Digital image registration

As known from the previous chapter, the Anger camera shows the -events detected by the 2D position sensitive and energy selective detector on the display of the oscilloscope according to their real position, triggering the X, Y position signals by the Z signal.

The image of activity distribution of the examined object on the applied photo technical material is formed from the numbers of timed flashes on the display of the oscilloscope caused by γ-events. The purpose is to produce digital image simultaneously with the analogue image registration with the sampling of the analogue X, Y signals by the Z trigger signal (synchron signal) in order to further image analysis and evaluation done now digitally. The digital image is composed of elemental square shaped object due to the same order of digitization of X, Y signals. The elemental square shaped object is called pixel whose fineness depends at the degree of digitization (number of bits) (See Figure 1).

 
Hereinafter, we study how to match a digital image point, i.e. pixel, with the place of a -event. The digitalization procedure is possible to overview in Figure 2.. The Anger system camera produce the X, Y, Z signals for both its analogue visualization device and at the same time for another output for further processing as well.
First step in the digitizing procedure is the sampling of X,Y analogue signal (Sample & Hold S/H) and then the sampled signal will pass through analog-digital conversion (ADC) to transform the analogue information into digital code. If the applied ADC has an “N” bit order - resolution -, then the coordintes information signal by signal needs at least a 2N bit storage capacity register to be stored. In the case of FRAME command, the Interface Control Device regards the 2N-bit-long register as an address and executes the following operations with the special, so-called dedicated BUFFER MEMORY of the digitalizing unit:

I. BUFFER MEMORY := 0 /* RESET (i.e. set zero) the complete BUFFER MEMORY*/
II. <buff.mem.addr.> = <buff.mem.addr.> + 1 /*basic equation of MCA*/
III. Process
IV. Archiving the registered FRAMEs.

The used <AD> symbol means the content of the AD address. In this particular case, <buff.mem.addr.> is the content of the memory addressed by the 2N-bit-long register, i.e. the content of the memory addressed by coordinate signals. Consequently, if Xi, Yi coordinate more frequently the occurs, the more times the same memory address will be increased by “1”, in other words, the content of the addressed memory content becomes larger and larger. This is how the measured radioactive activity distribution of the material is mapped to content distribution of memory compartments by the fundamental equation of MCA (Multi Channel Analyzer). That is called FRAME imaging. The size of the constructed FRAME image matrix is (2^N*2^N).

 
By a control device, this image can be displayed even position-wisely on the screen of the image processor. As it can be also seen in the Figure 2, the size of the special BUFFER MEMORY serving the FRAME image collection cannot be smaller than 22N. If the memory is word-ordering, then the memory cannot contain less than 22N memory words. One word consists from at least two bytes, so it can be 3 or 4 bytes too. The other really important data which comes around apropos of digital image shooting and processing is so-called pixel size (PS) related to a given detector:

PS = D_DET / 2^N

where D_DET determining size parameter of a detector (e.g. diameter in case of circular detector) and N is the digitization order.

In the case of rectangular detectors, where there are usually different X_side and Y_side, (i.e. X_side≠Y_side) cosequently pixel sizes may be different in the direction of x and y using the same digitization order.

PS_x = X_side /2^N
PS_y = Y_side /2^N

Concerning digital image processing, there is a tight physical connection between the resolution of a detector and the wished pixel size.

The following questsion is raised: what is the optimal pixel size for a particular 2D position sensitive detector with a resolution described by a FWHM (full width half maximum)? According to the Nyquist-Shannon sampling theorem, the connection between optimal pixel size and detector resolution is the following:

PS_opt ≅ FWHM/3

Choosing a significantly larger pixel size, we will get a degraded resolution after digitization. Choosing a significantly smaller pixel size than the optimal, we will not be able to achieve better resolution than the FWHM, at best we get a picture seeming more pleasant if the noise makes it possible.

The other method for raw data registration is the LIST registration - i.e. registration of segment of sample function - method used mainly for research and unique purposes. In this case, the LIST control signal activates the Interface Control Device producing the following data structure in the BUFFER MEMORY:

Before the start of the data collection the whole area of the BUFFER MEMORY is initiated to zero.

Using this method, a list is created in the memory from the incident’s coordinates with time interrupting registration marked with " ", and with ECG interrupting registration if ECG gating is used (this is nothing more than an observation of a running process with noise during a predefined time i.e. a registration of sampled process). If the assigned BUFFER MEMORY is completely full, it is necessary to save “very quickly”, then set to zero the ssigned area and start the new data record acquiring. The drawback of the LIST acquisition mode is the large dedicated memory demand. On the other hand, its major advantage is the possibility of the related FRAME image construction based on various criteria by applying software MCA algorithm. Supposing one FRAME parametrization method did not satisfy the expectation, another framing can be started from the beginning as if nothing happened (from the LIST raw data collection).

According to Figure 2, both at LIST mode and at FRAME mode timing is done by the Z signal. The great benefit of the digital image registration arises from the opportunity to carry out picture part emphasizing, noise filtering (background subtraction, smoothing), enlargement, etc… with different processing algorithms after finishing the collection. If one of the image processing operations doesn’t provide the expected result, then another procedure can be started even from the beginning again. The greatest benefit of this technique is the following-up as well as quantitative evaluation of the dynamic of biochemical procedures. By means of ROI (Region Of Interest) application it is possible to designate the interested part of the image where the time activity function of applied organ selective radio-pharmaceutical is important. The time dependent biochemical procedure can be characterized by quantitative evaluation. Basically, the functional operation of various organs can be evaluated by different quantitative processing algorithms.

 
Hereinafter, we show the significance of the quantitative processing concerning nuclear medical imaging through a few examinations. In the Figure 3, a quantitative and qualitative evaluation of an ECG gated heart examination can be seen based on a model with a periodic approach. The Figure 4 shows a dynamic examination of kidney function during time. Figure 5 demonstrates the results of a bone metastasis examination which is important concerning oncology.

 

 

---