Time-frequency analysis of biomedical signals

In the last few years interest has been dedicated on the time-frequency analysis, especially in applications to biological signal processing. In fact, analysis in the frequency domain is a well standardized tool for the quantification of many clinical and physiological phenomena. The main limitation is constituted by the required stationarity of the signal in the analysis window, while a great number of biological situations are characterized instead by a dynamical evolution. In this paper the principal methods of time-frequency analysis are briefly described: the Cohen class of distributions, the wavelet transform and the recursive autoregressive estimation. Examples of application are shown on different clinical protocols: time-variant analysis of the heart rate variability signal during drug infusion, the evaluation of the dynamical power distribution in the electroencephalogram during surgery, the single sweep analysis of evoked potentials and the analysis of the magnetic resonance imaging spectroscopic signal. Advantages and disadvantages of the methods are highlighted, in particular we want to put into evidence how the different features of the presented procedures are able to enhance different characteristics of the signals.