HippoScreen

Solution

HippoScreen's Stress EEG Assessment (SEA) System combines an EEG amplifier, a clinician-facing GUI, and an AI analytics stack to turn 90-second brainwave recordings into an objective depression-risk score. The amplifier collects and preprocesses EEG signals, while the GUI controls the test workflow and presents results to medical staff. In the compute layer, HippoScreen mixes proprietary algorithms with scikit-learn machine-learning methods for feature extraction, feature selection, clustering, classification, and regression, plus PyTorch deep-learning models such as SCCNet, EEGNet, and Shallow ConvNet. Intel VTune Profiler was used to inspect CPU and software thread behavior across Python environments and identify thread oversubscription; HippoScreen then tuned execution to a five-thread sweet spot. Intel Extension for Scikit-learn and Intel Optimization for PyTorch accelerated model build and validation cycles, helping the team test parameter combinations faster and shorten delivery of diagnostic output. The page names no storage product, but it does cite a large cross-center EEG database supporting the broader solution.

Data flow

Patients' 90-second EEG brainwave signals are captured by the EEG amplifier, processed inside the SEA System, and passed through preprocessing, feature extraction, feature selection, and classifier stages implemented with proprietary algorithms, scikit-learn, and PyTorch models. The system then produces a numerical evaluation index representing depression probability for doctors through the GUI.