Computed Radiography

Draft Document: AAPM Task Group 10 October 1997; version 3.1

Page 1

1. INTRODUCTION 1 2. SYSTEM OVERVIEW 2

A. A. Photostimulable phosphors and photostimulated luminescence 3 B. PSP characteristics and image formation properties 4 C. C. The readout process 5 D. D. Characteristic curve response 6

3. PROCESSING THE RAW PSP IMAGE 7 A. Readout parameters 8 B. Image grayscale processing 9 C. Display processing 10

4. IMAGE DEMOGRAPHICS AND EXPOSURE INDICATORS 11 A. Image acquisition and processing parameters 12 B. Exposure indicators 13 C. Exposure concerns of PSP systems 14

5. PSP SYSTEM IMAGE CHARACTERISTICS 15 A. Spatial resolution 16 B. Contrast resolution 17 C. Detective Quantum Efficiency 18 D. Image display 19

6. SYSTEM CONFIGURATIONS AND DIGITAL SOFT-COPY INTERFACES 20 7. GENERIC FUNCTIONAL SPECIFICATIONS OF PSP SYSTEMS 21

A. Phosphor plates and cassettes 22 B. Plate throughput 23 C. Spatial resolution 24 D. Contrast sensitivity 25 E. Dynamic range 26 F. Miscellaneous considerations for bid specifications 27 G. Clinical implementation issues 28

7. ACCEPTANCE TESTING 29 A. Preliminary communication with vendor engineer/specialist 30 B. Component inventory 31 C. Initial adjustments 32 D. Specific testing procedures 33

8. ARTIFACTS 34 A. Image artifacts 35 B. Software artifacts 36 C. Object artifacts 37 D. Film artifacts 38

9. QUALITY CONTROL AND PERIODIC MAINTENANCE 39 A. Daily (technologist) 40 B. Weekly (technologist) 41 C. Monthly (technologist) 42 D. Semi-Annually / Annually (physicist) 43

10. CONCLUSIONS 44 11. REFERENCES 45

46


Page 2

Acceptance Testing and Quality Control of 1 Photo Stimulable Phosphor Imaging Systems 2

Report of Task Group #10 3 American Association of Physicists in Medicine 4

5 Task Group #10 Members: J. Anthony Seibert (Chair), Terri Bogucki, Ted Ciona, Jon Dugan, Walter 6 Huda, Andrew Karellas, John Mercier, Ehsan Samei, Jeff Shepard, Brent Stewart, Orhan Suleiman, Doug Tucker, 7 Robert A. Uzenoff, John Weiser, Chuck Willis 8 9

Table of Contents 10 Page # 11

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 System overview 14 Photostimulable phosphors and photostimulated luminescence 15 PSP characteristics and image formation properties 16 The readout process 17 Characteristic Curve Response 18 Processing the raw PSP image 19 Readout parameters 20 Image grayscale processing 21 Display processing 22 Image demographics and exposure indicators 23 Image acquisition and processing parameters 24 Exposure indicators 25 Exposure concerns of PSP systems 26 PSP system image characteristics 27 Spatial resolution 28 Contrast resolution 29 Detective Quantum Efficiency 30 Image display 31 System configurations and digital soft-copy interfaces 32 Generic Functional Specifications of PSP systems 33 Phosphor plates and cassettes 34 Plate throughput 35 Spatial resolution 36 Contrast sensitivity 37 Dynamic range 38 Miscellaneous considerations for bid specifications 39 Clinical implementation issues 40 Acceptance Testing 41 Preliminary communication with vendor engineer and specialist 42 Initial adjustments, tools and equipment 43 Component inventory 44 Specific testing procedures -- tools and equipment 45 Artifacts 46 47 Quality Control and Periodic Maintenance . . . . . . . . . . . . . . . . . . . . . . . 48


Page 3

1 Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 4 Bibliography. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 6 Appendices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1. Manufacturers: addresses, contact information 8 2. Specific system details and testing procedures: Fuji Photo Film, Inc. and related PSP systems 9 3. Specific system details and testing procedures: Eastman Kodak Digital Science 10 4. Specific system details and testing procedures: Agfa 11 5. Sample acceptance testing and quality control forms 12

13


Page 4

ABSTRACT 1

2

Photostimulable phosphor (PSP) imaging employs reusable imaging plates and associated hardware and software 3

to acquire and display digital projection radiographs. This is a recent addition to diagnostic imaging technology. 4

Procedures are needed to guide the diagnostic radiological physicist in the evaluation and continuous quality 5

improvement of PSP imaging practice. This document includes overview material, generic functional 6

specifications, testing methodology, and a bibliography. We describe generic, non-invasive tests that are 7

applicable to a variety of PSP units. Manufacturers appendices describe specifications, machine-specific 8

attributes, and tests. 9

10

INTRODUCTION 11

12

The primary purpose of this document is to guide the clinical medical physicist in the acceptance testing of 13

photostimulable phosphor (PSP) imaging systems. PSP imaging devices are known by a number of names 14

including, computed radiography (CR), storage phosphor imaging, digital storage phosphor imaging, and digital 15

luminescence radiography. In the digital form, PSP images are readily integrated into a Picture Archiving and 16

Communications System (PACS). The tests we describe are appropriate for PSP systems in either integrated or 17

stand-alone applications. Digital imaging technology is rapidly evolving: this guide represents the state of 18

technology as of its writing. Proper application of this guide involves supplementing with current literature and 19

specific manufacturers technical data. A secondary purpose is to provide a consolidated source of information 20

regarding device functionality, testing, and clinical practice of PSP imaging. This document provides the physicist 21

with a means to conduct initial acceptance testing, interpret results, and establish baseline performance. A subset 22

of these tests can be extended to routine quality control. 23

24

25

SYSTEM OVERVIEW 26

In order to test an imaging device, an understanding of its basic operating principles is necessary. The following 27

text provides a basic discussion of those principles. 28

29

PSP Image acquisition 30

The photo-stimulable phosphor (PSP) stores absorbed x-ray energy in crystal structure traps, and is sometimes 31

referred to as a "storage" phosphor. This trapped energy can be released if stimulated by additional light energy of 32

the proper wavelength by the process of photostimulated luminescence (PSL). Acquisition and display of the PSP 33

image can be considered in five generalized steps illustrated in Figure 1 below. 34

35

36


Page 5

PSP Image acquisition

X-ray system

Patient

PSP detector

ComputedRadiograph

1.

ImageReader

2.Image

Scaling

3.Image

Recorder

4.

unexposed

exposed

5.

1

Figure 1. PSP Image acquisition and processing. 2

3

4

The unexposed PSP detector, placed in a cassette, replaces the screen-film receptor. Using x-ray imaging 5

techniques similar to screen-film imaging, an electronic latent image, in the form of trapped electrons is 6

imprinted on the PSP receptor by absorption of the photons transmitted through the object. At this point, the 7

unobservable latent image is processed by placing the PSP cassette into an image reader, where the image 8

receptor is extracted from the cassette and raster-scanned with a highly focused laser light of low energy. A higher 9

energy photostimulated luminescence (PSL) signal is emitted, the intensity of which is proportional to the number 10

of x-ray photons that were absorbed in the local area of the receptor. The PSL signal is channeled to a 11

photomultiplier tube, converted to a voltage, digitized with an analog to digital converter, and stored in a digital 12

image matrix. After PSP detector is totally scanned, analysis of the raw digital data locates the pertinent areas of 13

the useful image. Scaling of the data with well-defined computer algorithms creates a grayscale image that mimics 14

the analog film image. Finally, the image is recorded on film, or viewed on a digital image monitor. In terms of 15

acquisition, the PSP system closely emulates the conventional screen-film detector paradigm. As this report will 16

detail, however, there are also several important differences and issues that the user must understand and be aware 17

of to take full advantage of PSP imaging capabilities. 18

19

PSP characteristics and image formation properties 20

PSP devices are based on the principle of photostimulated luminescence [Takahashi, et al. 1983; Takahashi, 1984, deLeeuw 21 et al, 1987, and vonSeggern, et al, 1988]. When an x-ray photon deposits energy in the PSP material, the energy can be released 22

by three different physical processes. Fluorescence is the prompt release of energy in the form of light. This 23

process is the basis of conventional radiographic intensification screens. PSP imaging plates also emit 24

fluorescence in sufficient quantity to expose conventional radiographic film [Chotas, 91, Mc Mahon 91], however this is not 25

the intended method of imaging. PSP materials store some of the deposited energy in defects in their crystal 26

structure, thus they are sometimes called storage phosphors. This stored energy constitutes the latent image. Over 27


Page 6

time, the latent image fades spontaneously by the process of phosphorescence. If stimulated to light of the proper 1

wavelength, the process of stimulated luminescence can release the trapped energy. The emitted light constitutes 2

the signal for creating the digital image [Sonoda 83]. 3

4

PSP receptor characteristics. Many compounds possess the property of PSL [REF]. Few of these materials 5

have characteristics desirable for radiography, i.e. a stimulation-absorption peak at a wavelength produced by 6

common lasers, a stimulated emission peak readily absorbed by common photomultiplier tube input phosphors, and 7

retention of the latent image without significant signal loss due to phosphorescence [Luckey, 1975]. The compounds 8

that most closely meet these requirements are alkali-earth halides. Commercial products have been introduced 9

based on RbCl, BaFBr:Eu2+, BaF(BrI):Eu2+, BaSrFBr:Eu2+. A cross-section of the PSP receptor is illustrated in 10

Figure __ [Willis, in press] 11

12

13

14

15

16

17

18

19

20

21

Figure ___. Cross sectional views of the Fuji (left) and Kodak (right) PSP receptors are shown, indicating 22 the various structures comprising the receptor and cassette holder, and exemplifying differences that exist from 23 each manufacturer. Adapted from Willis[]. 24

25 Doping. Trace amounts of impurities, such as Eu2+, are added the PSP to alter its structure and physical 26

properties. The trace impurity is also called an activator. Eu2+ replaces the alkali earth in the crystal, forming a 27

luminescence center. 28

29

Absorption Process. Ionization by absorption of x-rays (or UV radiation) forms electron/hole pairs in the 30

PSP crystal. An electron/hole pair raises Eu2+ to an excited state, Eu3+. Eu3+ produces visible light when it returns 31

to the ground state, Eu2+. Stored energy (in the form of trapped electrons) forms the latent image. There are 32

currently two major theories for the PSP mechanism a bimolecular recombination model [Takahashi 83], and a 33

photostimulable luminescence complex (PSLC) model [vonSeggern, 87] to explain the energy absorption process and 34

subsequent formation of luminescence centers. Physical processes occurring in BaFBr:Eu2+ using the latter theory 35

appears to closely approximate the experimental findings. In this model, the PSLC is a metastable complex at 36

higher energy (F-center) in close proximity to an Eu3+-Eu2+ recombination center. X-rays absorbed in the PSP 37


Page 7

induce the formation of holes and electrons, which either activate an inactive PSLC by being captured by an 1

F-center, or form an active PSLC via formation and recombination of exitons explained by F-center physics 2 [vonSeggern, 87]. In either situation, the number of active PSLCs created (number of electrons trapped in the 3

metastable site) are proportional to the x-ray dose to the phosphor, critical to the success of the phosphor as an 4

image receptor. 5

6

Photostimulable Luminescence Complex (PSLC)

BaFBr:Eu 2+

Laser stimulation

2.0 eV

Conduction band

Valence band

relax

PSL3.0 eV

tunnelingrecombination

Eu

Eu2+Eu 3+/

F F +/4f 5d6

4f 7

phonon

8.3 eV

Electrons fill PSLC complexes (F centers) in numbers proportional to incident x-ray intensity

incident x-rays

e -

Photostimulated Luminescence:

BaFBr:Eu 2+

2 eV

Conduction band

Valence band

8.3 eV

>6 eV FF +

Eu Eu2+ 3+

Excitation process

PSL process

excitation energy

stimulation energy

3 eV

PSL energy

7

Figure --. An energy diagram of the excitation and photo-stimulated luminescence processes in a BaFBr:Eu2+ 8 phosphor. On the left is the representation of the interactions proposed by von Seggern, etal []. On the right is the 9 proposed energy diagram of Takahashi, etal [] Incident x-rays form an electron latent image in a meta-stable 10 F center site that can be processed with a low energy laser beam, producing the desired luminescent signals. is 11 the decay constant of the indicated process above. 12

13 X-ray absorption efficiency of BaFBr:Eu is compared to Gd2O2S:Tb (rare-earth screens) for typical 14

thicknesses of material encountered, as shown by attenuation curves illustrated in Figure --. Between ~35 to ~50 15

keV, the BaFBr phosphor is actually a better x-ray attenuator due to the lower K-edge absorption of barium; 16

however, below and above this range, the gadolinium rare-earth phosphor is superior. A typical beam spectrum 17

incident on the PSP phosphor often requires greater exposure to achieve similar quantum statistics compared to a 18

400 speed rare-earth receptor. In addition, high absorption probability of x-rays below the k-edge of the PSP 19

receptor, where a significant fraction of lower energy scattered x-ray distribution occurs, causes a greater 20

sensitivity to scatter (thus reference to the PSP as a scatter sponge in this context). 21

22


Page 8

X-ray Absorption Efficiency

0 20 40 60 80 100 120 1400

0.2

0.4

0.6

0.8

1

Energy (keV)

Phot

on a

bsor

ptio

n fr

actio

n

Gd2O2S, 120 mg/cm

BaFBr, 100 mg/cm

BaFBr, 50 mg/cm

1

Figure--. This plot compares the absorption efficiency of PSP and rare-earth x-ray phosphors as a function of 2 energy. The thicknesses are representative of a standard 400 speed conventional screen, a standard resolution 3 PSP phosphor plate (100 mg/cm2), and a high resolution PSP phosphor plate (50 mg/cm2). 4

5

Fading. Fading of the trapped signal will occur exponentially over time, through spontaneous 6

phosphorescence. A typical imaging plate will lose about 25% of the stored signal between 10 minutes to 8 hours 7

after an exposure, and more slowly afterwards [Kato, 94]. Fading introduces uncertainties in output signal that can be 8

controlled by introducing a fixed delay between exposure and readout [ref??] to allow decay of the prompt 9

phosphorescence of the stored signal. 10

11

Stimulation and Emission. The electronic latent image imprinted on the exposed BaFBr:Eu phosphor 12

corresponds to the activated PLSCs (F-centers), whose local population of electrons is directly proportional to the 13

incident x-ray flux for a wide range of exposures, typically exceeding 10,000 to 1 (four orders of exposure 14

magnitude). Stimulation of the Eu3+- F-center complex and release of the stored electrons requires a minimum 15

energy of ~2eV, most easily deposited by a highly focused laser light source of a given wavelength. Lasers 16

produced by HeNe (=633 nm) and diode (680 nm) sources are most often used. The incident laser energy 17

excites electrons in the local F-center sites of the phosphor. According to von Seggern [vonSeggern, 87], two subsequent 18

energy pathways within the phosphor matrix are possibleto return to the F-center site without escape, or to 19

tunnel to an adjacent Eu3+ complex. The latter event is more probable, where the electron cascades to an 20

intermediate energy state with the release of a non-light emitting phonon. A light photon of 3 eV energy 21

immediately follows as the electron continues to drop through the electron orbitals of the Eu3+ complex to the more 22

stable Eu2+ energy level. Figure ___ shows a plot of the energy spectra of the laser-induced electron stimulation 23

and subsequent light emission. Note that different phosphor formulations will impact the stimulation energies; 24

thus it is important for optimal results that the PSP receptors be matched with the energy of the stimulating laser 25

source. 26

27


Page 9

Stimulation Emission

300400500600700800

Rel

ativ

e in

tens

ity

0.0

0.5

1.0

, nm

HeNeLaser

633 nm

Energy, eV3 42.521.5

Stimulation and Emission Spectra

1.75

DiodeLaser

680 nm

BaF(Br,I):Eu+

BaFBr:Eu+

1 2

Figure __. Stimulation and emission spectra for BaFBr:Eu 2+ and BaFBr0.85I0.15:Eu 2+ storage phosphors 3 demonstrate the energy sensitivity of different phosphor formulations and the energy separation of the excitation 4 and emission events. Selective optical filtration isolates the light emission intensity from the incident laser 5 intensity. In absolute terms, intensity of the emitted light is significantly lower. (Figure adapted from reference 6 [vonSeggern, 87]) 7 8

The readout process 9

Laser Scanning. Produced by either a HeNe or diode laser source, the laser beam is routed through 10

several optical components prior to scanning the phosphor plate. First, a beam splitter uses a portion of the laser 11

output to monitor and compensate for intensity fluctuations through the use of a reference detector. This is 12

important, as the intensity of the stimulated light is dependent on the power of the stimulating laser [Bogucki, 95]. The 13

major portion of the laser energy reflects off scanning mirror (rotating polygonal or oscillating flat reflector), 14

through an optical filter, shutter, and lens assembly, providing a synchronized scanning beam. To maintain a 15

constant focus and linear sweeping velocity across the PSP plate, the beam passes through an f- lens to a 16

stationary mirror (typically a cylindrical and flat mirror combination). The laser spot distribution on the phosphor 17

is adjusted to have a gaussian profile with a 1/e2 diameter of approximately 100 m in most reader systems. 18

Simplified system architecture of the PSP reader components is illustrated in Figure ___. 19


Page 10

PMTBeam deflector

LaserSource

Light channeling guide

Plate translation: Sub-scan direction

Laser beam: Scan direction

Output Signal

Reference detector

Beam splitter

Cylindrical mirrorf-thetalens

Amplifier

ADC

To Image

Processor

1

Figure ___. Major components of a PSP reader include the stimulating laser source, a beam splitter, oscillating 2 beam deflector, f- lens, cylindrical reflecting mirror, light collection guide, and photomultiplier tube (PMT). The 3 plate is translated in a continuous motion through the laser beam scan by pinch rollers. All component functions 4 are orchestrated by digital computer. In some readers, multiple PMTs are used for capturing the signal. 5

6 The speed of the laser beam across the phosphor plate is adjusted according to the luminescent signal 7

decay time constant (~0.8 s for BaFBr:Eu2+) following excitation, which is the major factor limiting the readout 8

time. Laser beam power determines the fraction of the stored energy released, and impacts the scan time, 9

phosphorescent lag effects, and residual signal. Higher laser power can release more of the trapped electrons, but 10

the tradeoff is a loss of spatial resolution caused by increased depth of the laser beam and increased spread of the 11

stimulated light in the phosphor layer. 12

At the end of the scanned line, the laser beam is retraced to the start. Since the phosphor plate is 13

simultaneously moving, the translation speed is adjusted such that the next sweep of the laser beam initiates 14

another scan line with spacing equal to the effective sampling pitch along the fast sweep direction. This ensures 15

that sample dimensions are equal in the x and y directions. Scanning and translation of the plate continues in a 16

raster fashion over the total phosphor area. Scan direction, laser scan direction, or fast-scan direction is the 17

terminology that refers to the direction along the path of the laser beam deflection. Slow scan, plate scan, or sub-18

scan direction refers to the phosphor plate travel direction. Phosphor plate translation speed is selected for a given-19

sized plate in order to advance the plate incrementally with a single pass of the laser so that the effective sample 20

size is equal in the scan and sub-scan dimensions. The 1/e2 diameter of the laser spot at the surface of the imaging 21

plate is fixed in all present commercial systems, and imposes an upper limit to spatial resolution in both 22


Page 11

dimensions [REF] . The intensity of PSL as the laser passes across the plate is proportional to the x-ray energy 1

absorbed by that area of the plate. Characteristics of the plate readout geometry are shown in Figure ___. 2

3

Scan Direction

Sub-scan Direction

4

Figure --. A diagram of the raster-scan of the phosphor detector indicates the fast scan (laser scan) direction and 5 the sub-scan (plate scan) direction. Note the slightly skewed angle of the readout lines relative to the edge of the 6 phosphor plate, due to the simultaneous laser beam scanning and linear plate translation. 7

8 Residual latent image signals are contained on the phosphor plate after readout. Erasure of the plate using 9

a high intensity light source is accomplished prior to return to the inventory. Unless an extreme overexposure 10

occurs, essentially all of the residual trapped electrons are effectively removed during the erase cycle. On some 11

systems, the erasure of the plate is a function of the overall exposure, whereby longer exposures require a longer 12

erasure cycle. A summary of the PSP receptor cycle is illustrated in Figure --. 13

14

x-ray exposure

laser beam scan

light erasure

plate exposure:create latent image

plate readout:extract latent imagevia PSL

plate erasure:remove residual signal

PSP

Base support

reuse ofphosphor

plate

un-exposed phosphor plate

15

Figure --. The phosphor plate cycle is depicted above. An unexposed plate is comprised of the PSP material 16 layered on a base support and protected by a thin, transparent coating. Exposure to x-rays creates latent image 17 centers of electrons semi-stable energy traps in the crystal structure. Latent image processing is accomplished with 18


Page 12

a raster-scanned low power laser beam (e.g., 20-milliwatt HeNe laser @633 nm). Trapped electrons are released 1 from the luminescent centers and produce light that is collected by a light guide assembly and directed to a 2 photomultiplier tube. Residual trapped electrons are removed with a high intensity light source, and the plate is 3 returned to the inventory for reuse. 4

5

Detection and conversion of the PSL signal. PSL is emitted in all directions from the phosphor screen. 6

An optical collection system (mirror cavity or acrylic light collecting guide positioned at the laser phosphor 7

interface along the scan direction) captures a portion of the emitted light, and channels it to the photocathode of 8

the PMT (or PMTs) of the reader assembly. Detection sensitivity of the photocathode material is matched to the 9

wavelength of the PSL (e.g., ~400 nm). Photoelectrons emitted from the photocathode are accelerated and 10

amplified through a series of dynodes within the PMT. Gain (and thus detector speed) is varied by adjustment of 11

the voltage placed on the dynodes, so that a useful output current is obtained for a given (clinical) exposure for 12

proper image quality. Dynamic range of the PMT output signal is much greater than that of the phosphor plate, 13

allowing high signal gain over a wide range of exposures. Light intensity variations correspond to incident 14

radiation exposure variations linearly over a range of 1-10,000 or four orders of magnitude. Digitization of the 15

output signal requires the determination of a minimum and maximum signal range, as most clinically relevant 16

transmitted exposure variations occur over about a dynamic range of 100-400. In some PSP readers, a low energy 17

laser pre-scan coarsely samples the exposed PSP receptor and determines the useful exposure range. The gain of 18

the PMT is then adjusted (increased or decreased) to optimally digitize the PSL resulting from the subsequent 19

high-energy laser scan. In most systems, the PMT amplifier is pre-adjusted to be sensitive to the PSL resulting 20

from an exposure range corresponding from 2.5810 9 C/kg (0.01 mR) to 2.5810 5 C/kg (100 mR). 21

Most PSP reader systems then compress the PMT output signal with an analog logarithmic amplifier or a 22

square root amplifier. Logarithmic conversion provides a linear relationship of incident exposure to output signal 23

amplitude; square-root amplification provides a linear relationship with the quantum noise associated with the 24

exposure. In either case, the total dynamic range of signal is compressed to preserve digitization accuracy over a 25

limited number of discrete graylevels. 26

27

Digitization. Digitization is a two step process of converting an analog signal into a discrete digital value. 28

The signal must be sampled and quantized. Sampling determines the location and size of the PSL signal from a 29

specific area of the PSP receptor, and quantizing determines the average value of the signal amplitude within the 30

sample area. The output of the PMT is measured at a specific temporal frequency coordinated with the laser scan 31

rate, and quantized to a discrete integer value dependent on the amplitude of the signal and the total number of 32

possible digital values. An Analog to Digital Converter (ADC) converts the PMT signals at a rate much faster 33

than the fast scan rate of the laser (on the order of 2000 times faster, corresponding to the number of pixels in the 34

scan direction). A pixel clock coordinates the time at which a particular signal is encoded to a physical position on 35

the scan line. Therefore, the ratio between the ADC sampling rate and the fast scan (line) rate determines the 36

pixel dimension in the scan direction. The translation speed of the phosphor plate in the sub scan direction 37


Page 13

coordinates with the fast scan pixel dimension so that the width of the line is equal to the length of the pixel (i.e., 1

the pixels are square). The pixel size is typically between 100 and 200 m, dependent on the dimensions of the 2

phosphor receptor. 3

Although the analog output from the PMT has an infinite range of possible values between a minimum 4

and maximum voltage, the ADC breaks the signal into a series of discrete integer values (analog to digital units) 5

for encoding of the signal amplitude. The number of bits used to approximate the analog signal, or pixel depth 6

determines the number of integer values. PSP systems typically have 10, 12 or 16 bit ADCs, so there are 210 = 7

1024, to 212= 4096, to 216=65536 possible values for a given analog signal amplitude. One manufacturer (Kodak) 8

uses a 16-bit digitization to implement a digital logarithmic transformation to the final 12-bit/pixel image. Other 9

system manufacturers use an analog logarithmic amplifier (Fuji), or a square-root amplifier (Agfa) on the pre-10

digitized signal. Analog amplification avoids quantization errors in the signal estimate when the number of ADC 11

bits (quantization levels) is limited [Seibert, 95]. 12

13

Characteristic Curve Response 14

15

PSP plate

Film-screen(400 speed)

0.01 0.1 1 10 1001

10

100

1,000

10,000

Incident exposure, mR

Rela

tive

inte

nsity

of

PSL

Film

Opt

ical

Den

sity

0

1

2

3

4

Underexposed

Overexposed

Correctly exposed

16 17

Figure__ The characteristic curve of rare-earth screen-film (400 speed) and the PSP receptor are compared. 18 Exposure ranges superimposed on the PSP curve roughly indicate the exposure range for screen film response of a 19 200 speed system. 20

21

Figure ___ illustrates the characteristic curve response of a typical PSP receptor to a 400-speed screen-22

film system. A linear, wide latitude response to variations in incident exposure is characteristic of the phosphor 23

plate, while film is optimally sensitive to a restricted range of exposures. For screen-film detectors, which serve as 24

both the acquisition and display medium, it is necessary to tune the detector (film) contrast and radiographic speed 25


Page 14

to a narrow exposure range to achieve images with optimal contrast and minimal noise characteristics. PSP 1

receptors are not constrained by the same requirements because the acquisition and display events occur separately 2

so that compensation for under- and over-exposures is possible by the algorithms applied to the digital data. 3

However, identification of useful signal range must be accomplished prior to the autoranging and contrast 4

enhancement of the output image. In addition, since under- or overexposed images can be masked by the 5

system, a method to track exposures on an image by image basis is necessary to recognize those situations that 6

exceed the proper exposure range so that appropriate action can be taken to resolve any problems. 7

PROCESSING THE RAW PSP IMAGE 8

Readout Parameters 9

Wanted vs. Unwanted Image Signals. In conventional screen-film radiography, the x-ray technologist 10

adjusts the exposure technique to put the desired range of image signals on the linear portion of the H&D curve. 11

The unwanted image signals from unattenuated x-rays fall into the shoulder (high exposure range) of the curve, 12

and the unwanted image signals beyond the edges of the collimators fall into the toe (low exposure range). The 13

PSP system must similarly encode the useful image signal, to provide maximum contrast sensitivity through look-14

up-table adjustments of the digital values. Just as the radiographic technique and the image receptor are selected 15

for the specific anatomic view, the PSP readout algorithms make adjustments to the digital image specific to the 16

anatomy. 17

Partitioned pattern recognition. The first task for some PSP systems is to determine the number and 18

orientation of views in the raw digital data on the exposed receptor. Each view can then be analyzed 19

independently. While multiple views on a single cassette are good practice in conventional radiography, it can be a 20

possible complication for PSP radiography. 21

Exposure field recognition. Within an exposure field, it is important for the PSP reader to distinguish the 22

useful region of the image by locating the edges of collimation. Some PSP systems further segment the image by 23

defining of the edges of the anatomic region. Once the useful image is correctly located, the PSP system can 24

disregard the image information beyond the collimator boundaries when performing further analysis. 25

Histogram analysis. The method for determining the useful signal range for most PSP systems requires 26

the construction of a gray-scale histogram of the image, a graph of pixel value on the x-axis and frequency of 27

occurrence on the y-axis (i.e., a spectrum of pixel values). Figure ___ shows an example of noiseless histograms. 28

29


Page 15

Uniformunexposedfilm

Uniformexposedfilm

Linearlyexposedfilm

min max

min max

Freq

uenc

y (#

eve

nts) min max

1 2

Figure --. This simple example of image histograms depicts the outcome of an unexposed, uniformly 3 exposed, and linearly exposed PSP receptor without any additive noise. 4

5

The general shape of a histogram is dependent on the anatomy and the radiographic techniques employed 6

for the image acquisition. All PSP readers employ an analysis algorithm to identify and classify the components of 7

the histogram that correspond to bone, soft tissue, skin, contrast media, collimation, unattenuated x-rays and other 8

signals. This allows the discrimination of the useful and unimportant areas of the image so that the image 9

grayscale range can be properly rendered. An example of a chest-specific histogram is shown in Figure __. 10

11

0 200 400 600 800 1,0000

2,000

4,000

6,000

8,000

10,000

12,000

Digital number

Freq

uenc

y

High attenuation(e.g., mediastinum)

Low attenuation(e.g., lungs)

Histogram

12 Figure ___. A chest histogram illustrates the various components of the frequency distribution of pixel 13

values within the active area of the image, corresponding to anatomical variations. 14 15

The result of histogram analysis allows the normalization of raw image data for standard conditions of 16

speed, contrast, and latitude determined by the digital number analysis. Rescaling and contrast enhancements are 17

optimized for the specific patient examination to render the appropriate grayscale characteristics of the final output 18

image. Each manufacturer implements a specific method for this normalization procedure. With some systems, the 19


Page 16

latent image information is identified initially and resampled to a smaller range of digital values to minimize 1

quantization errors. Any errors in identification of the exposure range can be irreversible and require re-acquisition 2

of the image. Other systems digitize the full dynamic range of the PSL signal and then apply non-destructive 3

algorithms to the digital data. Whichever methods are used, the pertinent image information on the phosphor plate 4

must be identified for subsequent display grayscale and/or frequency processing, as the shape and information 5

content of the histogram affects the processing of the image. An example of finding and linearly processing the 6

image signal, also known as autoranging, is described in Figure __ for two exposure scenarios (typical of the 7

processing by Fuji PSP systems). In each case, the proper range of digital values is obtained from the PSL 8

produced by the incident radiation on the phosphor plate. 9

Histogram

0 511 1023100 10-1 100 102 3101

101

10-1

100

102

Exposure input

Rela

tive

PSL

Raw Digital Output

min max

0 200 400 600 800 1,0000

200

400

600

800

1,000

Raw Input digital number

Out

put d

igita

l num

ber

0 511 1023

min max

Grayscale transformation:Input digital number

transformed to mimic film response.

100 10-1 100 102 3101

101

10-1

100

102

Exposure input

Rela

tive

PSL

Raw Digital OutputHistogram

Overexposure:

Typical exposure:

10

Figure --. Autoranging of incident exposure into a corresponding digital number range is accomplished by 11 analyzing the image histogram. (A) Minimum and maximum values of the histogram are mapped to minimum 12 and maximum digital values (10 bit range in this case.) (B) Overexposure results in higher PSL signals that shift 13 the histogram distribution to a higher digital number range, but the system compensates by adjusting the amplitude 14 gain (digital or analog) to compensate. A grayscale transformation of the linear signals into a non-linear 15 relationship by a digital transformation table occurs as depicted on the right hand side of the figure. 16

17 Display Processing. 18

PSP images are matrices of digital pixel values that are readily manipulated to produce alternative image 19

presentations. Three broad categories of processing include image contrast variation, spatial frequency content 20

modification, or special image algorithm implementation. 21

PSP systems manufacturers provide sophisticated computer hardware and software to process images. 22

Some OEMs and third party vendors provide similar functionality for remote processing of image data. No 23

comprehensive source of information about manufacture specific algorithms and implementation exists. This is due 24


Page 17

in part to the immature nature of the PSP marketplace and digital image processing in the practice of radiology. 1

Significant misunderstanding exists, both inside and outside the manufacturer community, about the proper use of 2

image processing software. Selection and optimization of processing parameters is a non-trivial task that 3

potentially requires many thousands of man-hours by highly skilled staff [Vuylsteke, 97 AAPM Summer School]. A common 4

problem is that the range of processing parameters far exceeds clinically useful values and leads to gross over-5

processing artifacts. Modification of processing parameters should not be undertaken lightly. 6

Contrast Processing. Because of small differences in attenuation of the human body, the PSP data has 7

very little inherent contrast. To increase the visibility of anatomic detail, manufacturers provide contrast-8

processing software. The purpose of contrast processing is to create an image data set with contrast similar to 9

conventional screen-film images, or to enhance the conspicuity of desirable features. This type of processing is 10

also referred to as Gradation Processing, Tone Scaling and Contrast Enhancement by various vendors. 11

There are two different methods implemented for contrast processing. The most common technique 12

employs remapping individual pixel values according to user controlled look-up tables (LUTs). Both Fuji and 13

Kodaks default contrast processing alters local image contrast using this technique. Fuji uses four different 14

parameters (GA, GC, GT and GS) to control this processing [Gingold 94 paper on factors], and Kodak uses two (average 15

density and LUT start) [Kodak/Bogucki]. The Fuji processing provides selection of the basic curve shape (GT) that 16

mimic commercially available screen-film systems, the ability to increase or decrease gradient (GC and GA), and 17

overall brightness (GS). Kodak provides for the selection of one of several pre-defined LUTs. A second type of 18

contrast processing implemented by Fuji is known as dynamic range control (DRC). DRC attempts to alter the 19

images global contrast without significantly altering local contrast. This is used to enhance contrast in low signal 20

regions (mediastinum and sub-diaphragm) or high signal regions (air contrast, skin margins) [Kobayashi, SPIE paper, early 21 1990s]. DRC processing is optional, and controlled by three user-selectable parameters for each anatomical menu 22

selection: the kernel size, curve type and boost factor. Agfas image processing relies upon a MUlti Scale Image 23

Contrast Amplification (MUSICA). This processing represents the image as a set of coefficients corresponding to 24

image features at different levels of decomposition. In MUSICA the image is decomposed according to the 25

Laplacian pyramid transform [Burt PJ, Adelson EH: The Laplacian pyramid as a compact image code. IEEE Trans on Comm 1983; 31(4):532-540]. 26

Contrast enhancement is achieved by modifying the coefficients of the Laplacian pyramid. Two software-27

controlled parameters are typically used to modify coefficients. 28

Frequency Processing. One purpose of digital image processing is to enhance the conspicuity of features 29

within the data. Frequency processing enhances features within the image that can be characterized by their 30

specific spatial frequency. Several techniques exist in the literature to accomplish this goal, including Fourier 31

filters [] and blurred-mask subtraction [], and wavelet filtering []. Early users of PSP systems routinely printed each 32

image twice on a single film using different presentations, one presentation designed to mimic the appearance of a 33

conventional screen-film combination, the other with significant amounts of edge-enhancement. This practice is 34

not routinely followed in the U.S. because of user preference for larger image size and a more traditional 35

appearance of the image. 36


Page 18

Both Fuji and Kodak implement blurred-mask subtraction techniques []. In this technique, the original 1

image is blurred by convolution with a uniform kernel of a selected size. The blurred image is then subtracted from 2

the original image, resulting in an image contains predominantly high frequency information. Multiplying each 3

pixel by a user-defined enhancement factor modulates the high frequency information. Adding the resultant image 4

to the original image and normalizing the data set creates the frequency-enhanced image. User selectable 5

parameters include kernel size (RN for Fuji, mask size for Kodak), and enhancement factors (RE for Fuji, boost for 6

Kodak). In addition, both manufacturers provide the capability of spatially localizing enhancement base on gray-7

scale value in the original image. The RT parameter specifies a function whose input is pixel value and output 8

ranges from 0 to 1. The output of the function is multiplied by the RE value to determine the final amount of pixel 9

enhancement. Likewise, Kodak provides for the selection of density localized boost functions. Figure __ 10

summarizes the enhancement technique. 11

12

Solid: original responseDash: low pass filtered

Spatial frequency

Original - filtered

Res

pons

e

low

Sum

Difference:

lowlow

Difference + OriginalEdge Enhanced:

high highhigh 13

Figure__. The steps required for edge enhancement: Left: an original image frequency response (solid line) is 14 blurred by a convolution filter to eliminate high frequency signals (dashed line). Middle: subtracting the blurred 15 image from the original creates a difference signal with frequency components dependent on the amount of 16 blurring. Right: the difference signal is added back to the original image and normalized to provide a mid- to high 17 frequency boost in the filtered image. 18

19

Agfas method for frequency enhancing images closely follows that of their contrast processing. 20

Frequency enhancement of the image is achieved by selective modification of coefficients of the decomposed 21

image. The enhancement of the image becomes apparent upon reconstitution of the image. Two parameters are 22

used to control the modification of the decomposed image coefficients. 23

Generalized image grayscale enhancement and frequency processing examples of a PSP chest image are illustrated 24

in Figure ___ 25


Page 19

1 Figure __. Example chest images demonstrating the flexibility of PSP systems and variable contrast enhancement 2 available. Left: original raw chest image without contrast enhancement. Left center: contrast enhancement 3 applied. Right center: black-bone or reversed contrastoften helpful in identifying tube placement. Right: 4 edge-enhanced image. 5 6

Other Image Processing. Manufacturers have developed special processing software to address limited 7

applications for PSP. These include Dual Energy Subtraction, Tomographic Artifact Suppression and Dynamic 8

Range Control (DRC) supplied by Fuji [], and Noise Suppression supplied by Agfa[]. 9

10

IMAGE DEMOGRAPHICS AND EXPOSURE INDICATORS 11

Demographics 12

It is very important to understand and be able to decode the information available on the hard copy film or 13

the soft-copy image, independent of the PSP system installed. Review of the specific manufacturers user manual 14

will contain all of the pertinent information. A summary of the information should be posted at all reading areas 15

and the PSP equipment. In addition to the standard institution and patient demographic information, several 16

important image processing parameters are listed, including image magnification/reduction factors, type of LUTs 17

used for processing, frequency enhancement settings, latitude of the image data, and incident exposure 18

information, among other vendor-specific factors. Even though a PSP system is manufactured by a specific 19

company, re-sellers will brand their own demographics, notations, and positions on the image, or limit features, 20

e.g. mark reprinted film with different parameters not the same as the original manufacturer. The user must be 21

aware of these changes, and not assume that the information is presented in an identical way or provide identical 22

results. One must refer to the specific user/application manual for the PSP equipment to be used or tested for these 23

details. 24

25

Exposure indicators 26

The PSP system can provide proper optical density or image luminance for under or over exposures 27

because of a wide latitude response and ability to scale the signal. Potential problems with inappropriate 28

techniques can therefore be masked. As a result, it is important to have an indicator of the average incident 29

exposure on the imaging plate to verify proper radiological techniques. Each PSP manufacturer has a specific 30

method for providing this information. In the case of Fuji, a sensitivity number is provided, which is an indication 31

of the amount of amplification necessary to adjust the image information to the correct digital range, and is 32


Page 20

inversely related to the incident exposure. In the case of Kodak, an Exposure Indicator provides a value directly 1

proportional to an average exposure, while Agfa provides a relative exposure value, lgm, based upon a history of 2

previous exposures. A calibration of the system is necessary for reporting accurate results, and is described in the 3

acceptance testing procedures. 4

Fuji PSP systems utilize a sensitivity number to provide an estimate of the incident exposure on the plate 5

transmitted through the object (if any) for the automatic and semi-automatic modes of operation. Under normal 6

processing conditions for the standard resolution (ST) plates, the system sensitivity number for an unattenuated 80 7

kVp beam is given as [Fuji tech rv#3, 93]: 8

S 200exposure (mR)

9

10

A low incident exposure on the phosphor receptor creates fewer activated luminescence centers, and results in a 11

lower PSL signal. Amplification of the signal is required to obtain the optimal analog signal range for digitization. 12

The amount of amplification (or de-amplification when an overexposure occurs) is indicated by the system 13

sensitivity value. As opposed to screen-film systems, PSP systems offer flexibility in exposure level choices. When 14

the system sensitivity number is equal to 200 with the semi-automatic or automatic mode, an average 15

photostimulated luminescence within the area sensed by the reader is estimated as 1 mR (80 kVp, no object, no 16

added x-ray tube filtration other than inherent). This corresponds to a digital value of 511 (the central value of the 17

10 bit grayscale range), and to roughly the speed of a 200 speed screen-film combination. (A 400-speed 18

screen/film combination requires approximately half of the incident exposure). For the fixed sensitivity mode 19

available with the Fuji PSP system, the sensitivity number is set by the user, making the system perform similar to 20

a screen-film detector. Calibration of the system sensitivity response is part of the acceptance test procedures. 21

Since the energy of the x-ray beam determines the relative absorption by the PSP receptor, the system sensitivity 22

response varies with kVp and beam filtration. Fuji does not harden the x-ray beam used to calibrate the system, 23

which limits the usefulness of S as an estimate of the incident exposure on the detector for clinical applications. 24

Kodak PSP systems utilize an exposure index, a value reported by the reader that is directly proportional 25

to the average log incident exposure on the plate, and is calculated as [Bogucki, 95]: 26

27

EI 1000 exposure in mR) + 2000log( 28

29

An exposure of 1 mR (80 kVp, 0.5 mm Cu, 1 mm Al filtration) results in an exposure index of 2000. An exposure 30

of 10 mR leads to an exposure index of 3000, and an exposure of 0.1 mR will result in a value of 1000 for a 31

calibrated system. Doubling the screen exposure results in an increase of 300 in the exposure index value. When 32

using high-resolution PSP receptors, the exposure index has different ranges (see Kodak appendix). 33


Page 21

Agfa PSP systems have a dose-monitoring tool that uses a relative exposure paradigm, available as an 1

option to their systems (and highly recommended to be installed)[Agfa literature]. A dose value, called lgm, is 2

calculated for each scanned image in a similar fashion to the above-described methods, and logged into a lgm 3

database. After ~100 images of the same examination, the lgm mean value and standard deviation are stored as 4

the reference value (unless continued compilation is requested by the user). Exposure level comparisons of future 5

images are made with the reference value, and a graphical indicator is displayed in the text fields of each image. If 6

a given exposure exceeds a pre-determined threshold limit, a black warning bar is printed and warning messages 7

are logged into a database file. The lgm database and calculated data for all exams can be listed at user request[Agfa 8 literature]. 9

10

Exposure concerns when using PSP systems 11

The exposure indicator estimate of the incident exposure to the PSP receptor is sensitive to segmentation 12

algorithms, effective energy of the beam (kVp, filtration), delay between exposure and readout, positioning of the 13

patient relative to the phosphor, and the source-image distance, among other factors. Because the PSP system 14

provides a nearly optimal display of the anatomical information independent of exposure, this number is a very 15

important aspect of quality assurance, patient care, and training issues. Recent publications [Seibert etal, 96; Huda etal 96] 16

indicate the optimal exposure range for most clinical imaging procedures requires an x-ray technique 17

corresponding to a ~200 speed screen-film detector system, based upon the empirical analysis of images and 18

characteristics of the PSP image acquisition process. PSP receptors that are underexposed can be identified by 19

increased quantum mottle caused by an insufficient x-ray flux, resulting in a reduced signal to noise ratio and loss 20

of contrast detectability. For certain studies with detection tasks not requiring a high signal to noise ratio (e.g., 21

naso-gastric tube placement), exposures can be reduced significantly. On the other hand, overexposures are not as 22

easily identified by appearance only and usually do not impact the usefulness of the image, but represent a 23

disservice to patient care and proper radiation safety regulations. It is important to make the manufacturer aware of 24

the need to provide the exposure information in a retrievable database, and also provide visual cues on each printed 25

film or digital soft-copy that will alert the radiologists and technologists that the exposures are outside normal 26

limits. Technologists are advised to adjust their techniques similar to screen-film imaging, particularly for grid and 27

no-grid examinations. This will keep the visible noise reasonably consistent from image to image. 28

29

PSP IMAGE CHARACTERISTICS 30

Spatial Resolution 31

High contrast (limiting) resolution in PSP is determined by several factors. Physical limits imposed by the 32

composition and thickness of the phosphor plate, the size of the laser spot, temporal lag of the PSL, and light 33 scattering within the phosphor contributes to the modulation and loss of the pre-sampled signal. The finite 34

diameter of the laser spot incident on the phosphor layer and the spread of PSL, particularly at depth, contribute to 35

unsharpness, as shown in Figure __. Digital image pixel size is between 100 and 200 m, and determines the 36


Page 22

maximum spatial resolution of the system, up to physical limits imposed by the composition of the imaging plate 1

and the size of the laser spot. Digital sampling confines the maximum spatial frequencies contained in the output 2

image to a maximum determined by the Nyquist frequency, equal to the inverse of twice the pixel dimension, 3

(2x)-1. Unlike conventional cassettes, smaller phosphor plates will often provide better limiting resolution than 4

larger plates, because the pixel size is related to the plate dimension. Resolution sharpness can be increased with 5

the use of a thinner phosphor layer using high-resolution PSP receptors (see Figure __); however, the classical 6

tradeoff of detection efficiency and higher radiation dose must be considered. Phosphorescence lag causes the 7

spatial resolution in the fast scan direction to be slightly less than that in the sub-scan direction as depicted by the 8

MTF curves in Figure __, although one might expect more precision from an electro-optical motion than from a 9

mechanical motion. 10

11

Incident Laser Beam

Light guide assembly and PMT

Protective Layer

Phosphor Layer

Base Support

LightScattering

Laser Light Spread

PhotostimulatedLuminescence

"Effective" readout diameter 12

Figure __. The effective area of the phosphor simultaneously stimulated by the laser is determined by the incident 13 laser diameter, laser light spread within the phosphor, and the distribution of the PSL collected by the light guide 14 assembly. This spread reduces the modulation of higher frequency signals. Adapted from Kato[Kato, 94]. 15

16 Pre-sampled MTF Curves

Standard and High Resolution Phosphor Plates

0 2 4 6 8 100

0.2

0.4

0.6

0.8

1

Spatial Frequency (lp/mm)

MTF

Scan

Subscan

17 Figure __. Typical results for pre-sampled MTF curves with PSP receptors are illustrated. The curve pair on the 18 left are for standard resolution (thick phosphor) and on the right for high resolution (thin phosphor). Solid and 19 dashed lines distinguish the scan and subscan MTFs, respectively. Adapted from Dobbins [Dobbins, etal 95]. . 20

21


Page 23

Aliasing (pre-sampled high frequency signals beyond the Nyquist frequency reflected back into the image 1

at lower spatial frequencies), can negatively affect the PSP image. This artifactual signal is caused by inadequate 2

sampling, which in turn is limited by the size of the pixel and the digital image matrix. For instance, if the 3

intrinsic resolution limit of the phosphor plate is 5 lp/mm, and the pixel sampling rate is 5 pixels/mm, or 2.5 4

lp/mm, the spatial frequencies in the signal spectrum beyond 2.5 lp/mm will be reflected back into the image below 5

2.5 lp/mm. Aliasing can be controlled in the (fast) scan direction by application of a low-pass filter to reduce or 6

eliminate these high frequency signals. This is not possible, however, in the subscan direction. Aliasing will 7

potentially be greater with high-resolution phosphor plates due to improved frequency response beyond the Nyquist 8

limit (particularly for 200 m pixel sampling with Nyquist frequency = 2.5 lp/mm), as shown in with the pre-9

sampled MTF curves in Figure __. The impact of aliasing on image quality is scene dependent, enhances image 10

noise, and reduces the detective quantum efficiency of the PSP receptor. A notable example is the aliased signals 11

caused by anti-scatter grids with lead strip frequencies beyond the Nyquist frequency. On the other hand, 12

significant signal modulation still remaining at the Nyquist frequency provides the appearance of high spatial 13

resolution due to the enhanced contrast response of small objects in the image. 14

15

16

17

18

Figure __ Image demonstrating the visible effects of aliasing caused by a low frequency anti-scatter grid. 19

20

Contrast Resolution 21

The minimum difference in a noiseless signal that can be represented between digital pixels in the 22

image depends on the total number of possible code values (quantization levels), as well as the target signal 23

amplitude relative to the background. In most PSP systems, pixel values change with the logarithm of 24

photostimulated luminescence, or equally with the logarithm of radiation dose to the plate, so the numerical 25

difference between pixel values is the contrast. Contrast sensitivity or detectability of a PSP system depends not 26

only on the number of bits used to represent each pixel, but also by the gain of the system (e.g., # electrons/x-ray 27

photon, # x-ray photons per analog to digital unit) and overall noise amplitude relative to the contrast difference. 28

The ability to differentiate a signal in the image is strongly dependent on the inherent subject contrast (kVp, scatter 29

acceptance), amount of noise (x-ray, luminance, electronic, fixed pattern noise sources), image viewing conditions, 30

and the limitations of the observer to discern regions of low contrast with respect to size. 31

Contrast detectability that is provided by the PSP image is, in general, similar to the screen-film image. 32

As a digital detector, the PSP device permits the separation of latent image acquisition and display processing 33

steps. This allows the ability to vary the radiographic contrast of the displayed image with the application of 34

examination specific gradation, tonescale, or other image manipulations. Without digital enhancement, the visible 35

(radiographic) contrast of the resultant image would be extremely low (see the characteristic curve response in 36


Page 24

Figure __). Unlike screen-film detectors which are contrast limited at a particular radiographic speed (the classic 1

tradeoff between detector latitude and film contrast) the PSP image contrast is noise limited. There are several 2

noise sources that contribute to the overall noise in the image. The random variation of absorbed x-rays in the PSP 3

receptor determines the quantum noise component. Stimulated luminance variations during the readout process 4

contribute significant variations in the output signal. Quantization noise adds inaccuracies in the determination of 5

the discrete digital signal amplitude values (this is dependent upon the bit depth of the ADC, typically 10 to 12 bits 6

in current systems). Electronic noise sources cause a further variation in the output signal. To approximate the 7

typical image noise in a 400-speed film (and thus achieve similar contrast detectability), the PSP receptor (using 8

standard resolution plates) requires a higher x-ray photon flux by about a factor of 2 times [Seibert etal 96]. This is 9

chiefly associated with the lower detection efficiency of the phosphor plate. 10

11

Detective Quantum Efficiency 12

The Detective Quantum Efficiency describes the efficiency of information detection with respect to spatial 13

frequency. It is dependent on the quantum detection efficiency of the screen and the noise associated with each 14

process involved in creating the final image. This includes the number of trapped electrons per absorbed x-ray 15

photon, noise in the stimulation and emission of the latent image, noise in the conversion to an electronic signal, 16

noise associated with the digitization, and noise occurring in the final output image presentation. The large area, 17

zero frequency DQE of a storage phosphor has been described as [Barnes, 93, Lubinsky 87]: 18

DQEX

[1 + CV(E)][1+ CV(el)][1 + CV(S)] + < g >PSPabs

-1 19

where: Xabs = fraction of incident x-ray photons absorbed in the phosphor layer 20

CV(E) = coefficient of variation of the x-ray energy absorbed in the phosphor layer 21

CV(el) = coefficient of variation in the number of trapped electrons for a given absorbed energy 22

CV(S) = coefficient of variation of the light signal emerging from the phosphor for a given number of 23

trapped electrons 24

= the average number of photoelectrons detected at the photomultiplier per absorbed x-ray (the large-25

area response function) 26

Xabs is energy dependent as plotted in Figure ___. CV(E) depends on the overlap of the spectrum with the k-edge 27

of barium and the fraction of K characteristic x-ray escape. For an 80 kVp x-ray beam transmitted through the 28

patient, a value of ~0.15 has been estimated, similar to the CsI phosphor used in image intensifiers [Barnes, 93]. 29

Hundreds of electrons are trapped in the phosphor F-centers per absorbed x-ray photon, making CV(el) relatively 30

small (


Page 25

resolution phosphor plates. These values approximate the published findings of Dobbins[Dobbins etal 95] and Hillen[Hillen 1 etal 87]. Values of DQE(f) has been thoroughly investigated for several generations of storage phosphor imaging 2

plates[Dobbins etal 95]. A steady improvement in the development of phosphor plate technology and subsequent 3

detection efficiency as a function of spatial frequency has been demonstrated. Comparisons of the latest generation 4

PSP phosphor plates to screen-film detectors is very favorable in terms of overall response and image quality. 5

Image Display 6

Laser Film Printers convert the digital images to film images to mimic the conventional screen-film 7

radiography paradigm, where the film is trans-illuminated for viewing. With some PSP systems, the image size 8

must be reduced (de-magnified) by a variable amount, depending upon the phosphor plate size and output film 9

format. Hard copy presentation of the PSP image commits the user to a single rendering, obviating a major 10

advantage of convenient display processing. In order to provide two different grayscale/edge enhancement 11

renderings, the image size may be further reduced to accommodate two images on a single film. This two-on-one 12

format requires a reduction to 50% of the 35 43 cm (14 17 inch) field of view on small format PSP films (~26 13

36 cm). Size reductions complicate direct measurements, comparison to historical studies on film/screen, and 14

inter-comparison of views where the size reduction may be different. Full field of view printing is available on 35 15

43 cm format film, with large sampling matrices up to approximately 4000 4000 pixels (~3500 4300 by one 16

manufacturer) to provide high spatial resolution on the order of 5 lp/mm over the full field of view. 17

CRT Monitors are used for soft-copy display. Digital images from the PSP reader are displayed on CRT 18

monitors for a variety of purposes, including verification of correct patient positioning, Quality Control review and 19

image modification, primary diagnosis, and clinical reference. The capabilities of the monitors, the image 20

processing toolkits available in their associated workstations, and the criticality of their display properties vary 21

according to their function. CRT monitors provide for simultaneous viewing of images throughout the hospital 22

and for real-time modification of image appearance by the observer. CRT monitors share a number of 23

characteristics, including lower luminance levels than a standard lightbox or film alternator, an image produced by 24

fluorescence emission rather than by trans-illumination, an inherently nonlinear display transfer function, potential 25

for fading, geometric distortion, and defocusing. If the monitor is linked to production of hard copy images, 26

matching the appearance of the image on the monitor with the film is an important consideration. The adverse 27

effect of high ambient light levels on the appearance of the image is more problematic with a CRT than with a 28

trans-illuminator because of the lower CRT luminance. In addition, CRT phosphors produce different colors and 29

have difference phosphorescence lag when changing images. An increased emphasis on the acceptance testing and 30

quality control of display monitors and viewing conditions is necessary to ensure optimal image rendition. 31

32

SYSTEM CONFIGURATIONS AND DIGITAL SOFT-COPY INTERFACES 33

The diagnostic radiological physicist is likely to encounter PSP devices manufactured by any of several 34

vendors. These devices often represent different generations of technology and exist in different functional 35


Page 26

configurations. The specific system configuration significantly affects how the physicist conducts acceptance tests. 1

The display media, display processing and content of the digital image data file varies depending on system 2

configuration. 3

Many PSP devices operate as general-purpose devices inside or outside the radiology department. In this 4

application, an imaging plate is inserted into the device and a dedicated laser camera produces a film. Other PSP 5

devices are dedicated to acquisition of upright examinations of the Thorax (Konica, Fuji, and Kodak) and may be 6

integrated into the x-ray generator. Some PSP devices are constructed into a radiographic table (Fuji FCR 7502, 7

FCR 9502). In some hospitals, PSP devices are operated independently with dedicated laser printer. In other 8

hospitals, PSP devices are used to acquire digital data for a sophisticated image and information management 9

system (IMACS). 10

Two decades of research and development, combined with advances in computer technology, have 11

resulted in major improvements in PSP devices [Kato, 94, Fuji Tech Rev#2]. The integration of PSP devices and the 12

development of commercial PSP devices have followed a parallel path. A PSP device is a system that must provide 13

the user with several functions to be clinically useful. In addition to acquiring digital data representing the 14

projected x-ray beam, the integrated system must provide a facility to process, store and render the resulting image 15

data for display. This section reviews possible configurations of devices available at this time. 16

Early Clinical PSP Devices. The first PSP imaging devices were developed in the early to mid 1980s. 17

The first devices in the U.S. were clinically implemented by Philips Medical Systems in 1983, the Fuji Computed 18

Radiography (FCR) 101 and FCR 201 [ CBMerritt chest imaging summer school]. These devices were termed the central 19

processing type [Fuji Technical Review #2]. These early devices, now obsolete, were large enough to fill an average size x-20

ray room, expensive to obtain and operate, and slow to process plates. Image data could only be printed on a 21

dedicated laser camera. No convenient mechanism was provided to move image data outside of the manufacturers 22

domain. These early systems were not commercially successful, however, they did establish the data processing 23

model to be used for the next generation of PSP devices. 24

Independent PSP reader with dedicated laser printer. The second commercially available generation of 25

PSP devices developed by Fuji was the 7000 series, termed the distributed processing type [Fuji Technical Review #2]. The 26

system was marketed in the U.S. by Fuji through OEM relationships as a replacement for screen-film imaging. 27

Most of these systems were standalone, providing all of the necessary data processing support in a single functional 28

unit. Film output to a dedicated laser camera remained the standard method of rendering images, and no method 29

for storing or reprocessing image data was initially provided. Laser printers that could accept multiple PSP device 30

inputs provided some economy of scale. 31

Independent PSP Reader with Optional Digital Output. The OEMs, most notably Philips Medical 32

Systems, realized the potential of integrating output from PSP devices into other information management systems. 33

In conjunction with AT&T Bell Labs, Philips developed the Easy Vision Radiographic Workstation that included 34

a proprietary hardware and software interface to the FCR 7000. Early attempts to develop remote data processing 35

and display workstations were limited by the state of the PSP device interfaces and computer technology. It was 36


Page 27

clear at that time that significant benefits to the practice of radiology could be gained by separating the acquisition, 1

storage, transport and display of medical images [ref. U Arizona]. 2

The first commercially successful PSP device was the AC family developed by Fuji. By this time, the size 3

and cost of the device had been significantly reduced. Processing throughput was markedly improved over earlier 4

versions. The device was no longer marketed as a departmental replacement for screen-film systems, but rather as 5

a method to provide imaging solutions for the emergency room and intensive care units. The AC-1 was a 6

standalone system with an attached film processor. The same laser that scanned the imaging plate was used to 7

expose the laser printer film. No capability for storing or reprocessing of digital data was provided. 8

As PSP systems evolved, so did the radiology departments ability to provide digital image storage, 9

transport and rendering of digital images. Shortly after the introduction of the AC-1 system, Fuji introduced the 10

AC-1+ and AC-2 systems. These systems provided optional access to digital data from the proprietary Data 11

Management System (DMS) via a Small Computer Systems Interface (SCSI) adapter. The Fuji Digital Laser 12

Reader - Digital Acquisition System Manager (FDLR-DASM) was supplied by a third-party vendor (Analogics, 13

Inc. Maynard, MA). The AC-2 did not have an attached film processor, but had an option to print laser film to a 14

magazine. The FLDR-DASM was also an option for the FCR 7000 series scanners. Siemens Gammasonics 15

(Chicago, IL), another OEM, developed a Macintosh-based Computed Radiography Acquisition Workstation using 16

the DASM, and transmitted data via a proprietary network protocol. For the first time, digital image data could be 17

moved from the PSP device to a remote computer system. 18

Independent PSP Reader with Quality Control (QC) Workstation. The data available through the 19

FDLR-DASM, however, was not fully processed by the PSP device, and required additional image processing to 20

match the appearance of the film output. Early use of these interfaces was largely limited to academic institutions 21 [Templeton, et al. 1992, Journal of Digital Imaging]. Fuji subsequently introduced the HIC-654 computer workstation to interface to 22

the AC-1+ and AC-2 via the proprietary DMS interface. The HIC-654 provided the capability to temporarily store 23

data onto a local disk drive, reprocess image data stored on disk, print the reprocessed data and provided processed 24

through a FDLR-DASM. 25

Independent PSP Reader with QC Workstation and Networked Laser Printer. Kodak introduced its first 26

PSP device, the Kodak Ektascan Storage Phosphor Reader (KESPR) 3000 series in 1992. This was the first PSP 27

system that was designed as a data acquisition node for output to a network for image storage, hard copy recording, 28

and diagnostic soft-copy display. The KESPR consisted of the PSP reader device interfaced to a dedicated computer 29

workstation that provided image storage and reprocessing capabilities. Fully processed images could be moved 30

from the computer workstation to remote computer systems and shared laser printers in accordance with the ACR-31

NEMA Version 2.0 standard for medical image communications. 32

Networked PSP reader with Networked QC Workstation and Networked Laser Printer. CEMAX 33

(Milpitas, CA), a third party vendor, introduced the first commercially available network interface adapter (NIA) 34

for a PSP device. The NIA relied on the FLDR-DASM and connected to a standard Ethernet network. Another 35

third party vendor, DeJarnette Research (Towson, MD) introduced a similar NIA that conformed to the ACR-36


Page 28

NEMA DICOM 3.0 standard. Analogics has recently developed the SD100 SuperDASM that fulfills the same 1

function. GE (Milwaukee, WI) has also developed a combined QC workstation and NIA for Fuji PSP devices. At 2

this writing, Fuji, Kodak and Agfa produce PSP readers. All manufacturers have the capability to transfer data 3

processed data to attached laser cameras and to remote computer systems and laser printers via networks using the 4

DICOM3 communications protocol. At this writing, Agfa is the only PSP manufacturer whose PSP reader 5

communicates directly to the network according to DICOM3 conventions. 6

Non-standard Access to Digital PSP Data. In the quest for access to digital PSP data, a number of ersatz 7

methods were developed, including screen dumps of the video driver of a workstation to its local hard drive and 8

down-sampling of full-resolution PSP data. It is important to recognize that data captured by these methods did 9

not include display processing and did not include the full gray-scale resolution or pixel matrix of the original PSP 10

image file. 11

Acquisition and Association of Patient Demographic and Exam Information. Early PSP devices 12

required the operator to manually enter patient demographic and exam information associated with each PSP 13

image. As PSP devices became integrated into radiology operations, more efficient methods were developed to 14

acquire this data including creation of magnetic cards, bar codes, and ultimately, functional interfaces with the 15

Radiology Information System. In order to perform acceptance tests on integrated PSP systems, the physicist may 16

have to create phantom patients in the RIS corresponding to the planned test exposures. 17

Variations in the Size, Header, and Content of the PSP Image Data File. 18

Questions to direct the completion of this section: 19

Is image processing applied to pixel data? 20

Are image processing instructions contained in the header? Are they applied later for display? 21

What is the pixel transfer function of the PSP device? 22

What is the matrix size of the image? 23

How many bits per pixel? 24

Depends on vendor, configuration, conventions. 25

26

27

6. GENERIC FUNCTIONAL SPECIFICATIONS OF PSP SYSTEMS 28

Functional specifications related to typical capabilities/specifications are listed based upon a recent 29

review of vendor literature. It is highly recommended to communicate with marketing specialists and system 30

engineers to determine the up-to-date capabilities/specifications of a particular PSP system prior to purchase, 31

installation and testing. 32

Phosphor plate and cassettes of several sizes are available for PSP systems. The most popular sizes 33

include 35 cm 43 cm (14 17), 35 cm 35 cm (14 14), 24 cm 30 cm (10 12), 24 cm 24 cm (10 34

10) and 18 cm 24 cm (8 10). Specialized cassettes (e.g., 20 cm 20 cm) are also available as options 35


Page 29

from certain manufacturers. The time required reading the phosphor plate is dependent on the plate size. Larger 1

sizes usually take longer to read, and decrease overall system throughput. Spatial resolution is also affected by 2

phosphor plate size. In general, the larger the plate size, the poorer the limiting resolution. Plate inventory should 3

be sufficient to eliminate delays due to accessibility of plates, and not by the throughput of the PSP reader. Related 4

to the second item, it is highly recommended to purchase at least two PSP systems in busy work environments so 5

that backup capability is available in the event of system malfunctions. Regarding types of phosphor plates, 6

standard resolution and high-resolution versions are offered, and must be considered with respect to the imaging 7

applications. A tradeoff of detection efficiency occurs with high resolution plates, requiring on the order of three 8

times more exposure to achieve similar signal to noise ratio. 9

Plate throughput maximum of ~30 plates per hour up to ~110 plates per hour are specified by the various 10

manufacturers, depending on the equipment and options purchased. PSL decay time is the major limit to the 11

throughput speed, although in some systems, plate handling and erasure requirements can add substantial time for 12

the transit of a plate through the reader. Some PSP systems have internal stackers or external automatic handling 13

capabilities to allow the user to leave or very quickly retrieve an unexposed plate without having to wait for the 14

total readout process. 15

Spatial resolution specifications are dependent on the reading and recording laser sampling rates over a 16

given field of view (phosphor plate size), and therefore effective pixel size. Almost all PSP systems utilize a laser 17

beam with an effective 100 m diameter spot size on the phosphor. The output sample size is determined by the 18

number of pixels across the imaging plate, and in most cases is larger for the larger plate sizes. Imaging plate 19

characteristics such as phosphor coating thickness (e.g., standard versus high resolution IP), protective coating 20

layer thickness, finite laser beam dimensions (spot size), light scatter in the phosphor, and frequency response of 21

electrical circuits will degrade the limiting resolution. A range of specifications exist, dependent on the size of the 22

imaging plate, the type of PSP reader, memory options, and image output options, among others. In general, 23

intrinsic spatial resolution typically ranges between 2.5 to 5 lp/mm (0.2 mm to 0.1 mm object detail), somewhat 24

inferior to a 400 speed screen film intrinsic resolution capability of about 7 lp/mm (0.07 mm object detail). 25

There are several generations of IP types available from the manufacturers, and plates with different 26

characteristics. Standard resolution and high resolution plates are often used in the same PSP reader. The 27

former is typically used for all applications in general radiography; while the latter is used for extremities and 28

mammography applications. The thickness of the standard resolution plates is approximately 2 times greater than 29

that of the high-resolution plates. This provides lower spatial resolution, but higher detection efficiency and better 30

contrast resolution. Discussion of resolution factors, MTF measurements and detective quantum efficiency issues 31

are found in the literature [Kato,94;Dobbins etal 95]. CRT (soft copy) displays also influence spatial resolution, and will 32

likely be the limiting factor for displaying image matrix sizes that exceed the bandwidth and number of TV lines of 33

the monitor. However, by displaying a portion of the image by geometric zooming to the intrinsic resolution limit 34

(without pixel replication or interpolation), the monitor does not limit the spatial resolution, but only the display 35

field of view. 36


Page 30

Contrast sensitivity, for an optimally tuned PSP system, the contrast sensitivity is chiefly a function of the 1

image acquisition techniques (kVp, antiscatter grid, geometry, etc.) However, a major equipment issue is the bit 2

depth of the pixel. Ten bits has been shown to be sufficient for film recording [Kato, 94]. There is a potential to lose 3

image information if the scaling algorithms or histogram analysis is improperly applied, despite subsequent image 4

processing. Twelve-bit pixel accuracy is preferred when (user) image processing is performed. The total 5

information contained on the plate is encoded in the digital image. Even if inappropriate processing is applied, 6

reprocessing is still possible in order to display the pertinent information. 7

Dynamic range and incident exposure sensitivity of the PSP receptor extends from approximately 0.01 8

mR up to 100 mR (a range of 10,000 or 104). In some systems, a high gain setting can reduce the lowest 9

detectable exposure to 0.001 mR, but this also reduces the high end to 10 mR maximum. A logarithmic 10

amplification linearizes the exposure-luminance response curve. (Note: in Agfa systems, a square-root 11

amplification is used in lieu of log amplification.) Intrinsic detector and subject contrasts are typically very low, 12

and not clinically optimal. (Four decades of dynamic range is attributed because of this tremendous exposure 13

response however, rarely are four decades of dynamic range required or desired for diagnostic radiology 14

applications.) The range of exposures containing the useful image information is identified with image analysis of 15

the digital distribution on the raw image, usually by histogram analysis. Examination specific algorithms evaluate 16

the distribution and shape of the resultant histogram, followed by linear/non-linear contrast stretching and 17

enhancement to mimic screen-film presentations and/or radiologist preference. 18

19

Desirable Specifications and Features. 20

Phosphor plates, cassettes, grids, identification terminals 21

Enough phosphor plates and corresponding cassettes should be ordered to meet 1.5 times the peak dema

Documents

Computed Radiography