from MRI images. NeuroImage, 244:118610. doi:10.1016/j.neuroimage.2021.118610. (sclimbic variant.)", |
| "Billot B., Bocchetta M., Todd E., Dalca A. V., Rohrer J. D., Iglesias J. E. (2020). Automated segmentation of the hypothalamus and associated subunits in brain MRI. NeuroImage, 223:117287. doi:10.1016/j.neuroimage.2020.117287. (hypothalamic_subunits variant.)", |
| "mri_entowm_seg, mri_mcadura_seg, mri_vsinus_seg are FreeSurfer utility variants of the sclimbic architecture; the underlying network is the same and the FreeSurfer release notes document their introduction without separate primary publications." |
| ], |
| "schema": "https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/meta_schema_20240725.json", |
| "task": "Small-FOV brain structure segmentation (FSM family)", |
| "version": "0.0.0" |
| } |
