Guidelines & Rules 📜
The challenge happens 100% online across two epic stages. Read up on the mechanics, how you will be scored, and the ground rules to make sure your team is ready to dominate! 🚀
🗺️ The Game Plan (Competition Structure)
🛠️ Stage 1: The Build (Asynchronous)
Deadline: 25 March 2026
Submit your project via the official portal:
- 📄 Tech Report (Max 5 pages): Detail your problem statement, dataset justification, ML/DL methodology, preprocessing steps, results, interpretability, and clinical impact.
- 🎥 Video Pitch (Max 3 mins): Sell your solution! Explain how your mathematical/computational solution brings "hope" to healthcare.
- 💻 Code Repo (Recommended): Link your GitHub/GitLab with reproducible code so we can review your magic.
*Judges will shortlist the Top 5 teams based on technical rigor, innovation, interpretability, and clinical relevance.
🔥 Stage 2: The Grand Final (Synchronous)
Date: 11 April 2026
The Top 5 teams will battle it out live!
- 🌐 Format: Live Pitch via Zoom/Webex.
- ⏱️ Duration: 10 mins presentation + 5 mins Q&A.
- 🎯 Focus: Prove your model's clinical applicability, safety, interpretability, reproducibility, and real-world impact.
📊 How to Win (Evaluation Criteria)
Wondering how the judges will score your project? Here is the exact breakdown of the 100-point rubric:
| Criterion | Weight (%) |
|---|---|
| 🧮 Mathematical/Algorithmic Rigor | 25 |
| 📈 Model Performance & Validation | 20 |
| 🧠 Problem Definition & Healthcare Relevance | 15 |
| 💡 Innovation & Creativity | 15 |
| 🛡️ Clinical Interpretability & Safety | 10 |
| 🌟 “Hope” Impact & Practical Implications | 10 |
| 🗣️ Presentation Quality & Clarity | 5 |
| Total Score | 100 |
⚖️ The Ground Rules
Play fair and code hard. Ensure your team follows these regulations to avoid disqualification:
- 🚫 No Plagiarism: Your project must be 100% original work created specifically for this challenge.
- 📂 Strict Dataset Compliance: You must exclusively use one (or more) of the official datasets from the PhysioNet list provided (bigP3BCI, Brugada-HUCA, Myocardial Perfusion SPECT, or HYGD). External labeled datasets are strictly not permitted.
- 🤖 Ethical AI Only: Your models need to adhere to strict ethical principles:
- Prioritize model interpretability and explainability (e.g., saliency maps, feature importance).
- Include safety validations and consider potential harms.
- Include an honest discussion of your model’s limitations, data leakage prevention, and failure modes.
- Plan for responsible deployment.
- 🔓 Open Source (Encouraged): While optional, we highly encourage sharing your code for transparency and reproducibility.
- ⏰ No Late Submissions: The Stage 1 deadline is absolute. Submissions received after the cutoff will not be reviewed.
- 🧑⚖️ Judges’ Word is Final: The decision of the judging panel cannot be contested.
- 🤝 Squad Loyalty: Your team composition must remain exactly the same from registration through the finals.
💻 Tech Specs & Tools
Bring your favorite stack! You have complete freedom over your tech, but here is what we recommend based on the datasets:
- Programming Languages: Python 🐍 (highly recommended), R, MATLAB, or Julia.
- Core Scientific Stack: NumPy, Pandas, SciPy, scikit-learn, PyTorch, TensorFlow, or Keras.
- Physiological Signal Tooling (EEG/ECG): *
WFDB(for ECG in WFDB format)MNE-Python(for EEG/ERP workflows)
- Medical Imaging Tooling (DICOM/NIfTI):
pydicom(for DICOM images)nibabel/SimpleITK(for NIfTI and medical image IO)MONAI(Deep learning framework for medical imaging - optional but powerful)
- Hardware & Environments: You are in charge of your own compute! Free cloud platforms like Google Colab or Kaggle Notebooks are perfect for initial prototyping. Deep learning on medical images (SPECT/Fundus) may benefit from GPU acceleration. Pro-tip: Use
requirements.txtorcondaenvironments to ensure your submitted code is easily reproducible!
📚 Official Learning Resources
Get familiar with the data before you start coding:
- The Platform: PhysioNet
- Dataset 1 (EEG): bigP3BCI Documentation
- Dataset 2 (ECG): Brugada-HUCA Documentation
- Dataset 3 (SPECT): Myocardial Perfusion SPECT Documentation
- Dataset 4 (Fundus): HYGD (Glaucoma) Documentation
Tip: Brush up on segmentation validation methodologies (Dice score, Hausdorff distance) if tackling imaging, and signal robustness best practices if analyzing EEG/ECG!