BrainSync
BrainSync focuses on neural rehabilitation, by leveraging an EEGNet model to predict motor intent using EEG signals. It includes a doctor panel on a website for tracking appointments and analyzing EEG
Created on 18th April 2025
•
BrainSync
BrainSync focuses on neural rehabilitation, by leveraging an EEGNet model to predict motor intent using EEG signals. It includes a doctor panel on a website for tracking appointments and analyzing EEG
The problem BrainSync solves
Non-Quantifiable Data: Doctors can’t see if a patient is truly engaging neural pathways [Dr. Steven C. Cramer, Stroke Engine Report].
Doctors and Clinicians: The portal simplifies appointment scheduling and EEG data analysis, enhancing efficiency and accuracy in treatment planning.
Patients with Neurological Conditions: NeuroRehab aids in recovering motor and cognitive functions with webcam-based exercises (e.g., Mirror Arms, Reach and Grab). NeuroRehab’s gamified interface (scores, badges, leaderboards) and multilingual support (English/Hindi) make rehabilitation engaging and accessible at home.
Reduced Hardware Costs: Hardware costs are reduced by using a 64-channel EEG electrode headset, instead of the 20k clinical rigs and by keeping EEG equipment with doctors and offering a software-based solution.
Challenges we ran into
EEGNet Model Overfitting: Training EEGNet overfitted quickly due to limited data diversity. We reduced model capacity by adjusting filters and layers, and removed noise, hence refining model for stability.
Class Imbalance for Rest State: The rest state dominated the dataset, skewing predictions. We used class weighting for minority classes, improving generalization.
Data Handling of EDF Time Series Data on Website: Processing large EDF files caused delays. We chunked data into manageable segments ensuring smooth portal functionality.
Python Dependencies Handling: Managing libraries (e.g., MNE, TensorFlow, Scikit-learn, PyEDFlib, OpenCV) faced version conflicts. We created a requirements.txt, used venv, and tested with pip check to resolve issues, ensuring a robust backend.
OpenCV Integration Challenges: In NeuroRehab, OpenCV struggled with real-time video processing, causing frame drops during MediaPipe Pose tracking. We optimized by reducing resolution, improving performance for exercises like Pattern Tracing.
Technologies used