Comments on SPECT, a Little on Filtering
- Processing the SPECT data
- SPECT
- Back projection technique
- Generates a 3-dimension from a series of 2-dimensional images.
- Reconstruction algorithms
- Photon from each pixel is translated to all other pixels perpendicular to the detector - array sum.
- Adjacent pixels are affected.
- Creates a star pattern on hot lesions or point source.
- Reduces blurring by subtracting counts.
- Not suggested for low count images
- False positive
- Back projection technique
- First pre-filter collected counts (convoluted kernel or filter function).
- Each project then represents a count profile
- Reduces star pattern
- Linear attenuation correction may be applied - CT attenuation correction is preferred
- Frequency - number of cycles per second (cps) in a electromagnetic spectrum
- Related to sine wave - peaks and valleys
- peaks represent high-frequency noise
- valleys represent low-frequency noise or background.
- Nyquist frequency - contains the highest frequency and highest resolution. An image with a higher frequency will not be correctly reproduced. (Greater the 0.X cycles/pixel the better the resolution)
- Nyquist is affected by low-count-rate; results in too much statistical fluctuation.
- Ramp filters rule out low-frequency bkg and accept high frequency noise. Acceptable in high count images because of the statistical insignificance.
- Hamming filters reject high frequency noise and accept low frequency bkg.
- The use of both filters is a compromise between both high and low frequencies.
The use of filters in order - low to high noise/resolution/smoothing
