ilex-hub/brats_segresnet.1

BraTS SegResNet (multimodal brain-tumour sub-region seg) -- BraTS SegResNet (MONAI brats_mri_segmentation v0.5.4)

Description

MONAI brats_mri_segmentation (SegResNet; Myronenko 2018), ported to JAX / Equinox from the upstream PyTorch bundle. A ResNet-style 3D encoder/decoder (pre-activation GroupNorm+ReLU+Conv residual blocks, strided-conv downsampling, additive skips, non-trainable trilinear upsampling) that segments the three BraTS tumour sub-regions from a 4-channel multimodal MRI volume. The VAE branch used for autoencoder-regularised training is out of scope; this port is the segmentation forward only.

Intended use

Brain-tumour sub-region segmentation from 4-channel multimodal MRI (channel order T1c, T1, T2, FLAIR), per-channel nonzero z-score normalised, each spatial dim a multiple of 8. Returns 3 raw-logit channels for the overlapping sub-regions tumour core (TC), whole tumour (WT), and enhancing tumour (ET); apply a per-channel sigmoid and threshold at 0.5 (multi-label, not softmax). The v0 bundle is the network forward only; the bundle's intensity normalisation and sliding-window inference are not vendored.

Usage

from ilex.models.brats_segresnet import BraTSSegResNet
model = BraTSSegResNet.from_pretrained('ilex-hub/brats_segresnet.1')

Authors

Myronenko A.; MONAI Consortium

Citation

Myronenko A. (2018). 3D MRI brain tumor segmentation using autoencoder regularization. BrainLes 2018 (MICCAI workshop). arXiv:1810.11654. Distributed as the MONAI Model Zoo brats_mri_segmentation bundle.

References

• Myronenko A. (2018). 3D MRI brain tumor segmentation using autoencoder regularization. BrainLes 2018 (MICCAI workshop). arXiv:1810.11654. https://arxiv.org/abs/1810.11654
• MONAI Model Zoo: brats_mri_segmentation. https://github.com/Project-MONAI/model-zoo/tree/dev/models/brats_mri_segmentation
• Bundle weights: https://huggingface.co/MONAI/brats_mri_segmentation

License

HF Hub license tag: apache-2.0

Effective terms: Apache-2.0 (MONAI Consortium) on both the SegResNet network code (monai.networks.nets.SegResNet) and the brats_mri_segmentation bundle weights. No commercial restrictions; no gating required. The underlying BraTS training data has its own challenge terms, but the released weights are Apache-2.0. The ilex JAX / Equinox port code is separately licensed under Apache-2.0 / GPL-3.0.

Upstream license reference: https://github.com/Project-MONAI/model-zoo/blob/dev/models/brats_mri_segmentation/LICENSE

Copyright

Network architecture and pretrained weights: copyright (c) MONAI Consortium, released under the Apache-2.0 License. JAX / Equinox port: copyright (c) the ilex authors, released under the Apache-2.0 / GPL-3.0 dual license used by ilex itself.

Upstream source

Original weights / reference implementation: https://github.com/Project-MONAI/model-zoo/tree/dev/models/brats_mri_segmentation

Provenance

This artefact was produced by ilex's save/load pipeline. The architecture is implemented in ilex.models.brats_segresnet.BraTSSegResNet and the weights have been converted from their upstream format. See the upstream source above for the canonical reference.