diff --git a/README.md b/README.md index d58360dd2..32ae1cbac 100644 --- a/README.md +++ b/README.md @@ -83,6 +83,7 @@ In addition, you can find a table of the basic supported fields, modalities, vie
  • Segmentation
  • DeepGrow
  • DeepEdit
  • +
  • VISTA3D (Universal)
  • SAM2 (2D/3D)
  • @@ -345,6 +346,11 @@ monailabel start_server --app apps/radiology --studies datasets/Task09_Spleen/im - [MONAILabel: Multi-organ Segmentation with 3D Slicer](https://github.com/Project-MONAI/tutorials/blob/main/monailabel/monailabel_monaibundle_3dslicer_multiorgan_seg.ipynb): Multi-organ segmentation with CT scans in 3D Slicer. - [MONAILabel: Whole Body CT Segmentation with 3D Slicer](https://github.com/Project-MONAI/tutorials/blob/main/monailabel/monailabel_wholebody_totalSegmentator_3dslicer.ipynb): Whole body (104 structures) segmentation with CT scans. - [MONAILabel: Lung nodule CT Detection with 3D Slicer](https://github.com/Project-MONAI/tutorials/blob/main/monailabel/monailabel_monaibundle_3dslicer_lung_nodule_detection.ipynb): Lung nodule detection task with CT scans. +- **VISTA3D App**: + - Viewer: [OHIF](https://ohif.org/) | [3D Slicer](https://www.slicer.org/) | Datastore: Local/DICOM Web | Task: Multi-organ Segmentation + - [MONAILabel: VISTA3D Quick Start](docs/source/tutorials/vista3d_quickstart.md): Fast setup guide for VISTA3D universal segmentation model. + - [MONAILabel: VISTA3D Complete Tutorial](docs/source/tutorials/vista3d_tutorial.md): Comprehensive tutorial for VISTA3D model setup, interactive prompting, and multi-organ segmentation workflows with 117+ anatomical structures. + - [MONAILabel: VISTA3D Configuration Examples](docs/source/tutorials/vista3d_configurations.md): Sample configurations for different VISTA3D deployment scenarios. - **Pathology App**: - Viewer: [QuPath](https://qupath.github.io/) | Datastore: Local | Task: Segmentation - [MONAILabel: Nuclei Segmentation with QuPath](https://github.com/Project-MONAI/tutorials/blob/main/monailabel/monailabel_pathology_nuclei_segmentation_QuPath.ipynb) Nuclei segmentation with QuPath setup and Nuclick models. diff --git a/docs/source/index.rst b/docs/source/index.rst index 95bd2c749..fd9bdb719 100644 --- a/docs/source/index.rst +++ b/docs/source/index.rst @@ -80,6 +80,7 @@ Table of Contents whatsnew installation quickstart + tutorials modules appdeployment apis diff --git a/docs/source/tutorials.rst b/docs/source/tutorials.rst new file mode 100644 index 000000000..c9576fc07 --- /dev/null +++ b/docs/source/tutorials.rst @@ -0,0 +1,95 @@ +.. comment + Copyright (c) MONAI Consortium + Licensed under the Apache License, Version 2.0 (the "License"); + you may not use this file except in compliance with the License. + You may obtain a copy of the License at + http://www.apache.org/licenses/LICENSE-2.0 + Unless required by applicable law or agreed to in writing, software + distributed under the License is distributed on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + See the License for the specific language governing permissions and + limitations under the License. + +========= +Tutorials +========= + +This section provides comprehensive tutorials for using MONAI Label with advanced models and viewers. + +VISTA3D Universal Segmentation +=============================== + +VISTA3D is a universal foundation model for 3D CT segmentation that can segment 117+ anatomical structures +including organs, bones, vessels, and soft tissues. It supports both automatic and interactive segmentation workflows. + +**Quick Start** + +For immediate setup and basic usage, see the `VISTA3D Quick Start Guide`_. + +**Complete Tutorial** + +For comprehensive instructions including advanced configuration, troubleshooting, and clinical workflows, +see the `VISTA3D Complete Tutorial`_. + +**Configuration Examples** + +For sample configurations covering different deployment scenarios (memory optimization, PACS integration, +multi-model setup, etc.), see the `VISTA3D Configuration Examples`_. + +Key Features +------------ + +- **Universal Segmentation**: Supports 117+ anatomical classes +- **Interactive Prompting**: Point-based and class-based interactive segmentation +- **Multi-Viewer Support**: Works with OHIF and 3D Slicer +- **High Performance**: Optimized for clinical workflows +- **Flexible Deployment**: Local, cloud, and PACS integration options + +Supported Viewers +----------------- + +VISTA3D integration is available through: + +- **OHIF Viewer**: Web-based DICOM viewer with full VISTA3D support +- **3D Slicer**: Desktop application with VISTA3D capabilities + +Prerequisites +------------- + +- Python 3.8+ with MONAI Label installed +- NVIDIA GPU with 8GB+ VRAM (16GB+ recommended) +- CT data in DICOM or NIfTI format +- Compatible viewer (OHIF or 3D Slicer) + +Basic Usage +----------- + +1. **Setup**:: + + # Download bundle app + monailabel apps --download --name monaibundle --output apps + + # Start VISTA3D server + monailabel start_server \\ + --app apps/monaibundle \\ + --studies datasets/ct_volumes \\ + --conf models vista3d \\ + --conf preload true + +2. **Automatic Segmentation**: Load CT volume and run auto-segmentation for 117+ structures + +3. **Interactive Prompting**: Use point prompts to refine segmentation of specific anatomy + +4. **Class Selection**: Choose anatomical categories (organs, bones, vessels) for targeted segmentation + +For detailed instructions, examples, and troubleshooting, refer to the tutorial links above. + +Additional Resources +==================== + +For more MONAI Label tutorials covering other models and use cases, visit the +`MONAI Tutorials Repository `_. + +.. _VISTA3D Quick Start Guide: tutorials/vista3d_quickstart.md +.. _VISTA3D Complete Tutorial: tutorials/vista3d_tutorial.md +.. _VISTA3D Configuration Examples: tutorials/vista3d_configurations.md \ No newline at end of file diff --git a/docs/source/tutorials/README.md b/docs/source/tutorials/README.md new file mode 100644 index 000000000..5c0b47e9c --- /dev/null +++ b/docs/source/tutorials/README.md @@ -0,0 +1,23 @@ +# MONAI Label Tutorials + +This directory contains comprehensive tutorials for using MONAI Label with various models and viewers. + +## Available Tutorials + +### VISTA3D Universal Segmentation +- **[VISTA3D Quick Start](vista3d_quickstart.md)** - Get started with VISTA3D in minutes +- **[VISTA3D Complete Tutorial](vista3d_tutorial.md)** - Comprehensive guide for VISTA3D setup and usage +- **[VISTA3D Configuration Examples](vista3d_configurations.md)** - Sample configurations for different deployment scenarios + +### Getting Started +1. Start with the [Quick Start Guide](vista3d_quickstart.md) for immediate setup +2. Read the [Complete Tutorial](vista3d_tutorial.md) for detailed instructions +3. Use [Configuration Examples](vista3d_configurations.md) for specific deployment needs + +### Prerequisites +- MONAI Label installed +- Compatible viewer (OHIF, 3D Slicer) +- CT data for segmentation +- GPU recommended for optimal performance + +For general MONAI Label tutorials, visit the [MONAI Tutorials Repository](https://github.com/Project-MONAI/tutorials/tree/main/monailabel). \ No newline at end of file diff --git a/docs/source/tutorials/vista3d_configurations.md b/docs/source/tutorials/vista3d_configurations.md new file mode 100644 index 000000000..298fc2daa --- /dev/null +++ b/docs/source/tutorials/vista3d_configurations.md @@ -0,0 +1,187 @@ +# VISTA3D Configuration Examples + +This document provides sample configurations for integrating VISTA3D with MONAI Label. + +## Basic VISTA3D Server Configuration + +### 1. Standard Setup +```bash +monailabel start_server \ + --app apps/monaibundle \ + --studies datasets/ct_volumes \ + --conf models vista3d \ + --conf preload true \ + --conf skip_trainers true +``` + +### 2. Memory-Optimized Setup +```bash +monailabel start_server \ + --app apps/monaibundle \ + --studies datasets/ct_volumes \ + --conf models vista3d \ + --conf vista3d_patch_size "[64,64,64]" \ + --conf vista3d_overlap 0.125 \ + --conf vista3d_cpu_fallback true +``` + +### 3. High-Performance Setup +```bash +monailabel start_server \ + --app apps/monaibundle \ + --studies datasets/ct_volumes \ + --conf models vista3d \ + --conf vista3d_patch_size "[128,128,128]" \ + --conf vista3d_overlap 0.5 \ + --conf preload true \ + --conf workers 4 +``` + +## VISTA3D with DICOM Web + +### 4. PACS Integration +```bash +monailabel start_server \ + --app apps/monaibundle \ + --studies "http://pacs.hospital.com/dicom-web" \ + --studies_user "username" \ + --studies_password "password" \ + --conf models vista3d \ + --conf preload true +``` + +## Custom VISTA3D Bundle Configuration + +### 5. Custom Bundle Path +```bash +monailabel start_server \ + --app apps/monaibundle \ + --studies datasets/ct_volumes \ + --conf models "vista3d:/path/to/custom/vista3d/bundle" \ + --conf vista3d_prompting true +``` + +## Multi-Model Setup + +### 6. VISTA3D with Other Models +```bash +monailabel start_server \ + --app apps/monaibundle \ + --studies datasets/ct_volumes \ + --conf models "vista3d,spleen_ct_segmentation,swin_unetr_btcv_segmentation" \ + --conf preload true +``` + +## VISTA3D with Interactive Features + +### 7. Full Interactive Setup +```bash +monailabel start_server \ + --app apps/monaibundle \ + --studies datasets/ct_volumes \ + --conf models vista3d \ + --conf vista3d_prompting true \ + --conf vista3d_point_prompts true \ + --conf vista3d_class_prompts true \ + --conf preload true +``` + +## Development and Testing + +### 8. Debug Configuration +```bash +monailabel start_server \ + --app apps/monaibundle \ + --studies datasets/ct_volumes \ + --conf models vista3d \ + --conf debug true \ + --conf vista3d_cache_transforms false \ + --log_level DEBUG +``` + +### 9. Performance Profiling +```bash +monailabel start_server \ + --app apps/monaibundle \ + --studies datasets/ct_volumes \ + --conf models vista3d \ + --conf profile true \ + --conf vista3d_profile_inference true +``` + +## Docker Deployment + +### 10. Docker with VISTA3D +```bash +docker run --gpus all --rm -ti \ + -p 8000:8000 \ + -v /data/ct_volumes:/data/studies \ + -v /data/models:/data/models \ + projectmonai/monailabel:latest \ + monailabel start_server \ + --app /opt/monailabel/sample-apps/monaibundle \ + --studies /data/studies \ + --conf models vista3d \ + --conf preload true \ + --host 0.0.0.0 +``` + +## Configuration Parameters + +| Parameter | Description | Default | Example | +|-----------|-------------|---------|---------| +| `vista3d_patch_size` | Input patch size for inference | `[96,96,96]` | `"[64,64,64]"` | +| `vista3d_overlap` | Sliding window overlap ratio | `0.25` | `0.5` | +| `vista3d_prompting` | Enable interactive prompting | `false` | `true` | +| `vista3d_point_prompts` | Enable point-based prompting | `true` | `false` | +| `vista3d_class_prompts` | Enable class-based prompting | `true` | `false` | +| `vista3d_cpu_fallback` | Use CPU when GPU memory insufficient | `false` | `true` | +| `vista3d_cache_transforms` | Cache preprocessing transforms | `true` | `false` | +| `vista3d_profile_inference` | Profile inference performance | `false` | `true` | + +## Environment Variables + +```bash +# Set GPU memory fraction +export CUDA_VISIBLE_DEVICES=0 +export PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb:512 + +# Set cache directories +export MONAI_CACHE_DIR=/data/cache +export VISTA3D_CACHE_DIR=/data/vista3d_cache + +# Start server with environment +monailabel start_server \ + --app apps/monaibundle \ + --studies datasets/ct_volumes \ + --conf models vista3d +``` + +## Troubleshooting Common Configurations + +### GPU Memory Issues +```bash +# Reduce memory usage +--conf vista3d_patch_size "[48,48,48]" +--conf vista3d_overlap 0.125 +--conf vista3d_cpu_fallback true +``` + +### Network/Download Issues +```bash +# Use local model cache +--conf zoo_source local +--conf model_path /local/path/to/vista3d + +# Use alternative zoo repo +--conf zoo_repo "custom-repo/model-zoo" +``` + +### Performance Optimization +```bash +# Optimize for speed +--conf vista3d_patch_size "[128,128,128]" +--conf vista3d_overlap 0.5 +--conf workers 8 +--conf preload true +``` \ No newline at end of file diff --git a/docs/source/tutorials/vista3d_quickstart.md b/docs/source/tutorials/vista3d_quickstart.md new file mode 100644 index 000000000..0503eea20 --- /dev/null +++ b/docs/source/tutorials/vista3d_quickstart.md @@ -0,0 +1,79 @@ +# VISTA3D Quick Start Guide + +This quick start guide helps you get VISTA3D running with MONAI Label in minutes. + +## What is VISTA3D? + +VISTA3D is a universal foundation model for 3D CT segmentation that can segment 117+ anatomical structures including organs, bones, vessels, and soft tissues. It supports both automatic and interactive segmentation workflows. + +## Quick Setup + +### 1. Prerequisites +- Python 3.8+ +- NVIDIA GPU with 8GB+ VRAM +- MONAI Label installed + +### 2. Download and Start +```bash +# Download the bundle app +monailabel apps --download --name monaibundle --output apps + +# Start VISTA3D server +monailabel start_server \ + --app apps/monaibundle \ + --studies /path/to/ct/volumes \ + --conf models vista3d \ + --conf preload true +``` + +### 3. Connect Viewer + +**For OHIF Viewer:** +```bash +cd apps/monaibundle +# Follow OHIF setup instructions in main tutorial +``` + +**For 3D Slicer:** +1. Install MONAI Label extension +2. Connect to server at `http://localhost:8000` +3. VISTA3D models will appear automatically + +## Basic Usage + +### Automatic Segmentation +1. Load CT volume in your viewer +2. Select VISTA3D model +3. Run auto-segmentation for 117+ structures + +### Interactive Prompting +1. Choose target anatomy (e.g., liver, kidney) +2. Click positive points on the structure +3. Right-click negative points (optional) +4. Generate refined segmentation + +### Class Selection +1. Choose anatomical categories: + - Organs (liver, spleen, kidneys, etc.) + - Bones (vertebrae, ribs, pelvis, etc.) + - Vessels (aorta, vena cava, etc.) +2. Run targeted segmentation + +## Common Use Cases + +- **Clinical**: Multi-organ analysis, surgical planning +- **Research**: Anatomical quantification, population studies +- **AI Development**: Training data generation, model evaluation + +## Need Help? + +- 📖 [Complete VISTA3D Tutorial](vista3d_tutorial.md) +- 🐛 [Report Issues](https://github.com/Project-MONAI/MONAILabel/issues) +- 💬 [Community Support](https://projectmonai.slack.com/) + +## Next Steps + +- Explore advanced configuration options +- Set up automated workflows +- Integrate with clinical systems +- Customize for specific use cases \ No newline at end of file diff --git a/docs/source/tutorials/vista3d_tutorial.md b/docs/source/tutorials/vista3d_tutorial.md new file mode 100644 index 000000000..b50ec548d --- /dev/null +++ b/docs/source/tutorials/vista3d_tutorial.md @@ -0,0 +1,273 @@ +# VISTA3D Tutorial for MONAI Label + +VISTA3D (Versatile Imaging SegmenTation and Annotation model for 3D Computed Tomography) is an advanced foundation model for 3D medical image segmentation that can segment a wide variety of anatomical structures in CT volumes. + +## Overview + +VISTA3D is designed to work as a universal segmentation model that can handle multiple anatomical structures simultaneously. It supports both automatic segmentation and interactive segmentation with point prompts, making it highly versatile for clinical workflows. + +### Key Features + +- **Universal Segmentation**: Supports 117+ anatomical classes including organs, bones, vessels, and tumors +- **Interactive Prompting**: Point-based interactive segmentation for precise control +- **Class-based Prompting**: Segment specific anatomical structures by class selection +- **3D Native**: Designed specifically for 3D CT volumes +- **High Performance**: State-of-the-art segmentation accuracy across diverse anatomy + +## Supported Viewers + +VISTA3D integration is currently available through: + +- **OHIF Viewer**: Web-based DICOM viewer with full VISTA3D support +- **3D Slicer**: Desktop application with VISTA3D capabilities (requires bundle integration) + +## Prerequisites + +### System Requirements + +- **GPU**: NVIDIA GPU with at least 8GB VRAM (16GB+ recommended) +- **RAM**: 16GB+ system memory +- **Python**: Python 3.8+ with MONAI Label installed +- **Storage**: Several GB for model weights + +### Software Installation + +1. Install MONAI Label with VISTA3D dependencies: +```bash +pip install monailabel[vista3d] +``` + +2. For development/latest features: +```bash +git clone https://github.com/Project-MONAI/MONAILabel +cd MONAILabel +pip install -e . +``` + +## Data Preparation + +VISTA3D works with standard CT DICOM series or NIfTI volumes. Prepare your data as follows: + +### For Local Datastore +``` +datasets/ +├── ct_volume_001.nii.gz +├── ct_volume_002.nii.gz +└── ... +``` + +### For DICOM Web +Ensure your PACS/DICOM server supports DICOMWeb standard for direct integration. + +## Setting Up VISTA3D + +### Method 1: Using MONAIBundle App (Recommended) + +The VISTA3D model can be integrated through the MONAI Bundle application: + +```bash +# Download the monaibundle app +monailabel apps --download --name monaibundle --output apps + +# Start server with VISTA3D model (when available in model zoo) +monailabel start_server \ + --app apps/monaibundle \ + --studies datasets/ct_volumes \ + --conf models vista3d_v1.0.0 \ + --conf preload true +``` + +### Method 2: Custom VISTA3D Configuration + +For advanced users wanting to customize VISTA3D behavior: + +1. Create a custom configuration in your bundle app: +```python +# lib/configs/vista3d_config.py +from monailabel.tasks.infer.bundle import BundleInferTask + +class VISTA3DConfig: + def __init__(self): + self.model_name = "vista3d" + self.bundle_path = "path/to/vista3d/bundle" + self.labels = { + "liver": 1, + "spleen": 2, + "kidney_left": 3, + "kidney_right": 4, + # ... additional anatomical labels + } +``` + +## Usage Examples + +### 1. Automatic Segmentation + +Start the server and use the Auto Segmentation feature: + +```bash +# Start MONAI Label server +monailabel start_server \ + --app apps/monaibundle \ + --studies datasets/ct_volumes \ + --conf models vista3d_v1.0.0 + +# Open OHIF viewer and navigate to: +# http://localhost:3000/viewer?StudyInstanceUIDs= +``` + +In the OHIF viewer: +1. Load a CT volume +2. Open MONAI Label plugin panel +3. Select "Auto Segmentation" tab +4. Choose VISTA3D model +5. Click "Run Segmentation" for automatic multi-organ segmentation + +### 2. Interactive Point Prompting + +For targeted segmentation with point prompts: + +1. Load CT volume in OHIF viewer +2. Select "Point Prompts" tab in MONAI Label panel +3. Choose VISTA3D model +4. Select target anatomical structure from dropdown +5. Click on the structure in the image to add positive points +6. Right-click to add negative points (background) +7. Click "Run Inference" to generate segmentation + +### 3. Class-based Prompting + +For segmenting specific anatomical categories: + +1. Select "Class Prompts" tab +2. Choose VISTA3D model +3. Select anatomical categories: + - **Organs**: Liver, spleen, kidneys, etc. + - **Bones**: Vertebrae, ribs, pelvis, etc. + - **Vessels**: Aorta, vena cava, pulmonary vessels, etc. +4. Click "Run Inference" to segment selected classes + +## Advanced Configuration + +### Model Parameters + +VISTA3D supports various inference parameters that can be configured: + +```bash +# Start with custom parameters +monailabel start_server \ + --app apps/monaibundle \ + --studies datasets \ + --conf models vista3d_v1.0.0 \ + --conf vista3d_patch_size "[96,96,96]" \ + --conf vista3d_overlap 0.25 \ + --conf vista3d_batch_size 1 +``` + +### Memory Optimization + +For systems with limited GPU memory: + +```bash +# Use smaller patch sizes and enable CPU fallback +monailabel start_server \ + --app apps/monaibundle \ + --studies datasets \ + --conf models vista3d_v1.0.0 \ + --conf vista3d_patch_size "[64,64,64]" \ + --conf vista3d_cpu_fallback true +``` + +## Viewer-Specific Instructions + +### OHIF Viewer Setup + +1. **Install OHIF with MONAI Label Extension**: +```bash +# Clone OHIF with MONAI Label extension +git clone https://github.com/Project-MONAI/MONAILabel +cd MONAILabel/plugins/ohifv3 + +# Install dependencies and start +npm install +npm run dev +``` + +2. **Configure OHIF for VISTA3D**: + - The OHIF extension automatically detects VISTA3D models + - Point prompts and class prompts are available in the MONAI Label panel + - Auto segmentation supports multi-organ prediction + +### 3D Slicer Integration + +For 3D Slicer users: + +1. **Install MONAI Label Extension**: + - Open 3D Slicer + - Go to Extension Manager + - Search and install "MONAI Label" + +2. **Connect to VISTA3D Server**: + - Open MONAI Label module + - Set server URL: `http://localhost:8000` + - VISTA3D models will appear in the model selection + +## Troubleshooting + +### Common Issues + +1. **GPU Memory Errors**: + - Reduce patch size: `--conf vista3d_patch_size "[48,48,48]"` + - Enable CPU inference: `--conf vista3d_device cpu` + +2. **Model Loading Failures**: + - Verify VISTA3D bundle is correctly downloaded + - Check network connectivity for model zoo access + - Ensure sufficient disk space for model weights + +3. **Slow Inference**: + - Verify GPU is being used: check `nvidia-smi` + - Optimize batch size based on available GPU memory + - Consider model quantization for faster inference + +### Performance Tips + +- **Preload Models**: Use `--conf preload true` for faster first inference +- **Batch Processing**: Process multiple volumes in sequence for efficiency +- **ROI Selection**: Use viewer tools to define regions of interest for faster processing + +## Example Workflows + +### Clinical Workflow: Abdominal Organ Segmentation + +1. Load abdominal CT scan in OHIF +2. Select VISTA3D model +3. Use "Class Prompts" to select: + - Liver + - Spleen + - Left/Right Kidneys + - Pancreas +4. Run segmentation and review results +5. Use point prompts to refine any inaccurate regions +6. Export segmentations for clinical use + +### Research Workflow: Comprehensive Anatomy Analysis + +1. Load full-body CT scan +2. Run automatic segmentation with VISTA3D +3. Review 117+ automatically segmented structures +4. Use interactive prompts to add/remove structures as needed +5. Export detailed anatomical segmentations for analysis + +## References + +- [VISTA3D Paper](https://arxiv.org/abs/2406.05285): Versatile Imaging SegmenTation and Annotation model for 3D Computed Tomography +- [MONAI Bundle Specification](https://docs.monai.io/en/latest/mb_specification.html) +- [MONAI Label Documentation](https://docs.monai.io/projects/label/en/latest/) + +## Support + +For VISTA3D-specific issues: +- Check [MONAI Label GitHub Issues](https://github.com/Project-MONAI/MONAILabel/issues) +- Join [MONAI Slack Community](https://projectmonai.slack.com/) +- Review [MONAI Label Discussions](https://github.com/Project-MONAI/MONAILabel/discussions) \ No newline at end of file diff --git a/sample-apps/monaibundle/README.md b/sample-apps/monaibundle/README.md index 9b9b67191..7ba150ef3 100644 --- a/sample-apps/monaibundle/README.md +++ b/sample-apps/monaibundle/README.md @@ -43,6 +43,7 @@ The MONAIBundle App currently supports most labeling models in the Model-Zoo. Yo | [spleen_deepedit_annotation](https://github.com/Project-MONAI/model-zoo/tree/dev/models/spleen_deepedit_annotation) | DeepEdit | Spleen| CT | An interactive method for 3D spleen Segmentation | | [lung_nodule_ct_detection](https://github.com/Project-MONAI/model-zoo/tree/dev/models/lung_nodule_ct_detection) | RetinaNet | Lung Nodule| CT | The detection model for 3D CT images | | [wholeBody_ct_segmentation](https://github.com/Project-MONAI/model-zoo/tree/dev/models/wholeBody_ct_segmentation) | SegResNet | 104 body structures| CT | The segmentation model for 104 tissue from 3D CT images (TotalSegmentator Dataset) | +| [vista3d](https://github.com/Project-MONAI/model-zoo/tree/dev/models/vista3d) | VISTA3D | 117+ anatomical structures | CT | Universal foundation model for 3D CT segmentation with interactive prompting and multi-organ support | ### How To Use the App @@ -65,6 +66,12 @@ monailabel start_server --app workspace/monaibundle --studies workspace/images - # Skip Training Tasks or Infer only mode monailabel start_server --app workspace/monaibundle --studies workspace/images --conf models spleen_ct_segmentation_v0.1.0 --conf skip_trainers true + +# Use VISTA3D for universal multi-organ segmentation +monailabel start_server --app workspace/monaibundle --studies workspace/images --conf models vista3d --conf preload true + +# Use VISTA3D with interactive prompting enabled +monailabel start_server --app workspace/monaibundle --studies workspace/images --conf models vista3d --conf vista3d_prompting true ``` **Specify bundle version** (Optional) @@ -197,4 +204,8 @@ To set configuration parameters for MONAI Label Server, use the `--conf < | skip_trainers | true, **false** | Skip adding training tasks (Run in Infer mode only) | | epistemic_max_samples | int | _Default value:_ 0 ; Epistemic scoring parameters | | epistemic_simulation_size | int | _Default value:_ 5 ; Epistemic simulation size parameters | -| epistemic_dropout | float | _Default value:_ 0.2 ; Epistemic scoring parameters: Dropout rate for scoring models | | +| epistemic_dropout | float | _Default value:_ 0.2 ; Epistemic scoring parameters: Dropout rate for scoring models | +| vista3d_prompting | true, **false** | Enable interactive point and class prompting for VISTA3D | +| vista3d_patch_size | array | _Default value:_ [96,96,96] ; Patch size for VISTA3D inference | +| vista3d_overlap | float | _Default value:_ 0.25 ; Sliding window overlap for VISTA3D | +| vista3d_cpu_fallback | true, **false** | Use CPU inference when GPU memory is insufficient | |