Brain Imaging & Analysis Center

usage: -expr bandpass flag slowfreq fastfreq scalefact

The BANDPASS function will filter out both low and high frequency signals from a time series data set. Usually, frequencies slower than SLOWFREQ and faster than FASTFREQ will be filtered out. For convenience, frequencies are typically expressed as wavelengths in seconds, so a slow frequency cut-off would be something like 30-60s and a high frequency something like 2-3s, depending on what frequency wavelengths you want to preserve. If SLOWFREQ is faster than FASTFREQ, then BANDPASS acts as a notch filter to remove frequencies between SLOWFREQ and FASTFREQ (instead of frequencies outside that range).

Note: For temporal filtering, the data set time dimension should have the correct data sampling units (e.g. TR interval) specified; if the time unit is 0 (or negative) a 1s sampling interval is assumed.

SCALEFACT is a scaling factor; FFT output intensity values will be divided by SCALEFACT. Default SCALEFACT is 1000.

The FLAG argument can be used to select filtering options.
FLAG:

• 1 - frequencies are in Hertz (default: frequencies are seconds)
• 2 - 2-d spatial filtering [not implemented yet] (default is temporal filtering)
• 4 - frequencies are number of cycles across time course (default: seconds or Hertz)
• 8 - make output wordsize same as input (default: output is FLOAT32)
• 16 - filter out 0 frequency (default: 0 freq mean signal is preserved)
• 32 - no filtering (test FFT → reverse FFT)

Example:

   -expr bandpass 0 30 3            ; filter out frequencies below 30s (0.033 Hz) and above 3s (0.33 Hz) 
   -expr bandpass 1 0.033 0.33      ; same as above, specified in Hertz
   -expr bandpass 4 10 100          ; filter out frequencies below 10 cycles and above 100 cycles 
   -expr bandpass 0 10 40           ; filter out frequencies between 10s (0.1 Hz) and 40s (0.025 Hz)

See Also:
fScan Home, fScan Manual, CmdLineOptions, Expression Command, Manual Help