Biomedical Engineering Reference
In-Depth Information
FIGURE 4.37:
Recurrence plots of the heartbeat interval time series at a control
time (a) and before a VT (b) with an embedding dimension of 6. The RP before
a life-threatening arrhythmia is characterized by big black rectangles, whereas RP
from the control series shows only small rectangles. From [Marwan et al., 2002].
4.2.2.4.5 Effectiveness of non-linear methods
Different non-linear methods
were compared in [Maestri et al., 2007]. The authors computed 11 non-linear HRV
indices found from: detrended fluctuation analysis, entropy estimators including
SaEn, empirical mode decomposition, and Poincare plots. The performance of these
measures was evaluated by comparison of results of the two 5-minute sessions con-
ducted on two consecutive days for a group of 42 healthy adults. The reliability of
measures was assessed by means of the test of no difference between two measure-
ments (paired
t
-test), and relative reliability was based on the interclass correlation
coefficient. The best results were obtained for Poincare plots, detrended fluctuation
analysis, and entropy measures; other indices performed rather poorly. These relia-
bility measures were based on short-term measurements and two data sets only, so
they have to be considered with care. In assessment of the utility of different methods
other measures such as: sensitivity and specificity in distinguishing different condi-
tions and robustness to noise have to be taken into account as well.
Although in principle, the above mentioned techniques have been shown to be
good tools for characterization of various complex systems, no major breakthrough
has yet been achieved by their application to biomedical signals including HRV anal-
ysis. No systematic study has been conducted by means of non-linear methods to
investigate large patient populations. Therefore they were not recommended for di-
agnostic applications by the Task Force of the European Society of Cardiology and
Search WWH ::
Custom Search