Cloud computing for seizure detection in implanted neural devices.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computational Neuroscience

Background:

• Implantable neural stimulators offer a promising treatment for medically refractory epilepsy.
• Accurate seizure detection is crucial for effective therapeutic stimulation delivery by closed-loop devices.
• Current seizure detection algorithms are often too computationally intensive for direct deployment on implantable devices.

Purpose of the Study:

• To develop and validate a novel framework for implementing computationally intensive seizure detection algorithms on implantable neural stimulator platforms.
• To couple on-board device capabilities with cloud computing resources for enhanced seizure detection accuracy.
• To improve the performance and patient outcomes of next-generation neuro-implantable devices.

Main Methods:

• A sensitive on-board gating algorithm was used to identify potential seizure epochs.
• Candidate seizure epochs were transmitted to a cloud-based platform for analysis.
• A precise seizure detection algorithm was applied in the cloud, leveraging significant computational resources.
• The strategy was tested on human data from the NeuroVista Seizure Advisory System.

Main Results:

• The on-board gating algorithm achieved high-sensitivity detection using a computationally efficient linear classifier.
• The cloud-based precision algorithm successfully identified all transmitted seizures.
• The joint approach significantly reduced the false positive rate to 0.03 per hour.
• Overall seizure detection accuracy reached 99% across all subjects.

Conclusions:

• A novel framework effectively integrates implantable devices with cloud computing for complex algorithm execution.
• This approach enables next-generation neuro-implantable devices to utilize sophisticated seizure detection algorithms.
• The demonstrated strategy has the potential to greatly improve device performance and patient outcomes in epilepsy management.