Update README.md
Browse files
README.md
CHANGED
|
@@ -24,7 +24,6 @@ Model weights for the paper **"Prompt-Guided Interactive Segmentation of Interst
|
|
| 24 |
|---|---|---|
|
| 25 |
| nnU-Net (Lung, ILD, ILD raw(without masking the lung from CT)) | Automatic Lung, ILD segmentation in thoracic CT | `[(https://huggingface.co/VasilisDedousis/ILD-SemiSegTool/tree/main/HuggingFace/nnUNet_weights)]` |
|
| 26 |
| MedSAM2-ILD | Prompt-guided interactive ILD refinement | `[(https://huggingface.co/VasilisDedousis/ILD-SemiSegTool/tree/main/HuggingFace/MedSAM2%2BnnUNet_weights)]` |
|
| 27 |
-
| nnU-Net (Airway) | Airway segmentation in thoracic CT | `[(https://huggingface.co/VasilisDedousis/ILD-SemiSegTool/tree/main/HuggingFace/nnUNet_weights/Airways)]` |
|
| 28 |
|
| 29 |
The models are served through a MONAI Label radiology app with an OHIF viewer and Orthanc storage. See the GitHub repository for setup.
|
| 30 |
|
|
@@ -39,8 +38,9 @@ Then follow the setup instructions in the GitHub README to place the weights in
|
|
| 39 |
|
| 40 |
## Training data
|
| 41 |
|
| 42 |
-
|
| 43 |
|
|
|
|
| 44 |
## Intended use
|
| 45 |
|
| 46 |
For research use only. These models are not validated for clinical use and must not be used for diagnosis or treatment decisions.
|
|
@@ -61,4 +61,6 @@ For research use only. These models are not validated for clinical use and must
|
|
| 61 |
## Contact
|
| 62 |
|
| 63 |
AI in Health and Nutrition (AIHN) Lab, ARTORG Center, University of Bern.
|
| 64 |
-
https://www.artorg.unibe.ch/research/aihn/index_eng.html
|
|
|
|
|
|
|
|
|
| 24 |
|---|---|---|
|
| 25 |
| nnU-Net (Lung, ILD, ILD raw(without masking the lung from CT)) | Automatic Lung, ILD segmentation in thoracic CT | `[(https://huggingface.co/VasilisDedousis/ILD-SemiSegTool/tree/main/HuggingFace/nnUNet_weights)]` |
|
| 26 |
| MedSAM2-ILD | Prompt-guided interactive ILD refinement | `[(https://huggingface.co/VasilisDedousis/ILD-SemiSegTool/tree/main/HuggingFace/MedSAM2%2BnnUNet_weights)]` |
|
|
|
|
| 27 |
|
| 28 |
The models are served through a MONAI Label radiology app with an OHIF viewer and Orthanc storage. See the GitHub repository for setup.
|
| 29 |
|
|
|
|
| 38 |
|
| 39 |
## Training data
|
| 40 |
|
| 41 |
+
## Training data
|
| 42 |
|
| 43 |
+
The models were trained on 306 thoracic CT scans (1 mm slices, hard-kernel reconstruction) from patients with fibrotic and obstructive lung disease, drawn from the LTRC repository and the cohort of Christe et al. (2019). Annotations cover seven ILD patterns (GGO, reticulation, consolidation, honeycombing, reticulation+GGO, traction bronchiectasis, emphysema) plus healthy tissue. Data were split 80/10/10 at scan level. See the [paper](https://arxiv.org/abs/2608.28453) for details.
|
| 44 |
## Intended use
|
| 45 |
|
| 46 |
For research use only. These models are not validated for clinical use and must not be used for diagnosis or treatment decisions.
|
|
|
|
| 61 |
## Contact
|
| 62 |
|
| 63 |
AI in Health and Nutrition (AIHN) Lab, ARTORG Center, University of Bern.
|
| 64 |
+
https://www.artorg.unibe.ch/research/aihn/index_eng.html
|
| 65 |
+
|
| 66 |
+
For questions, please open an [issue](https://github.com/AIHNlab/ILD-SemiSegTool/issues).
|