- Lack of infrastructure impacts diagnosis and monitoring of neuropsychological diseases
- Manual and tedious annotation done by certified EEG specialists
- Shortage of qualified neurologists and certified EEG specialists
- Lack of validated EEG database in existing commercial software for spike detection
- Restricted in-situ access to EEG data
- Prevalence: 55-60 million people worldwide
- 80% of patients from low and middle income countries
- Epilepsy-related cost – US$2.6billion/year (USA)
Traumatic Brain Injury
- 10M people/year worldwide (2.5M in US, 1M in EU)
- Inadequately prepared health systems to address the associate health outcomes
- NeuroBrowser (NB), a cloud-based software system addresses the clinical needs associated with EEG interpretation, including (1) remote access, (2) time savings, (3) automation, (4) statistical validation, and (5) clinical validation.
- Current database of thirty-five thousand spike signatures from 100 epileptic patients is the world’s largest database of expert annotated epileptic spikes
- Used as training set for the algorithms.
- Classifier development based on the concept of classifier ensembles and cascades: best for dealing with extreme pattern variability while providing exceptional computational efficiency
- The team has developed a combination of algorithms (machine learning and neural networks) to automate EEG interpretation.
- The system was trained and tested by applying a 4-fold cross-validation on an annotated database
- The team is finalising a larger database with Mass General Hospital, Boston, of greater than ninety thousand spikes from over one thousand patients
- Current IP position: US National Phase patent application filed. Signal processing know how and trade secrets
- Clinical validation will begin Q1 2018 and scheduled to be completed in Q2 2018
- The team intends to spin out a company to bring the product to market by Q2 2018. Investments and/or partnerships are welcome.
Dr Rahul Rathakrishnan / National University Health Systems