Abstract
Hidden state transitions are frequent events in complex biological systems like the brain. Accurately detecting these transitions from sequential measurements (e.g., EEG, MER, EMG, etc.) is pivotal in several applications at the interface between engineering and medicine, like neural prosthetics, brain-computer interface, and drug delivery, but the detection methodologies developed thus far generally suffer from a lack of robustness. We recently addressed this problem by developing a Bayesian detection paradigm that combines optimal control and Markov processes. The neural activity is described as a stochastic process generated by a Hidden Markov Model (HMM) and the detection policy minimizes a loss function of both probability of false positives and accuracy (i.e., lag between estimated and actual transition time). The policy results in a time-varying threshold that applies to the a posteriori Bayesian probability of state transition and automatically adapts to each newly acquired measurement, based on the evolution of the HMM and the relative loss for false positives and accuracy. An application of the proposed paradigm to the automatic online detection of seizures in drug-resistant epilepsy subjects is here reported.
Original language | English (US) |
---|---|
Title of host publication | A Systems Theoretic Approach to Systems and Synthetic Biology I |
Subtitle of host publication | Models and System Characterizations |
Publisher | Springer Netherlands |
Pages | 153-178 |
Number of pages | 26 |
ISBN (Electronic) | 9789401790413 |
ISBN (Print) | 940179040X, 9789401790406 |
DOIs | |
State | Published - Mar 1 2014 |
Keywords
- Bayesian estimation
- Drug-resistant epilepsy
- Hidden Markov model (HMM)
- Intracranial electroencephalogram (iEEG)
- Optimal control
- Seizure onset detection
ASJC Scopus subject areas
- General Medicine