The Performance of Information-Theoretic Criteria in Detecting the Number of Independent Signals in Multilead ECGs

Abstract:

Three different methods to detect the number of independent signals in multilead ECGs were evaluated by using different ECG realizations with known specifications for signal and noise: a threshold method (TM), the minimum description length (MDL) and Akaike’s information criterion (AIC).

The fundamental assumption in this kind of signal processing is that both the signal and the noise stem from independent stochastic generators. The consequence is that the detection of the number of signals is only possible if the noise is white, or if the noise properties have been specified.

The evaluation was performed with respect to the QRS complex of individual multilead ECG simulations and to the entire ensemble. In the simulated ECGs the number of independent signals was fixed (eight). It was found that, out of the three methods studied, the performance of MDL was the best, especially when the number of available observations in the noise (used to estimate the noise specifications) was moderate.

Key-Words

ECG - Information Theory - Signal Processing

• REFERENCES

• 1 Ahmed N, Rao KR. Orthogonal Transforms for Digital Signal Processing. Berlin: Springer Verlag; 1975
• 2 Klema VC, Laub AJ. The singular value decomposition: its computation and some applications. IEEE Trans Autom Control 1980; 25: 164-76.
• 3 Damen AA, Van der Karn J. The use of the singular value decomposition in electrocardiography. Med Biol Eng Comput 1982; 20: 473-82.
• 4 Lux RL, Burgess MJ, Wyatt RF, Evans AK, Vincent GM, Abildskov JA. Clinically practical lead systems for improved electrocardiography: comparison with precordial grids and conventional lead systems. Circulation 1979; 59: 356-63.
• 5 Evans AK, Lux RL, Burgess MJ, Wyatt RF, Abildskov JA. Redundancy reduction for improved display and analysis of body surface potential maps II: Temporal compression. Circ Res 1981; 49: 197-203.
• 6 Uijen GJH, Heringa A, van Oosterom A. Data reduction of body surface potentials by means of orthogonal expansions. IEEE Trans on Biomed Eng 1984; 31: 706-14.
• 7 Uijen GJH, van Oosterom A. On the detection of the number of signals in multilead ECGs. Meth Inform Med 1992; 31: 246-54.
• 8 Heringa A, Uijen GJH, van Dam RTh. A 64-channel system for body surface potential mapping. In: Electrocardiology ‘81. Anlozcy Z, Preda I. (eds). Budapest: Akademiai Kiado; 1982: 297-301.
• 9 Wax M, Kailath T. Detection of signals by information theoretic criteria. IEEE Trans Acoust Speech Signal Proc 1985; 33: 387-92.
• 10 Anderson TW. Asymptotic theory for principal component analysis. Ann Math Stat 1963; 34: 122-48.
• 11 Zhao LC, Krishnaiah PR, Bai ZD. On detection of the number of signals in presence of white noise. J Multivar Anal 1986; 20: 1-25.
• 12 Zhao LC, Krishnaiah PR, Bai ZD. On detection of the number of signals when the noise covariance matrix is arbitrary. J Multivar Anal 1986; 20: 26-49.
• 13 Jenkins GM, Watts DG. Spectral Analysis and its Applications. San Francisco: Hol-den-Day; 1969
• 14 Zhao LC, Krishnaiah PR, Bai ZD. Remarks in certain criteria for detection of number of signals. IEEE Trans Acoust Speech Signal Proc 1987; 35: 129-32.