Detection of event-related fMRI signal with matched filters

Introduction A common problem in biomedical signal processing is signal detection in noisy data. If approximate signal features such as shape and duration are known, the signal’s presence and position in time can be estimated by pattern recognition techniques. A matched filter approach, already shown to aide identification of physiological signals embedded within white [l] or colored 121 noise may have utility in extracting eventrelated fMR1 signals. The matched filter technique combines waveform pattern recognition with crosscorrelation analysis.