Glossary of Terms

The pixel pitch (d), is of great importance in dictating the limiting resolution of a digital system. According to sampling theory, only spatial frequencies below a certain level can be imaged accurately. This is known as the Nyquist frequency (Nv).

Frequencies higher than this will appear as a lower frequency component of the image. This is known as aliasing. When measuring high contrast spatial limiting resolution, aliasing can result in the dark and light bars being effectively reversed.
Theoretically, the first zero of the MTF is at twice the Nyquist frequency (assuming that the active pixel dimension is equal to the pixel pitch). If the signal between the Nyquist frequency and the first zero of the MTF is kept as low as possible, the amount of aliasing will be kept to a minimum.

The detector consists of a layer of a-Se with a matrix of electrodes on each face. The X-ray energy produces electron-hole pairs in the selenium layer, which are attracted towards the electrodes by an electric field. The charge is collected and readout using a TFT array.
The use of a-Se based photoconductor systems has the advantage that there is only one energy conversion while a phosphor requires both the conversion of X-ray energy to light and then light to charge. Therefore, a-Se based systems do not suffer from light spread prior to readout (figure 3). In common with a-Si based detectors, the limiting resolution of these detectors is limited by the effective pixel pitch.

Converts an analogue electrical signal to a discrete digital values so it can be processed in a digital system.

An ABC system control the x-ray tube output of a fluoroscopic system to maintain a constant image brightness at the image intensifier output.

An AEC for a film/screen system is set up by ensuring the correct optical density is achieved across a range of kVs. This method is not practical for digital imaging as the image will always be displayed optimally, irrespective of the exposure used. The AEC will need to be set up in collaboration with the radiology, medical physics departments and the supplier. The level of the exposure must be optimised and the receptor dose measured.
One other consideration is that sometimes when film/screen systems are replaced by a CR system then for simplicity the AEC is kept at the same settings, which allows film/screen still to be used. This may not be the optimal working level, in addition the CR plate will have a different response to change in the kV and so the AEC must be set up differently.

Is the increase in the mean energy of a x-ray beam as the lower energies are absorbed. Beam hardening by the use of filters in the x-ray tube reduce the lower energy x-rays that would be absorbed by the patient, thereby adding noting to the image but increasing the does to the patient.

This is useful for reducing the size of a data set into a manageable amount, and/or averaging data. A set of bins is defined as a set of evenly spaced ranges of values e.g. 2.5-3.5, 3.5-4.5 etc. Data is then collated into the appropriate bin depending on the value of the independent variable. Once all the data is collated into the bins, the dependent variable is the average of all the dependent variables in that bin. It can be a useful process to bin data for example in producing a edge spread function.

The smallest unit of computer memory, either on or off which equals one or zero. Bit is from the words binary digit.

See Image size and bit depth.

A device that moves a grid so as to blur the grid lines in an image.

The ratio of incident to transmitted radiation for a given grid (See also grid factor).

The smallest unit of computer storage which can represent a character. Byte is from the words binary term.

A CCD is an array of light sensitive elements which release electrons when light falls on them. The electrical signal produced is proportional to the intensity of the light.

CR is in first appearance similar to use as a film/screen system. The CR plate is in a cassette which will fit the table and vertical Bucky trays and can be used with a mobile unit on the wards. The plate is then scanned in a reading system similar in size to a daylight processor. This therefore makes the change to digital radiography easier. A DR system entails more changes in x-ray couch and vertical Bucky design and often changes to the x-ray tube assembly.
The X-ray photons interact with the image plate phosphor exciting electrons to a higher energy level. Although, some of these instantly de-excite and release light photons, some are stored within meta-stable energy traps formed by the Europium dopant where, due to thermal processes, they gradually de-excite. Therefore, it is important that the image plate is read as soon as possible post irradiation. The electron pattern in the phosphor makes up the image, as the number of excited electrons is directly proportional to the number of transmitted X-ray photons.

DQE is a frequency dependent measure of the efficiency of an imaging system. It is the square of the SNR of a detector compared to that of an ideal detector, assuming that the noise of an ideal detector is quantum limited. Therefore, DQE is a measure of the SNR squared in the image compared to that in the transmitted X-ray beam.
The image SNR squared of the system under test can be considered to be the NEQ. Quantum limited noise obeys Poisson statistics which states that the noise is proportional to the square root of the number of photons. Therefore, the SNR squared for the ideal detector becomes Q. The equation for DQE simplifies into the equation below:

Digital imaging devices have the potential for non-uniformity correction using a pixel dependent correction matrix calculated from acquisitions of nominally flat or uniform field exposures. The correction matrix is usually stored separately to the image data and applied during the image display process. The frequency and time required to acquire the calibration data varies considerably. Some systems may require calibration during service, whereas others may require daily calibration by the users, or the system may self calibrate.
Thermal noise produces a dark signal, which manifests as pixel values that are present without the application of X-rays. Some systems may use dark field calibration to correct for these processes. This involves the automatic acquisition of an image, without the application of X-rays, with a time equal to that of the actual diagnostic exposure. This is subtracted from the image data and does not slow the process to an unacceptable level.
On certain systems, the digital detector can be directly cooled to reduce noise. The cooling may be thermo-electric or by circulation of a cooling fluid. In the latter case a heat exchange device will have to be accommodated. A temperature interlock may be provided such that the system will only operate within the correct temperature range.

Digital Imaging and Communications in Medicine (DICOM) Version 3.0 is a protocol based standard to facilitate the transfer of digital images and associated information between devices manufactured by a range of vendors and to aid in the development of picture archive and communications systems (PACS).
When purchasing a system, a DICOM conformance statement should be received, this will inform how the device and software conform to the standard. It must be noted that a conformance statement indicates the particular sub set of DICOM conformance that is used. If a system does not produce images in DICOM format then a DICOM broker may be needed to convert the image to DICOM and allow the system to be connected to the PACS.

Digital imaging is used in a number of imaging techniques. The following table indicates what each is suitable for.

There are a number of technologies used for digital imaging in planar radiography. They can be divided into Computed Radiography (see CR) and Direct Digital Radiography (see DDR).

The main X-ray technologies used in DDR are:

• phosphor coupled to a read out device (amorphous-Silicon photodiode/TFT array or CCD)
• amorphous-Selenium/TFT in the form of a flat panel

This allows the subtraction of bone or soft tissue from an image, allowing soft tissue below bones to be easily visible. If an object containing only one type of material is imaged then the thickness at any point can be calculated providing the x-ray spectra, exposure and energy response of the detector is known. The body can be grouped into two sub groups of bone and soft tissue with similar x-ray absorption properties. If a body is imaged twice with different energy spectra, this can be undertaken at two kVp values or at the same kV with and without a metal filter, this then becomes a simultaneous equation, which can easily be solved for bone and soft tissue. There should be no movement between the images otherwise the images will need to be corrected, to reduce this problem, the images will need to be taken in rapid succession or simultaneously.

For flat panel detectors, a proportion of the detector contains the readout circuitry and will be insensitive to the incoming light photons or electrons. This leads to the concept of the fill factor (see equation below), which is the ratio of the sensitive area of the pixel to the effective area of the detector element itself.
Any improvements in resolution will require a reduced pixel pitch. The fill factor will decrease with improved resolution as the readout electronics will take up a larger proportion of the detector element and decrease the detector sensitivity.

Any signal can be produced by adding sine waves of different frequencies, amplitudes and phases. The Fourier transform splits the real image up into its component frequencies. Repeating the Fourier transform on the spatial frequency image returns it to the real space.

Low grid densities can cause interference patterns in the image called Moiré patterns. This can be solved by using moving grids or high density grids of over 60 lines per cm. When using CR, the grid lines must also be perpendicular to the scan lines in the reader.

The amount an exposure needs to be increased by when a grid is used to maintain the same image density (see also bucky factor).

Is the ratio of the height of the lead grid slats to the distance between the slats. The greater the grid ratio the more effective the grid is at removing scatter radiation from the beam.

Hospital Information Systems (HIS) and Radiology Information Systems (RIS) contain patient details and examination information. If the digital system and PACS is connected to a HIS/RIS this can speed up the workflow of a department by bringing the information to the department to give improved patient throughput. The patient details are then connected to the image, for easier reporting and image retrieval. The system uses HL7 (Health Level 7) language to transfer the patient details.

Compression of images is useful, the smaller an image is the more that can be stored and the faster the transmission. Compression can either be lossless or lossy. Reconstruction of the image from a lossless compression will be the same as the original image, while lossy will have some changes. Some system while lossy will claim to be “visually lossless”, i.e. there is no perceptible differences between the original and the reconstructed image. Compression ratio is defined as the ratio of the image size to the compressed image size such that a 4:1 has reduced the image size by a factor of 4.

Image quality is related to the radiation exposure received by the detector. Although a relatively low exposure will result in a noisy image, it may still contain sufficient information to be diagnostically acceptable. A high exposure will result in improved image quality since quantum noise is reduced. However, image quality improvement is not linear, it will eventually level off as the quantum noise becomes less dominant and decrease as the plate becomes overexposed. Ideally, a system should be set up to obtain adequate image quality for the lowest possible dose.

A CCD comprises a series of electrodes deposited on a semi-conductor substrate to form an array of capacitors. Voltages are applied to each of the electrodes to produce storage well locations. CCD technology requires an image phosphor to convert the X-rays into light photons which are read by the CCD camera. Photoelectric interactions within the semi-conductor result in the light photons being converted into electrical charge which resides in the charge storage wells. Voltage switching is used to transfer the charge from the storage wells to the read out circuitry.
For chest imaging systems, the dimensions of the CCD are typically smaller than the required field size. Either fibre optic bundles or lens/mirror systems are used to couple the image phosphor to the CCD and produce the demagnification required. The coupling efficiency decreases as the amount of demagnification increases, potentially increasing noise and thereby reducing SNR. More than one CCD can be used to cover the image phosphor. This is referred to as tiling of the detectors.

Ideally, the acquisition workstation should enable the following:-

• addition of anatomical markers
• demographics correction
• image annotation
• window and level adjustment
• electronic collimation
• magnification
• application of different LUTs.

The reporting workstation will be a high specification workstation with a comprehensive range of post-processing facilities, which include all of those listed previously, plus other functions including:-

• edge enhancement
• noise reduction
• tools for measuring pixel values, distances and angles
• zoom and roam.
  
  

A proportion of the signal from the previous image is added to the next image. Image retention should be kept low for radiographic imaging, while some fluoroscopy systems add lag to keep the noise low, this is suitable for images which are not changing rapidly.

The matrix size is the number of pixels or memory locations into which the image is divided. Thus the total number of pixels in a 1024 by 1024 matrix is 1,048,576, or approximately 1M.
Contrast resolution depends on the bit depth. The analogue value of the output from each pixel is converted to digital form, and the results stored at a separate location in a matrix. The number of grey levels available equals 2 to the power of the number of bits.

Number of grey levels per bit size

Clinical images require good contrast resolution, which can be difficult to achieve due to the noise inherent in a radiographic image. In order to achieve good contrast resolution, high bit depths are required.

Image stitching software is being developed by several manufacturers. Image stitching is designed to allow the visualisation of long areas of the body, such as the legs or spine. In CR the images are acquired by using several CR plates to cover the required anatomy and manufacturers are now introducing special holders where multiple plates are positioned vertically. In direct digital radiography the x-ray tube and detector are moved in step across the anatomy. Individual images are then electronically stitched together by the software.

See Spatial frequency

An LUT converts each pixel value into a new value. An LUT has two main uses:-

• a digital detector generally has a much wider dynamic range than the range of intensities in a clinical image, therefore an LUT is used to compress the data to cover only the clinically useful data
• an LUT may not be a linear relationship, but may be curved to enhance the contrast in the clinically useful densities. There may be a range of curves suitable for different clinical examinations.

The figure shows an example LUT used to compress a 12 bit image to a 10 bit image and apply a characteristic curve.

The MTF describes how well an imaging system transmits the spatial frequency information of the object being imaged. An ideal imaging system would transmit all spatial frequency information and would have an MTF of unity at all frequencies. The image would thus be a perfect representation of the object. Blurring or unsharpness introduced by the imaging system results in higher spatial frequencies not being transmitted as well as lower spatial frequency information. As a result, the MTF progressively decreases with increasing spatial frequency. There are a number of established methods for deriving the MTF of an imaging system. A common method is to calculate the MTF as the Fourier Transform of the Line Spread Function (LSF) or Point Spread Function (PSF).

The health protection agency (HPA) undertake surveys of doses appropriate to a particular type of radiograph or x-ray examination on patients of average size. The value is set at the level of the third quartile of the data set. This can be considered the dose level that action to reduce doses should be undertaken.

So far the digital technology has been discussed, however the great advantage of digital imaging is to be able to integrate it into a hospital wide system. Linking the digital x-ray into the HIS/RIS system and storing the images on a PACS system enables images to be reviewed at various reporting rooms, consultant’s offices, wards etc. Images can also be reported off-site using teleradiology.

The NEQ is a measure of the signal to noise ratio and can be considered as the effective number of X-ray quanta used in acquiring an image at a particular dose level, as a function of spatial frequency. NEQ is dependent upon MTF and the normalised Noise Power Spectrum (NNPS) of the imaging system (see equation below).

NPS – frequency dependant noise power spectrum,
MTF – frequency dependant modulation transfer function,
LAS – Large area signal, average pixel value in image.

The NPS is essentially a measure of the noise level over the spatial frequencies. It is normal to quote the normalised noise power spectrum (NNPS, equation below).

LAS – Large area signal, average pixel value in image.

See Aliasing and the Nyquist frequency

The X-ray detector is normally a thallium doped CsI phosphor, although other phosphors such as gadolinium oxysulphide (Gd2O2S) are also used. The X-rays interact with the phosphor and release light photons. Each type of phosphor will have advantages and disadvantages. For example, CsI is columnar in structure and acts as a light guide directing the emitted light photons towards the photodiode array and readout circuitry. Gd2O2S is not columnar, however, it does have the advantage that it releases more light per interacting X-ray than CsI.

DDR Phoshors

Image phosphor coupled to a-Si photodiode TFT array flat panel detector:
The light from the phosphor promotes charge in the photodiode array, with the amount of charge being proportional to the intensity of the incoming X-ray photons. Each TFT element is then sequentially addressed so that the charge in the photodiodes is readout and digitised.
The detector resolution is not only limited by the pixel pitch (distance between centres of pixels), but may also be reduced by the effect of light spread in the phosphor layer.
The light from the phosphor promotes charge in the photodiode array, with the amount of charge being proportional to the intensity of the incoming X-ray photons. Each TFT element is then sequentially addressed so that the charge in the photodiodes is readout and digitised.

• Acquisition and Reporting workstations
• Monitors
• Laser Printers

A pixel is the smallest element of a digitised picture. A smaller pixel size will generally give an improved spatial resolution in the image. The pixel pitch is the distance between the centres of adjacent pixels.

Posterisation effects occur when there are insufficient grey levels in the image to distinguish changes in density. These typically appear as bands with the same pixel value through the image and may be caused by having insufficient bits in an image or data being compressed by a Look Up Table.

A series of tests undertaken on a piece of equipment to ensure that it is functioning correctly.

See Acquisition and Reporting workstations.

CCDs are generally only up to 5 cm wide, this means that the image needs demagnifying and can result in loss of signal. An alternative method for covering the full image area is to use slot-scanning technology. A linear array of CCDs scans across the patient in conjunction with a narrow fan X-ray beam. This method may result in good scatter rejection and contrast differentiation but has a number of disadvantages including a long exposure time and high tube loading. Also, the alignment of the scanning radiation beam and the detectors requires tight mechanical tolerances and mechanical stability of the scanning mechanism.
This technology is generally used in mammography and chest imaging. Generally the detector is a phosphor connected to a linear array of CCDs.
The CCD is read using Time Delay Integration (TDI).

This defines the relationship between the pixel value and the exposure to the detector. There are different relationships possible, such as linear (with or without an offset) or logarithmic. The relationship needs to be known to linearise the image for tests such as NPS, MTF and uniformity.

Images are viewed in real space. However an image can be converted into frequency space using a Fourier transform. A spatial frequency is related to the size of objects in an image, where high frequency relates to rapid change (edges or noise) and low frequency relates to slow changes e.g. slight density differences. Spatial frequencies are useful for undertaking image processing and describing imaging properties of a digital detector.

A tiled array consists of a number of detectors abutted together to sample the whole image. However, there may be small areas on the read out device which are not sensitive due to the gaps between the detectors (typically about 100 microns).

During exposure, the X-ray fan-beam is continuously directed at the CCD detector. Charge is collected in each pixel of the CCD. The charge is transferred to the next pixel by electronic means. Each pixel accumulates charge for a set integration time before the charge is transferred to the next pixel. Concurrently, the detector moves in the opposite direction with the same velocity as the transfer of charge. This permits the charge packet to accumulate charge from a fixed point relative to the patient.