Datasets, Papers, Models, and Tools Documentation
Datasets
| # | Dataset Name | Type | Status | Source/Link |
|---|---|---|---|---|
| 1 | MICCAI BraTS 2019 Data Training | Nifti | Available | Kaggle - Brain Tumor Segmentation |
| 2 | Primary Nasopharyngeal Carcinoma MRI with Multi-modalities Segmentation | Dicom (.dcm) | Available | Zenodo |
| 3 | SinusSegment Repository Dataset (Nasal Cavity) | Unknown | Requested (No reply) | Requested via email to rheadkaul@gmail.com |
| 4 | BraTS 2023 Dataset | Nifti (likely) | Requested (No access) | Synapse |
| 5 | nnInteractive Research Paper Datasets | Various | Listed in paper | GQC SharePoint |
| 6 | NasalSeg Dataset (Automatic Segmentation of Nasal Cavity and Paranasal Sinuses) | NNRD | Available | Zenodo |
| 7 | HeadNeck CT | - | Local | Available locally |
| 8 | Dicom Datasets from Sudhir | Dicom | Local | Available locally |
| 9 | CTA Head and Neck Dataset | - | Volview URL | Kitware Data |
| 10 | MRA Head and Neck | - | Volview URL | Kitware Data |
| 11 | MRI Cardiac 3D Cine | - | Volview URL | Kitware Data |
| 12 | MRI PROSTATEx | - | Volview URL | Kitware Data |
Azure Blob Storage Datasets
| Dataset ID (RowKey) | Description | Container | Segmented |
|---|---|---|---|
| Headneck CT.zip | Headneck CT | zip-files | No |
| CTA-Head_and_Neck.zip | CTA-Head and Neck | zip-files | No |
| BraTS19_TCIA13_653_1.zip | BraTS19 TCIA13 653_1 | zip-files | No |
| BraTS19_TCIA13_654_1.zip | BraTS19 TCIA13 654_1 | zip-files | No |
The last two datasets (BraTS19_TCIA13_653_1.zip and BraTS19_TCIA13_654_1.zip) are from the BraTS 2019 dataset which is #1 in Datasets table above.
Referenced Papers
| # | Paper Title | Link/Reference |
|---|---|---|
| 1 | NNInteractive Research Paper | arXiv |
| 2 | Development of an Open-Source Algorithm for Automated Segmentation in Clinician-Led Paranasal Sinus Radiologic Research | GQC SharePoint |
| 3 | A Dataset of Primary Nasopharyngeal Carcinoma MRI with Multi-modalities Segmentation | Nature Scientific Data |
Model Directory
| # | Model Name/Description | Base Architecture | Training Dataset | Training Platform | Runtime | Link |
|---|---|---|---|---|---|---|
| 1 | Brain Tumor Segmentation U-Net (Zeeshan's Original) | U-Net | MICCAI BraTS 2019 | Kaggle | ~4 hours (CPU), ~1h 35m (L4 GPU) | Kaggle Notebook |
| 2 | Brain Tumor Segmentation U-Net with Weights and Biases | U-Net | MICCAI BraTS 2019 | Google Colab | - | Colab Notebook |
| 3 | Nasopharyngeal Carcinoma Segmentation Model | U-Net (Zeeshan's backbone) | Primary Nasopharyngeal Carcinoma MRI (converted to Nifti) | Google Colab | - | Custom notebook (dataset restructured to match BraTS format) |
| 4 | SinusSegment Model | Uses UNetPlusPlus Model | NasalSeg Dataset (Automatic Segmentation of Nasal Cavity and Paranasal Sinuses) | - | - | https://github.com/rheadkaul/SinusSegment |
Tools and Documentation
| Name/Description | Type | Link |
|---|---|---|
| ITK-SNAP Software | Software Tool | ITK-SNAP Downloads |
| ITK-SNAP DLS Documentation | Documentation | ITK-SNAP DLS Quick Start |
Relationships Diagram
Notes
Requested Datasets: SinusSegment repository dataset and BraTS 2023 dataset have been requested but access has not been granted yet.
Dataset Conversion: The Primary Nasopharyngeal Carcinoma MRI dataset (Dicom) was converted to Nifti format and restructured to match the BraTS 2019 dataset format for model training.
Model Training: Zeeshan's Brain Tumor Segmentation Uses UNET Model for training Ran for almost 4 hours without connecting to the L4 runtime GPU Ran for almost 1 hour 35 minutes connecting it to the L4 runtime GPU
Nasopharyngeal carcinoma notebook: Converted the DICOM files into NIfTI format and restructured the dataset so that it follows the same format as the brain tumor segmentation dataset used in Zeeshan’s Kaggle notebook. Using Zeeshan’s U-Net model as the backbone, I successfully trained the model on the nasopharyngeal carcinoma dataset.
Local Datasets: HeadNeck CT and Dicom datasets from Sudhir are available locally but may not have public links.