Long-term ST database: A reference for the development and evaluation of automated ischaemia detectors and for the study of the dynamics of myocardial ischaemia
An analysis of the Long Term ST Database (LTSTDB) was conducted to quantify the spatio-temporal dynamics of ischemic and non-ischemic episodes. For all 86 recordings the ischemic episode length is decribed by a lognormal distribution and the non-ischemic episode length by a generalized extreme value distribution. For the 15 recordings that possess orthogonal (EASI) leads sets we derived the 12 standard leads and analyzed the spatial time course (from the j-point to j+120 ms) of each episode over time to identify dominant trends. Although the magnitude of the ischemic episodes did not reveal any inter-subject trend (except for generally exhibiting Brownian-like motion), there appeared to be strong correlations with the heart rate (HR). Wavelet cross-spectral coupling with significance testing was then applied to the ST-amplitude and HR evolution over the course of each episode. In all subjects significant cross-spectral correlations were found at very low frequencies (&#60;0.04 Hz), as well as at respiration and baroreflex frequencies. This may indicate that the ischemic episodes are modulated by blood pressure and activity or HR-related phenomena and that all episodes in the LTSTDB may be of a ‘mixed’ type at some point in their duration. The dominant angle also showed significant correlation (p&#60;0.01) with the ST amplitude and HR changes at similar frequencies to those described above. All three protocols used to define ischemia in the LTSTDB gave similar results.