The proposed work offers enhancements to the MagicMarker software developed in Phase I. This software offers a new methodology for the display and analysis of long-term EEG records. Epochs of similar activity are grouped into segments and then states via two-pass hierarchial clustering. This results in clearly differentiated background, paroxysmal activity and patient state transitions. The underlying EEG is always available via hyper- links so that artifacts can be distinguished from """"""""real EEG."""""""" The Phase II work adds classification abilities (intelligence) to the Phase I software. Proposed are an expert-level seizure detector, an ICU abnormality detector and a user-defined activity detector. The effort includes the development of a large library of carefully analyzed and annotated prolonged EEG studies. Newborns, older children and adults will be included ensuring robust algorithms for all age groups. The proposed software greatly reduces staff requirements for long-term monitoring through intelligent notification (visual, audible and dial-up pager) of interesting events. This, in addition to the ability to monitor patients """"""""away from the lab,"""""""" provides more frequent patient checks and improved clinical outcomes.
The proposed software would be a valuable addition to any digital EEG because of the great timesavings it provides for both neurologists and EEG technicians. The software will be marketed along with current Persyst products (Insight, SpikeDetector and Prism), which support and are sold by virtually every major DEEG manufacturer.
