CASE STUDIES

99% Accuracy in Predicting Epilepsy with Machine Learning

Customer

A leading Healthcare Tech Company based in the USA

Problem Statement

The client was looking for a partner to Develop the algorithm for sensing human brain signals through a single bipolar electroencephalogram (EEG) to predict, detect, and monitor epileptic seizures and other neurological events.

SOLUTION

Used AI model on Android and iOS to detect seizure on the watch using technologies like NodeJS, HIPAA-compliant AWS cloud infrastructure
Used a supervised learning algorithm that uses a decision tree classifier approach
Developed Matlab software to train decision classifier using 1000+ hours of training data recorded from multiple patients
Developed Matlab software for real-time acquisition from DAQ and detection, dynamic learning, and offline review of data for doctors

BUSINESS IMPACT

It achieved 99% detection accuracy when trained for individual patients. It can operate in real-time, consuming a very low CPU load. Able to achieve detection and alert Leading US Medical University 2-3 minutes of seizure
Helped to alert the caregiver to provide help to the patient in time. Also, the data captured can be used for medicine tuning by doctors

SHARE THIS ARTICLE