Real-time seizure prediction informed by hidden markov model event states

An implantable medical device for predicting and treating electrical disturbances in tissue, the medical device including:

• an implantable telemetry unit (ITU) including a processor and an electrode circuit;
• and
• an implantable leads assembly (ILA) including a first electrode and a second electrode implantable in the tissue and electrically coupled to the electrode circuit of the ITU,
• wherein the processor of the ITU is configured to perform training by:
• receiving electrical signals input to the electrode circuit by the first and second electrodes,
• parsing the electrical signals into dynamic event states using Bayesian Non-Parametric Markov Switching, and
• modeling each event state as a multi-dimensional probability distribution,
• wherein the processor of the ITU is further configured to perform analysis of the electrical signals and therapy to the tissue by:
• receiving other electrical signals input to the electrode circuit by the first and second electrodes,
• applying the other electrical signals to the multi-dimensional distribution to predict future electrical disturbances in the tissue, and
• controlling the electrode circuit to apply an electrical therapy signal to the first and second electrodes to mitigate effects of the future electrical disturbances in the tissue.