Temporal pattern mining for multivariate clinical decision support.

Multivariate temporal data are collections of contiguous data values that reflect complex temporal changes over a given duration. Technological advances have resulted in significant amounts of such data in high-throughput disciplines, including EEG and iEEG data for effective and efficient healthcare informatics, and decision support. Most data analytics and data-mining algorithms are effective in capturing global trends, but fail to capture localized behavioral changes in large temporal data sets. We present a two-step algorithmic methodology to uncover temporal patterns and exploiting them for an efficient and accurate decision support system. This methodology aids the discovery of previously unknown, nontrivial, and potentially useful temporal patterns for enhanced patient-specific clinical decision support with high degrees of sensitivity and specificity. Classification results on multivariate time series iEEG data for epileptic seizure detection also demonstrate the efficacy and accuracy of the technique to uncover interesting and effective domain class-specific temporal patterns.