Processing Techniques :Processing Techniques Group 2 Natasha Dehombre Daimy Perez Neysis Vidal Gustavo Thondik


OBJECTIVES :OBJECTIVES Discuss filters and there functions Define noise and its purpose Explain frequency Discuss Ramp filters/filtering Explain windows and there specific functions Discuss 3-dimensional imaging Define data display Explain the importance of quantitation


FILTERS :FILTERS Filters enhance image quality without changing the raw data Used to reduce noise in processing of images They produce results that are easier to read and process Downside to filters are that using them in excess for any given image (overfiltering) will reduce resolution, increases noise, and causes blurring to an image


NOISE :NOISE Noise is caused by background radiation, scatter radiation, and/or statistical fluctuations in measurements They show up on a image as Undesired additives or artifacts


The Image on the Right (B) Has More Noise Than the Image on the Left (A) :The Image on the Right (B) Has More Noise Than the Image on the Left (A)


Example of data filtered using cutoff value of 0.11. Images become too smooth, and spatial resolution is lost. :Example of data filtered using cutoff value of 0.11. Images become too smooth, and spatial resolution is lost.


Example of data filtered using cutoff value of 0.21. Images show good structural information within data and are not overfiltered. :Example of data filtered using cutoff value of 0.21. Images show good structural information within data and are not overfiltered.


FREQUENCY :FREQUENCY The conversion from spatial to frequency domain is performed by presenting the repetitions of pixels in the mathematical terms of sine and cosine functions. We can define frequency as the number of peaks per unit distance (pixel to pixel) or cycle per pixel or centimeter.


FREQUENCY :FREQUENCY Consider running a profile through a row of pixels that represent a portion of a 64x64 low count static background image. The result would reveal a systematic series of peaks and valleys representing random count deviations across that row of pixels The sharp peaks are indicative of a combination of high frequency noise and low frequency back ground


FREQUENCY :FREQUENCY The maximum number of peaks possible would be in a situation in which noise and background occur in every other pixel, representing a maximum of 32 separate peaks. So the greatest frequency possible for a 64x64 matrix is 32 or 32/64=0.5 Extending the frequency beyond this point would only lead to reflection.


FREQUENCY :FREQUENCY The point at which a filters transfer function begins to fold (becomes a mirror image of itself) is known as the Nyquist frequency The filters selected must exclude background and noise while retaining the useful data


RAMP FILTER :RAMP FILTER When a point source is projected into each pixel representative of the acquired angle a ramp filter suppresses the “star” effect. Without this filter, the complete collection of data including background and frequency noise would be projected as a ray, producing an abundance of artifacts Background noise which exists in low frequency range, would be included in this projection but would have a smoothing effect in the final reconstruction


RAMP FILTER :RAMP FILTER With this in mind, the filter design employed in filtered back-projection must not only suppress the radiating spokes of the target but exclude the background noise that will contribute to the blurring as well Filtering techniques of this ramp filtered data that enables the user to achieve the proper balance of noise exclusion and useful data retention


WINDOWS :WINDOWS A window function is used to improved the signal to noise (S/N) ratio in reconstruction A rectangular window may be tailored in a variety of ways to suit the overlap of useful data and the high-frequency noise The rectangular window accepts all image data while excluding noise Window functions have a variety of names: Hann Hamming Parzen


WINDOW CON’D :WINDOW CON’D One of the commercially available types of filter-window combinations is the Ramp-Hann Most commercial software offers the operator a choice of cutoff value selections for this type of filter-window combination. The higher the frequency, the sharper the image Reducing the cutoff frequency will increase smoothing and degrade resolution The higher the # of counts in a reconstructed image, the less smoothing a window function should have


WINDOW CON’D :WINDOW CON’D Butterworth filter and Metz filter are very design friendly Easily tailored to encompass that portion of data by the user These filters reduce artifacts They improve sharpness and retain resolution


THREE DIMENTIONAL FILTERING :THREE DIMENTIONAL FILTERING The original concept of SPECT reconstruction relating to oblique tomographic display, was stacking of multiple 2D transaxial imagesto present a 3D impression The standard Ramp-Hann, Hamming or Butterworth filter, performed during the reconstruction process, reduces noise using data in x and y directions. Planar data may be filtered in the y direction with a spatial filter that has the following coefficients:


THREE DIMENTIONAL FILTERING :THREE DIMENTIONAL FILTERING 0 1 0 0 2 0 0 1 0 After reconstruction of the transaxial data sets, using a Ramp-Hann filter-window with a 0.5 frequency cutoff. This 1:2:1 filter operates pixel by pixel over each planar view, regardless of the slice width and reconstruction limits


THREE DIMENTIONAL FILTERING :THREE DIMENTIONAL FILTERING Planar filtering of the raw projections with a 9-point smoothing spatial filter is another three dimensional filtering effect and has the following coefficients: 1 2 1 2 4 2 1 2 1 And then reconstructing the smoothed projection data with ramp filter. Transaxial images should be filtered in all 3 directions (x,y,z) to obtain maximum reduction in noise structure.


DATA DISPLAY :DATA DISPLAY Reconstruction – Process of creating trans-axial slices from projection views Implementation of oblique, saggital, or coronal algorithms takes little time compared with initial reconstruction of trans-axial data, display format of existing processed data Data in each pixel element have already undergo filtering and attenuation correction for reconstruction process


DATA DISPLAY :DATA DISPLAY Rearrangement of transaxial images as a cube with x and y planes representing the cross-sectional 64 x 64 matrix slices and the z slices in the transaxial file


DATA DISPLAY :DATA DISPLAY Oblique protocols may be very effective in obtaining transaxial views from the brain studies Thallium protocol is very effective in straightening rotated brain Imagines suspending a heart in a cube or box, in a position that may be equivalent of a typical study


DATA DISPLAY :DATA DISPLAY Imagines suspending a heart in a cube or box, in a position that may be equivalent of a typical study


QUANTITATION :QUANTITATION Quantitative measurement – Relative and Absolute Absolute Quantization– Quantization measurement of radioactive concentration (mCi) Relative Quantitation – Configuration of the attenuating medium stays the same for all measurement, compare with each other: Ratio = Image1 (x,y) x Att (x,y) Image2 (x,y) x Att (x,y)


QUANTITATION :QUANTITATION SPEC Tl-201- Myocardial blood flow with quantitation: Used to diagnose transient ischemia or scar tissue


QUANTITATION :QUANTITATION Another application of quantitative SPECT is the measurement of regional cerebral blood flow (rCBF) with inhalation of Xe-133


Question 1 :Question 1 Name two application of quantitative SPECT


Answer :Answer Myocardial blood flow with quantitation: Used to diagnose transient ischemia or scar tissue Measurement of regional cerebral blood flow (rCBF) with inhalation of Xe-133


Question 2 :Question 2 T/F Oblique protocols may be very effective in obtaining transaxial views from the brain studies


Answer :Answer True


Question 3 :Question 3 T/F Data in each pixel element have already undergo filtering and attenuation correction for reconstruction process


Answer :Answer True


Question 4 :Question 4 T/F The higher the # of counts in a reconstructed image, the less smoothing a window function should have.


Answer :Answer True


Question 5 :Question 5 Which 2 filters reduce artifacts, improve sharpness and retain resolution?


Answer :Answer Butterworth filter and Metz filter


Question 6 :Question 6 T/F Reducing the cutoff frequency will decrease smoothing and degrade resolution


Answer :Answer False


Question 7 :Question 7 What would happen to an image if a technician applied over filtering?


Answer :Answer Smoothing out of useful data Loss of resolution Increase noise


Question 8 :Question 8 Filtering alters the raw data to degrade image quality for interpretation. T/F


Answer :Answer False (enhances image quality)


Question 9 :Question 9 What can we define frequency as?


Answer :Answer number of peaks per unit distance


Question 10 :Question 10 The point at which a filters transfer function begins to fold (becomes a mirror image of itself) is known as the


Answer :Answer Nyquist frequency