Inspiration
Remote EEG monitoring is becoming increasingly important in epilepsy care, ICU monitoring, and tele-neurology. But we noticed a hidden problem: once EEG data travels over a network, it becomes hard to tell whether an unusual signal is a real brain event, a transmission error, or tampered data. That uncertainty inspired us to build TrialFlux, a system that adds a trust layer to brain-signal transmission.
What it does
TrialFlux is a smart EEG transmission and integrity-checking system. It compresses brain-signal data intelligently, simulates network failures, and detects whether an anomaly came from a real brain event, a transmission error, or tampered data. It also helps identify where the problem occurred in the pipeline.
How we built it
We built TrialFlux as a layered software pipeline. First, we load EEG data from a dataset or synthetic generator. Then we split the signal into windows and detect clinically important events based on deviation from the local baseline. After that, we compress the important parts with higher fidelity, attach integrity metadata, and link segments together using a hash-chain style structure. Finally, we simulate network corruption and run an audit layer that classifies each anomaly as biological, corrupted, or fabricated.
Challenges we ran into
One major challenge was deciding how to distinguish between a real brain event and a damaged transmission. Another challenge was keeping the project software-only while still making it realistic enough for medical use cases. We also had to balance technical depth with hackathon time limits, so we focused on a rule-based, explainable MVP instead of overcomplicating the model. A further challenge was designing a clean workflow that could be clearly explained to judges in a short demo.
Accomplishments that we're proud of
We are proud that TrialFlux goes beyond simple EEG classification and instead focuses on data trust and provenance. We built a pipeline that is fully software-based, explainable, and realistic for tele-neurology use cases. Another accomplishment is that the project can distinguish not just abnormal signals, but also the source of the problem, which makes it much more useful in real-world healthcare settings.
What we learned
While working on TrialFlux, we learned how EEG data is structured, how important channels and timestamps are for signal interpretation, and how clinical events can be detected using simple signal features such as deviation, variance, and spike shape. We also learned how cryptographic tools like hash chains and Merkle chains can be used outside of finance and blockchain to verify the integrity of medical data.
What's next for TrialFlux
Next, we want to improve the event detection logic, make the classification more accurate, and support more real EEG datasets. We also want to build a stronger frontend dashboard, add better visualization of the anomaly timeline, and make the integrity audit more robust against advanced tampering and replay attacks. In the future, TrialFlux could be extended for ICU monitoring, seizure detection systems, and secure brain-computer interfaces.
Built With
- browser-based
- chaining
- css
- csv
- dataset
- deployment
- edf
- edf+
- formats:
- hash
- hashing
- hmac
- integrity
- javascript
- language:-python-backend-framework:-fastapi-or-flask-data-processing:-numpy
- layer:
- local
- merkle-style
- or
- pandas
- platform:
- scipy-signal-processing-/-eeg-handling:-mne-python-machine-learning:-scikit-learn-visualization:-plotly-or-matplotlib-frontend:-html
- server
- streamlit
