|
| 1 | +# What's new in 1.1 🎉🎉 |
| 2 | + |
| 3 | +- Digital pathology workflows |
| 4 | +- Experiment management for MONAI bundle |
| 5 | +- Auto3dSeg enhancements |
| 6 | +- New models in MONAI Model Zoo |
| 7 | +- State-of-the-art SurgToolLoc solution |
| 8 | + |
| 9 | +## Digital pathology workflows |
| 10 | + |
| 11 | + |
| 12 | + |
| 13 | +Hover-Net is a model for simultaneous segmentation and classification of nuclei in multi-tissue histology images (Graham et al. Medical Image Analysis, 2019). |
| 14 | +We have added support for this model in MONAI by implementing several new components, enhancing existing ones and providing pipelines and examples for training, validation and inference. |
| 15 | + |
| 16 | +Along with the modules release, new digital pathology analysis tutorials are made available: |
| 17 | + |
| 18 | +- [HoVerNet pipelines](https://github.com/Project-MONAI/tutorials/tree/main/pathology/hovernet) based on MONAI workflows for training, validation and inference |
| 19 | +- [HoVerNet tutorial](https://github.com/Project-MONAI/tutorials/blob/main/pathology/hovernet/hovernet_torch.ipynb) for training, validation and inference |
| 20 | +- NuClick (Interactive Annotation for Pathology) tutorials for [training](https://github.com/Project-MONAI/tutorials/blob/main/pathology/nuclick/nuclick_training_notebook.ipynb) |
| 21 | +and [inference](https://github.com/Project-MONAI/tutorials/blob/main/pathology/nuclick/nuclick_infer.ipynb) |
| 22 | +- Nuclei classification tutorials for [training](https://github.com/Project-MONAI/tutorials/blob/main/pathology/nuclick/nuclei_classification_training_notebook.ipynb) |
| 23 | +and [inference](https://github.com/Project-MONAI/tutorials/blob/main/pathology/nuclick/nuclei_classification_infer.ipynb) |
| 24 | + |
| 25 | +## Experiment management for MONAI bundle |
| 26 | + |
| 27 | + |
| 28 | + |
| 29 | +In this release, experiment management features are integrated with MONAI bundle. |
| 30 | +It provides essential APIs for managing the end-to-end model bundle lifecycle. |
| 31 | +Users can start tracking experiments by, for example, appending `--tracking "mlflow"` to the training or inference commands to enable the MLFlow-based management. |
| 32 | +By default, MLFlow will track the executed bundle config, model quality measurements, and source code versioning. |
| 33 | +For more details, please refer to the [tutorial](https://github.com/Project-MONAI/tutorials/blob/main/experiment_management/bundle_integrate_mlflow.ipynb). |
| 34 | + |
| 35 | +## Auto3dSeg enhancements |
| 36 | + |
| 37 | +Multiple improvements have been added in `Auto3DSeg` both in terms of |
| 38 | +usability and performance. |
| 39 | +- Multi-modality support is added and applied for |
| 40 | +automated segmentation of the HECKTOR22 challenge dataset, which includes input 3D |
| 41 | +CT and PET images of various resolutions and sizes. A tutorial example of |
| 42 | +running Auto3DSeg on the HECKTOR22 challenge dataset is available in MONAI |
| 43 | +Tutorials. The tutorial is based on [the HECKTOR22 challenge 1st place solution](https://arxiv.org/abs/2209.10809). |
| 44 | +- A new improved version of `Segresnet` Algo is now available in `AutoRunner`. |
| 45 | +In this version, data caching is more efficient and the preprocessing transforms are more flexible. |
| 46 | +The workflow progresses including the timings of steps are written to console output as well as a YAML file. |
| 47 | +- Automatic customization and optimization of the model training configuration |
| 48 | +can be achieved according to the GPU devices used. The feature |
| 49 | +focuses on determining parameters including batch size of model |
| 50 | +training and sliding-window inference, allocated devices for |
| 51 | +data in sliding-window inference. For more details about how to enable it, please see [the tutorials](https://github.com/Project-MONAI/tutorials/tree/main/auto3dseg). |
| 52 | + |
| 53 | +## New models in MONAI Model Zoo |
| 54 | + |
| 55 | +New pretrained models are being created and released [in the Model Zoo](https://monai.io/model-zoo.html). |
| 56 | +Notably, |
| 57 | + |
| 58 | +- The `mednist_reg` model demonstrates how to build image registration workflows in MONAI bundle |
| 59 | +format. The model uses a ResNet and spatial transformer for hand X-ray image registration based on |
| 60 | +[the registration_mednist tutorial](https://github.com/Project-MONAI/tutorials/blob/main/2d_registration/registration_mednist.ipynb), |
| 61 | +- `pathology_nuclei_segmentation_and_classification`, |
| 62 | + `pathology_nuclick_annotation`, and `pathology_nuclei_classification` bundles |
| 63 | + are built for [digital pathology image |
| 64 | + analysis](https://github.com/Project-MONAI/model-zoo/tree/dev/models/pathology_nuclei_segmentation_classification). |
| 65 | + |
| 66 | +For more details about how to use the models, please see [the tutorials](https://github.com/Project-MONAI/tutorials/tree/main/model_zoo). |
| 67 | + |
| 68 | +## State-of-the-art SurgToolLoc solution |
| 69 | + |
| 70 | +[SurgToolLoc](https://surgtoolloc.grand-challenge.org/Home/) is a part of the |
| 71 | +[EndoVis](https://endovis.grand-challenge.org/) challenge at [MICCAI 2022](https://conferences.miccai.org/2022/en/). |
| 72 | +The challenge focuses on endoscopic video analysis and is divided into (1) fully supervised tool classification |
| 73 | +and (2) weakly supervised tool classification/localization. |
| 74 | +Team NVIDIA won prizes by finishing [third](https://surgtoolloc.grand-challenge.org/results/) in both categories. |
| 75 | +The core components of the solutions [are released in MONAI](https://github.com/Project-MONAI/tutorials/tree/main/competitions/MICCAI/surgtoolloc). |
0 commit comments