---
license: cc-by-nc-4.0
pretty_name: "2025 Dr. Robert Gillies ML Workshop - Cancer Survival Prediction"
tags:
- medical
- cancer
- glioblastoma
- pathology
- clinical-data
- machine-learning
- competition
- survival-prediction
- medical-imaging
- whole-slide-imaging
size_categories:
- 100K<n<1M
configs:
- config_name: default
  data_files:
  - split: train
    path: "train/train.csv"
  - split: test
    path: "test/test.csv"
---

# 2025 Dr. Robert Gillies Machine Learning Workshop in Cancer - Hackathon Dataset

## Dataset Description

This dataset is part of the **2025 Dr. Robert Gillies Machine Learning Workshop** at Moffitt Cancer Center. It contains de-identified medical imaging, pathology reports, and clinical data for glioblastoma patients, designed for an **overall survival prediction challenge**.

### Dataset Splits

- **Training Set:** 600 patients with complete survival outcomes
- **Test Set:** 265 patients with withheld survival outcomes

### Training Dataset

- **Total Patients:** 600
- **Pathology Images:** 600 whole-slide tissue images (.tif format)
- **Clinical Records:** 600 patient records with comprehensive clinical and pathology data
- **Total Size:** ~481GB

### Test Dataset

- **Total Patients:** 265
- **Pathology Images:** 265 whole-slide tissue images (.tif format)
- **Clinical Records:** 265 patient records with **predictive features only** (survival outcomes withheld)
- **Total Size:** ~213GB
- **Purpose:** Final evaluation with ground truth outcomes held by organizers

### Dataset Structure

```
train/
├── train.csv                           # Clinical and pathology report data (with outcomes)
└── images/                             # Pathology whole-slide images
    ├── metadata.csv                    # Links image files to patient IDs
    ├── P0002.tif                       # Patient pathology slides (~100-174MB each)
    ├── P0003.tif
    └── ... (600 patients)

test/
├── test.csv                            # Clinical features WITHOUT survival outcomes
└── images/                             # Pathology whole-slide images
    ├── metadata.csv                    # Links image files to patient IDs
    ├── P0001.tif                       # Patient pathology slides (~100-174MB each)
    ├── P0004.tif
    └── ... (265 patients)
```

### Clinical Data Schema (Training Set)

The `train.csv` file in the training set contains the following fields:

**Demographics:**
- `patient_id`: Unique de-identified patient identifier
- `age_at_diagnosis`: Patient age at diagnosis
- `gender`: Patient gender
- `race`: Patient race
- `ethnicity`: Patient ethnicity

**Tumor Characteristics:**
- `primary_diagnosis`: Primary cancer diagnosis
- `tumor_grade`: Tumor grade classification
- `ajcc_m`: AJCC M stage (metastasis)
- `ajcc_n`: AJCC N stage (lymph nodes)
- `classification_of_tumor`: Tumor classification (primary/recurrence/progression)
- `morphology`: ICD-O-3 morphology code
- `tissue_origin`: Anatomical site of origin
- `laterality`: Side of body affected
- `prior_malignancy`: History of prior cancer
- `synchronous_malignancy`: Concurrent cancer diagnosis

**Clinical Outcomes:**
- `vital_status`: Patient vital status (Alive/Dead)
- `overall_survival_days`: Overall survival time in days
- `overall_survival_event`: Survival event indicator (0/1)
- `days_to_death`: Time to death event
- `days_to_last_followup`: Time to last follow-up
- `cause_of_death`: Cause of death
- `days_to_progression`: Time to disease progression
- `days_to_recurrence`: Time to disease recurrence
- `progression_or_recurrence`: Indicator of disease progression/recurrence
- `disease_response`: Response to treatment

**Treatment Information:**
- `treatment_types`: Types of treatments received
- `therapeutic_agents`: Specific drugs/agents used
- `days_to_treatment`: Time to treatment initiation
- `treatment_outcome`: Outcome of treatment

**Pathology:**
- `pathology_report`: Complete de-identified pathology report text

### Test Set Features

The `test.csv` file in the test set contains **21 predictive features only**. The following 9 survival outcome columns have been **removed** to prevent participants from calculating C-index:

**Withheld Columns:**
- `vital_status` (Alive/Dead status)
- `overall_survival_days` (survival time - required for C-index)
- `overall_survival_event` (event indicator - required for C-index)
- `days_to_death`
- `days_to_last_followup`
- `cause_of_death`
- `days_to_progression`
- `days_to_recurrence`
- `progression_or_recurrence`

**Available Test Features (21 columns):**
- All demographic fields (patient_id, age_at_diagnosis, gender, race, ethnicity)
- All tumor characteristics (primary_diagnosis, tumor_grade, ajcc_m, ajcc_n, classification_of_tumor, morphology, tissue_origin, laterality, prior_malignancy, synchronous_malignancy)
- Treatment information (treatment_types, therapeutic_agents, days_to_treatment, treatment_outcome, disease_response)
- Pathology report (pathology_report)

## Submission Format

For the overall survival prediction challenge, submit predictions as a CSV file with the following format:

```csv
patient_id,outcome,risk_score,confidence
P0001,1,0.742,0.89
P0004,0,0.234,0.76
P0007,1,0.891,0.92
```

**Column Definitions:**
- `patient_id`: Patient identifier (must match test set IDs)
- `outcome`: Predicted binary outcome (0 = censored/alive, 1 = event/death)
- `risk_score`: Predicted risk score (higher = higher risk of event)
- `confidence`: Model confidence in prediction (0-1 scale)

**Evaluation Metrics:**
1. **C-Index (Concordance Index)** - 40%
2. **AUROC (Area Under ROC Curve)** - 30%
3. **Accuracy** - 20%
4. **Calibration** - 10%

## Usage

### Important Notes

This is a **multimodal dataset** combining:
- **Large pathology images** (~100-174MB each, TIFF format, ~700GB total)
- **Clinical/tabular data** (CSV format with patient metadata)

Due to the large image file sizes, we recommend:
1. Loading CSV files directly using pandas with `hf://` protocol
2. Downloading specific images on-demand using `hf_hub_download()`
3. Using `streaming=True` if iterating through the dataset
4. Only downloading the full dataset if you have sufficient storage (~700GB)

### Loading the Dataset

```python
from datasets import load_dataset
import pandas as pd

# Load the training set (with outcomes)
train_dataset = load_dataset("Lab-Rasool/hackathon-2025", split="train")
train_data = pd.read_csv("hf://datasets/Lab-Rasool/hackathon-2025/train/train.csv")

# Load the test set (without outcomes)
test_dataset = load_dataset("Lab-Rasool/hackathon-2025", split="test")
test_data = pd.read_csv("hf://datasets/Lab-Rasool/hackathon-2025/test/test.csv")

# Access images with metadata
from huggingface_hub import hf_hub_download
from PIL import Image
import tifffile

# Download and open a specific image
train_img_path = hf_hub_download(
    repo_id="Lab-Rasool/hackathon-2025",
    filename="train/images/P0002.tif",
    repo_type="dataset"
)
train_image = tifffile.imread(train_img_path)

# Or using datasets library with streaming
train_ds = load_dataset("Lab-Rasool/hackathon-2025", split="train", streaming=True)
# Note: For large image files, consider downloading specific files as needed
```

### Working with Images and Clinical Data

This dataset combines large pathology images (TIFF files) with clinical/tabular data. Here's how to efficiently work with both:

```python
import pandas as pd
import tifffile
from huggingface_hub import hf_hub_download, snapshot_download

# Option 1: Download entire dataset (requires ~700GB storage)
# local_dir = snapshot_download(repo_id="Lab-Rasool/hackathon-2025", repo_type="dataset")

# Option 2: Download only CSV files and images on-demand (recommended)
# Load clinical data
train_csv = pd.read_csv("hf://datasets/Lab-Rasool/hackathon-2025/train/train.csv")
test_csv = pd.read_csv("hf://datasets/Lab-Rasool/hackathon-2025/test/test.csv")

# Load image metadata to link images with patient IDs
train_metadata = pd.read_csv("hf://datasets/Lab-Rasool/hackathon-2025/train/images/metadata.csv")
test_metadata = pd.read_csv("hf://datasets/Lab-Rasool/hackathon-2025/test/images/metadata.csv")

# Download specific images as needed
def load_patient_image(patient_id, split="train"):
    """Load pathology image for a specific patient"""
    img_path = hf_hub_download(
        repo_id="Lab-Rasool/hackathon-2025",
        filename=f"{split}/images/{patient_id}.tif",
        repo_type="dataset"
    )
    return tifffile.imread(img_path)

# Example: Load image and clinical data for patient P0002
patient_id = "P0002"
patient_image = load_patient_image(patient_id, split="train")
patient_clinical = train_csv[train_csv['patient_id'] == patient_id]

print(f"Image shape: {patient_image.shape}")
print(f"Clinical data:\n{patient_clinical}")
```

### Example Use Cases

**Image Analysis:**
```python
# Load pathology image with clinical data
import tifffile
import pandas as pd
from huggingface_hub import hf_hub_download

# Download clinical data
train_data = pd.read_csv("hf://datasets/Lab-Rasool/hackathon-2025/train/train.csv")

# Download specific image
img_path = hf_hub_download(
    repo_id="Lab-Rasool/hackathon-2025",
    filename="train/images/P0002.tif",
    repo_type="dataset"
)
image = tifffile.imread(img_path)

# Get corresponding clinical data
clinical_info = train_data[train_data['patient_id'] == 'P0002']
```

**NLP on Pathology Reports:**
```python
import pandas as pd

# Load clinical data with pathology reports
train_data = pd.read_csv("hf://datasets/Lab-Rasool/hackathon-2025/train/train.csv")

# Extract pathology reports
reports = train_data[['patient_id', 'pathology_report']]
```

**Survival Analysis (Training Set Only):**
```python
# Analyze survival outcomes (only available in training set)
survival_data = train_data[[
    'patient_id',
    'overall_survival_days',
    'overall_survival_event',
    'vital_status'
]]
```

**Making Predictions on Test Set:**
```python
import tifffile
import pandas as pd
from huggingface_hub import hf_hub_download

# Load test features
test_features = pd.read_csv("hf://datasets/Lab-Rasool/hackathon-2025/test/test.csv")

# Download and load a test image
test_img_path = hf_hub_download(
    repo_id="Lab-Rasool/hackathon-2025",
    filename="test/images/P0001.tif",
    repo_type="dataset"
)
test_image = tifffile.imread(test_img_path)

# Extract features for a patient
patient = test_features[test_features['patient_id'] == 'P0001']

# Make predictions with your model
# risk_score = your_model.predict(patient_features)

# Create submission
submission = pd.DataFrame({
    'patient_id': test_features['patient_id'],
    'outcome': predicted_outcomes,
    'risk_score': predicted_risk_scores,
    'confidence': prediction_confidences
})

submission.to_csv('my_submission.csv', index=False)
```
## License

This dataset is released under the **Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0)** license.

- ✅ Academic and research use permitted
- ✅ Educational use permitted
- ❌ Commercial use prohibited
- ✅ Derivative works permitted with attribution

## Important Notes

1. **Training Set:** Use for model development and training. Contains complete survival outcomes.
2. **Test Set:** Use for final predictions. Survival outcomes are withheld for evaluation.
3. **No Self-Evaluation:** Ground truth outcomes for the test set are held by competition organizers.
4. **Fair Use:** Do not attempt to obtain test set ground truth labels through external sources.
5. **Submission Required:** All 265 test patients must be included in submissions with matching patient IDs.