COMPLEX SYSTEMS
Assessment of Condition Change from Nonlinear Time Series
We developed a robust and reliable model-independent technique for measuring condition change from nonlinear time series. After (low frequency) artifacts have been filtered out, the data is represented as a discrete distribution function on the reconstructed attractor. New metrics are introduced that evaluate the distance between distribution functions. The metrics are properly renormalized to provide robust and sensitive relative measures of condition change. After validating the technique on several analytic models, we applied it to clinical, single-channel, scalp (non-invasive) EEG signals to detect timely precursors of epileptic seizure onset. The choice of this first real-life application was dictated by: (i) the availability of sufficient reliable data; (ii) the characteristic time scale of the underlying dynamics; and (iii) the relevance of the problem. The results are consistently and significantly superior to the warning provided by applying traditional nonlinear measures to sub-dural (invasive) data.
This method has several significant advantages over earlier approaches. Indeed, previous investigations used data from multi-channel data from sub-dural and depth electrodes, while in our analysis we use only one channel of scalp EEG data. We also demonstrate the successful use of new measures of condition changes for many seizure types, over a variety of clinical conditions: digital and analog EEG from several clinical sites, data sampling at 200 and 512 Hz, raw EEG data precision between 10-12 bits, presence of substantial noise in the raw data, etc. Data manipulation requires comparatively modest processor means and can be done 60 times faster than real time on a standard desktop computer (in other words, one hour of data can be completely processed, displayed, and archived in less than a minute).
Due to its robustness, accuracy, speed, and ease of implementation, this method is suitable to provide non-invasive, long-term, ambulatory, and/or remote monitoring and accurate dynamics assessment for various medical, engineering, and geophysical applications. We filed an invention disclosure and a patent application for the method and the apparatus. The license for the specific EEG application and apparatus has been acquired by a medical equipment company that is now implementing it.
[