US 5,124,913 AGrant
Rule-Based Technique to Automatically Determine the Final Scan Gain in Storage Phosphor Radiography
Issue Date:1992-06-23
•9 Claims
•6 Drawing Sheets
Abstract
In storage phosphor radiography, a portion of the x-ray energy transmitted through the subject is absorbed by a storage phosphor plate. When stimulated by visible light of the right wavelength, the phosphor plate emits light (at another wavelength) in proportion to the absorbed energy. To obtain the radiographic image, the plate can be scanned in a raster fashion and the emitted light can be photoelectrically detected. The detected signal is then amplified, digitized, processed and finally printed on film, or displayed on a cathode ray tube (CRT). Setting the amplification gain of this scanning process, the so-called final scan (or final read-out) gain, is of interest in maximizing the information extracted from the phosphor. One method of setting the final scan gain invokes the use of a preliminary read-out (pre-scan). A pre-scan can be conducted on the phosphor plate by use of a stimulating ray having a stimulating energy lower than the stimulating energy in the final scan. The present invention provides a technique that uses the pre-scan image histogram to automatically determine the final scan gain in storage phosphor radiography.
Metadata
Assignee
- Eastman Kodak Co.
Inventors
- Muhammed I. Sezan
- Ralph Schaetzing
Application Information
Application Number:US 4519167
Filing Date:1989-12-18
Priority Date:1989-12-18
Art Unit:231
Classifications
IPC:
G06F 1500
Field of Search:
364250413.13327.2 P;327.2 A;327.2 B;327.2 C;327.2 D;327.2 E;327.2 F;327.2 G;327.2 L ;484.1 R;484.1 B
Patent Drawings (6 sheets)
Description
TECHNICAL FIELD
This invention relates to a method of automatically determining the image read-out conditions in storage phosphor radiography systems. The method is based on the histogram of a preliminary image (pre-scan image), read out with a low power stimulating ray prior to the final scan.
BACKGROUND ART
In storage phosphor radiography systems, a storage phosphor is exposed to radiation, to produce a latent image in the storage phosphor. Subsequently, the storage phosphor is scan-simulated to release the latent image in the form of detectable radiation. One of the problems associated with storage phosphor radiography is to determine the intensity of stimulation required to produce an optimum read-out of the storage phosphor. The optimum stimulation intensity depends upon the range of energies stored in the phosphor. To this end, it has been proposed to conduct a preliminary scan-stimulation at low intensity to determine the range of energies stored in the storage phosphor. A final read-out scan-stimulation is then performed based on the results obtained from the preliminary read-out. The intensity of the final read-out scan is also referred to as final scan "gain". See for example European Patent application EP 00778677 A3, published Apr. 27, 1983 Suzuki and Horikawa, where they suggest the use of a preliminary scan as a means of determining the final scan and image processing conditions. Although they proposed storage phosphor radiography systems employing manual and automatic control units that utilize the pre-scan information to determine the final scan conditions, they did not disclose the details of how these units work.
In U.S. Pat. No. 4,682,029 issued Jul. 21, 1987 to Tanaka et al., they used the pre-scan image histogram to determine the minimum (S.sub.min), and the maximum (S.sub.max) signal levels that correspond to the "useful" image information. The final scan gain was determined such that (S.sub.min) and (S.sub.max) would become respectively the the signal levels Q.sub.min and Q.sub.max. At the output, predetermined transformation mapped the signal within the range [Q.sub.min, Q.sub.max ] to the desired output density range [D.sub.min, D.sub.max ]. In this manner, the useful image information was expressed within a predetermined range at the output. This technique used a "percent rule" to determine S.sub.min and S.sub.max from the pre-scan image histogram. The quantity S.sub.max was determined from a gray level that was occupied by 0.1 to 2.0% of the total number of picture elements and S.sub.min was determined from a gray level that was occupied by 0.05 to 1.0% of the total number of picture elements. The major drawback of this technique is that many gray levels may have the same relative percent population. No rule was disclosed to choose among the multiple possibilities.
In European patent application EP 0145982 A1, published Jun. 26, 1985, Tanaka et al. used a slightly different but equivalent perspective in considering the problem. They emphasized the automatic control of a scale factor introduced in the analog-to-digital (A/D) converter. That is, the final scan gain and the scaling, followed by A/D conversion, constituted the final scan conditions. As in U.S. Pat. No. 4,682,028 cited above, the pre-scan histogram was used to determine S.sub.min and S.sub.max. The scale factor was determined from the difference (S.sub.max -S.sub.min). In cases where the spatial extent of the radiation exposure field is limited to a certain anatomical structure (i.e., collimated X-rays), the value S.sub.min is determined mainly by the scattered radiation. This value is smaller than that obtained within the image portion of the radiation exposure field. As a result, the image contrast may decrease if this fact is not taken into consideration in determining the scale factor. In order to alleviate such detrimental effects, Tanaka et al. proposed a technique that required the computation of the histogram (h.sub.2) of the pre-scan data obtained from a sub-region of the storage phosphor plate in addition to the histogram (h.sub.1) of the data obtained from the entire storage phosphor plate (sub-region area normally occupied 20% to 80% of the total plate area). The quantities S.sub.min,1, S.sub.max,1 and S.sub.min,2, S.sub.max,2 were obtained from h.sub.1 and h.sub.2, respectively. (Usually, S.sub.min,1 <S.sub.min,2 and S.sub.max,1 =S.sub.max,2 =S.sub.max.) Tanaka et al. proposed a method to compute a value of S.sub.min from S.sub.min,1 and S.sub.min,2, that would be used to determine the scale factor.
In European patent application EP 0154880 A2, published Sep. 18, 1985 by Tanaka et al., the pre-scan data were collected only from selected sub-regions of the phosphor plate. A characteristic value, S.sub.ch, was calculated from the mean values of the gray levels within these sub-regions. The final scan gain was determined such that S.sub.ch would become the gray level Q.sub.av at the final scan. In the predetermined output transformation, the quantity Q.sub.av was mapped to a desirable output density level D.sub.av. The major disadvantage of this technique is that the location of the sub-regions that are used to collect pre-scan data are exam-, and possibly image-dependent.
In U.S. Pat. No. 4,652,999 issued Mar. 24, 1987, Higashi et al. proposed a configuration where the final scan gain and the image processing conditions were determined automatically from the pre-scan information. The so-called "automatic sensitivity adjusting function" (ASAF) determined the final scan gain based on the exam type and image recording conditions (e.g., chest exam and lung field magnification). The final scan gain was determined such that the image information presented to the output station was within a predetermined range [Q.sub.min, Q.sub.max ], which was mapped to some predetermined density range [D.sub.min, D.sub.max ] at the output. But, a desired D.sub.max may have been specified for the lung field only rather than for the entire image. In that case, the lung field can have the desired output dynamic range only if the x-rays were coned (collimated) onto the lung field (lung field magnification image). Higashi et al. addressed this problem by proposing a "secondary automatic gradation" unit which would ensure that the structure of interest, rather than the entire image, had the desired dynamic range for varying image recording conditions. This control unit was provided with the recording conditions and the value at the output of the ASAF unit. The working principles of the ASAF unit for determining the final scan gain were not disclosed.
As further experience with storage phosphor imaging systems has been gained, it has become apparent that further improvements in methods for adjusting the final read-out conditions based on a preliminary read-out are needed.
Another problem that has been discovered as experience has been gained is that unsatisfactory exposures are not discovered until the final image is read out, processed and displayed. This whole process can consume a good deal of computer time that is wasted if the image must be re-taken.
DISCLOSURE OF THE INVENTION
It is therefore the object of the present invention to provide an improved method for automatically determining the read-out conditions in a storage phosphor radiography system.
In the method according to this invention, peaks (or group of peaks, called clusters) that correspond to major anatomical structures and the background portion (if any) of the pre-scan image are detected. For example, FIG. 1 shows a histogram having 3 peaks 10, 12 and 14, forming a cluster 16. A cluster is composed of j(j=1,2, . . . ) peaks and reduces to a peak for j=1. In what follows, we use the terms `peak` and `cluster` interchangeably. The final scan gain is then determined such that the peaks are moved to desired gray level (or code value) locations in the final scan. By placing the peaks at desired gray level locations in the final scan, the associated anatomical structures will be at the desired gray levels. The desired locations depend on the exam type and on the image recording conditions. In general, the main consideration in determining the final scan gain is to be able to utilize effectively the available range of gray levels by using a gain as high as possible. At the same time, the gain should not be so high to cause useful image information to saturate at the maximum gray level (in general, only the background portion of the image is allowed to saturate at the maximum gray level).
One advantage of the proposed method over the previous pre-scan histogram based methods is the detection of individual histogram cluster supports instead of the entire histogram support, [S.sub.min, S.sub.max ]. Since histogram clusters, in general, correspond to major anatomical structures, this technique is more flexible than the others for maintaining the major structures at desired gray levels in the final image. Furthermore, the entire histogram support as used in the prior art methods, determined from the minimum and the maximum gray levels (S.sub.min, S.sub.max) present in the image, may not correctly represent the useful image information range. In particular, S.sub.min may underestimate the true minimum, S.sub.max, on the other hand, may overestimate the maximum value of the useful image information range. This can be because of the existence of insignificantly populated higher gray levels isolated from the major population range, or because of background peaks. The former situation is illustrated in FIG. 2 which illustrates a pre-scan histogram with isolated levels 18. The existence of such levels may be due to (1) nonuniformities in the x-ray beam, (2) x-ray noise, (3 ) x-ray scatter, (4) phosphor plate structure noise, and (5) scanner noise. In FIG. 2, the final scan gain based on S.sub.max would not fully utilize the allowable dynamic range of gray levels (e.g., [0,2.sup.B -1] for a B-bit digital radiography system). The gray level e.sub.2 in FIG. 2 is a better estimate for the maximum of the useful image information range. The method according to the present invention estimates the cluster supports [s.sub.1, e.sub.1 ] and [s.sub.2, e.sub.2 ], and bases the final scan gain on the value e.sub.2.
According to one aspect of the present invention, the final scan gain in a storage phosphor radiography system is adjusted by: performing a preliminary scan read out at low intensity; generating a histogram from the results of the preliminary scan read-out; locating clusters of peaks in the histogram; determining the peaks, clusters representing structures of interest; and adjusting the final scan gain such that the clusters of peaks representing anatomical structures of interest are located at desired gray levels in the output image, depending on exam type and image recording conditions.
According to another aspect of the invention, a gain factor g is computed from the histogram data, and is applied on a pixel by pixel basis to the pre-scan image to produce a quality control image. The quality control image display on a monitor prior to final read-out of the image so that an operator can order a re-take of the image if the quality is unacceptable.
BRIEF DESCRIPTION OF DRAWINGS
FIG. 1 is a histogram illustrating peaks and clusters;
FIG. 2 is a histogram illustrating insignificantly populated isolated levels;
FIG. 3 is a block diagram of a storage phosphor radiography system where the final-scan gain is determined by the method of the present invention;
FIG. 4 is a flowchart showing the final scan gain setting method according to the present invention;
FIG. 5 illustrates a pre-scan histogram of a chest exam;
FIG. 6 illustrates a pre-scan histogram of a lumbar spine exam; and
FIG. 7 illustrates a pre-scan histogram of a hands exam.
MODES OF CARRYING OUT THE INVENTION
FIG. 3 is a schematic diagram illustrating a storage phosphor radiography system for implementing the final scan gain determination method of the present invention. A switch 20 is at position 1 when a pre-scan is conducted at the scanner station 22. The pre-scan data are amplified in an amplifier 24 (g.sub.p), digitized in an A/D converter 26 and fed to a digital computer 28 that performs the gain setting algorithm. Exam type, image recording condition and system parameters are provided to the computer and hence the gain setting algorithm via a user interface 30. The gain setting algorithm determines a gain factor g which is then scaled by system parameters to determine the final scan gain g.sub.f. The gain factor g is used in computing a "quality control image" (QCI) displayed on a monitor at a quality control station 32, prior to the final read-out. The QCI is computed by pixel-by-pixel scaling of the pre-scan image by the gain factor g. An operator inspects QCI for motion artifacts or patient misplacement. In the case of severe motion artifacts or misplacements, the operator may call for a re-take, otherwise the patient is released prior to final read-out. The final read-out is conducted using a higher laser power (and possibly higher resolution) at the scanner station 22 with the switch 20 at position 2. The gain of the final scan amplifier 34 is set to g.sub.f where ##EQU1## and where g.sub.l is the gain introduced by the increase laser power during the final scan. The final scan data are digitized in an A/D converter 36 and supplied to an image processing station (IPS) 38. The image processing station 38 (IPS) is a digital computer that is programmed to implement various image processing algorithms. The IPS can incorporate tone-scale transformations, unsharp marking, whose parameters can be automatically configured. The final image, or multiple images processed differently, are printed on film, or displayed on CRT, or recorded on an archiving system at the output station, or encoded and transmitted to a remote site at an output station 40.
The method for setting the final scan gain according to the invention includes the following:
I. Generate the histogram, h(n), and the cumulative distribution function (CDF) (or normalized cumulative histogram), c(n), of the pre-scan image;
II. Detect the peaks (or clusters) in the histogram h(n);
III. Based on a set of "selection rules", determine one major set of peaks or clusters representing structures of interest that are to be used in gain setting;
IV. On the basis of exam-dependent rules and image recording condition, determine the gain factor g;
V. Using the system parameters, such as the pre-scan gain, the laser power, photomultiplier tube settings, etc., and the gain factor g computed in Step IV, compute the electronic gain g.sub.f that is to be used in the final scan.
A flowchart showing the steps employed in the gain setting method is illustrated in FIG. 4. We now discuss each of these five steps.
STEP I (42): We denote the histogram of the pre-scan image by h(n), n=0,1, . . . ,2.sup.B -1, where B is the number of bits per pixel used to represent the image. Then, CDF, c(n) is computed as ##EQU2## where M denotes the total number of pixels in the image.
STEP II (44): Peaks of the pre-scan histogram, h(n), are detected, for example by using the peak detection method disclosed in U.S. Pat. No. 4,731,863 issued Mar. 15, 1988 to Sezan et al. which is incorporated herein by reference.
In the peak detection algorithm, a peak detection function is generated from CDF. First, c(n) is smoothed by convolving with a uniform rectangular window w.sub.N (n) to produce a smoothed CDF, c.sub.N (n),: ##EQU3## where the uniform rectangular window is such that ##EQU4## and N is assumed to be odd. The smoothed CDF, c.sub.N (n), is subtracted from c(n) to generate the peak detection function r.sub.n : ##EQU5##
The following principles are applied to the peak detection function r.sub.N to estimate the start and end points of the peaks.
(i) A zero-crossing of the detection signal to negative values (henceforth, negative crossover) indicates the start of a peak. The gray level at which the negative crossover occurs is defined to be the estimate of a start point. For the ith peak, this gray level is denoted by s.sub.i. Similarly, the next negative crossover at the gray level s.sub.i+1 estimates the start of the next peak.
(ii) The gray level between two successive negative crossovers at which the detection signal attains its local maximum is defined to be the estimate of the end point of the peak. For the ith peak, this gray level is denoted by e.sub.i. The peaks are denoted by intervals defined by their start and end points, i.e., [s.sub.i, e.sub.i ].
The length of the window, N, determines the sensitivity of peak detection. The parameter N is therefore referred to as the `peak detection sensitivity parameter`. As the value of N is decreased, the peak-detection sensitivity increases. To detect peaks accurately, the above procedure (Step II) is iterated twice with two different window sizes, N=N.sub.1 and N=N.sub.2 (N.sub.1 >N.sub.2), and two sets of peaks are obtained:
where I.sub.2 .gtoreq.I.sub.1 because the sensitivity of the peak detection increases with decreasing window size. The other parameters in the peak detection algorithm are set to the values disclosed in U.S. Pat. No. 4,731,863.
STEP III (46): The purpose of this step is to select from both A.sub.1 and A.sub.2 a final set, A, of peaks (so-called "major" peaks) that will be used in gain calculation. The selection is performed on the basis of a number of rules. The rules determine the peaks (or groups of peaks) from A.sub.1 and A.sub.2 that significantly overlap with each other, and take only one representative peak (or peak cluster) into consideration in gain calculation.
The A.sub.2 -intervals (or equivalently the A.sub.2 -peaks), i.e., [s.sub.i.sup.2, e.sub.i.sup.2 ]'s may overlap with the A.sub.1 -intervals (or equivalently the A.sub.1 -peaks), i.e., [s.sub.i.sup.1, e.sub.i.sup.1 ]'s. If the relative population of pixels contained in the overlap exceeds a predetermined value then the overlap is said to be "significant". Nonoverlapping peaks, or insignificantly overlapping peaks are called "independent" peaks. The overlapping and the independent peaks are determined by the overlap detection procedure described in U.S. Pat. No. 4,731,863. To summarize, the set A of the major peaks are formed via the following rules (R1-R2):
R1. An A.sub.1 -peak qualifies for the set A if
(i) it is an independent peak, or
(ii) it is not an independent peak but the total number of the significant overlaps is less than t, where t is an empirically predetermined parameter. (If an A.sub.2 peak overlaps significantly with an A.sub.1 -peak then the overlap is said to be a `major overlap` if the ratio of the number of pixels contained in the overlap to the total number of pixels contained in the A.sub.1 -peak exceeds the value R.sub.maj.)
R2. An A.sub.2 -peak qualifies for the set A if
(i) it is an independent peak, or
(ii) it is not independent, but its overlap with the A.sub.1 -peak is a major overlap and the total number of A.sub.2 -peaks that have major overlaps with the A.sub.1 peak is at least t, or
(iii) it is not independent and its overlap with the A.sub.1 -peak is not a major one, but here exist at least t other A.sub.2 -peaks with major overlaps with that A.sub.1 -peak. In this case, adjacent peaks that do not have major overlaps with the A.sub.1 -peak are combined into single peaks.
The final set A can be defined as
The overlap detection algorithm is explained in detail in U.S. Pat. No. 4,731,863. The recommended values for the parameters of the overlap detection algorithm in the present invention are: N.sub.1 =2161; NN.sub.2 =541; R.sub.maj =0.80; and t=2. STEP IV (48): Given the major set of peaks determined in the previous step, the exam type (50), and the image recording condition (50), the gain factor is calculated using a rule base. The image recording condition (52) may be either (i) x-rays are collimated, or (ii) x-rays are not collimated. The exam types are classified into three major categories. That is, each incoming exam is classified into one of the following catagories: (i) chest, (ii) extremity, and (iii) abdomen. Each category has its own set of rules. These rules were determined experimentally from thousands of exams.
In each category, the rule base first determines whether or not the peaks in the final set correspond to an anatomical structure, background or a mixture of the two. Then, the gain factor is calculated with two main objectives: (1) to set the gain high enough such that the available gray level range is fully utilized, and (2) to set the gain low enough such that valuable image information is not saturated at the maximum gray level (in general, only the background portion of the image is allowed to saturate at the maximum gray level). In the following, we provide the rule base for three exam categories.
In our notation, capital letters denote the user-specified (predetermined) parameters. The subscripts denote the exam category, i.e., `c` for chest, `e` for extremity and `a` for abdomen. The superscripts denote the image recording condition and the modality of the histogram. For instance, Q.sub.e.sup.u,c denote the parameter for extremity exams when the x-rays are collimated and the pre-scan histogram is unimodal. We provide recommended values of the user-specified parameters for a 12-bit digital radiography system. These values have been obtained as a result of studying thousands of cases.
CHEST EXAMS
A. The Rule Base
1. IF the histogram is unimodal, THEN the gain is set such that e.sub.1 is mapped to gray level (or code value) Q.sub.c.sup.u : g=Q.sub.c.sup.u /e.sub.1
2. IF the histogram is not unimodal, THEN
(a) IF at least a predetermined percentage, P2% of the total number of pixels attain values in the interval [e.sub.2, q.sub.max -1] (where q.sub.max denotes the largest gray level present in the pre-scan image), THEN the gain is determined by the following rule: First, the local maximum of the histogram in the interval [s.sub.2, e.sub.2 ] is determined. Let m.sub.2 denote the code value at which the local maximum occurs. Then the interval [m.sub.2, q.sub.max -1] is searched for the smallest code value at which the histogram attains a value less than or equal to K1h(m.sub.2) (K1<1 is a predetermined coefficient),
(b) IF the percentage of pixels that attain values in the interval [e.sub.2, q.sub.max -1] is greater than or equal to a predetermined percentage, P1, but smaller than P2 THEN the previous rule is used with K2h(m.sub.2) (K1<K2<1 is a predetermined coefficient),
(c) IF the percentage of pixels that attain values in the interval [e.sub.2,q.sub.max -1] is less than P1 THEN
IF the x-rays are not collimated
i. IF the histogram is bimodal, THEN
ii. IF the histogram is not bimodal, THEN the gain is set such that e.sub.2 is mapped to code value Q.sub.c : g=Q.sub.c /e.sub.2.
IF the x-rays are collimated, THEN the gain is set as in 2(a).
B. Recommended Values of Parameters ##EQU6##
EXTREMITY EXAMS
A. The Rule Base
1. IF the histogram is unimodal, THEN
IF the x-rays are not collimated, THEN the gain is set as g=Q.sub.e.sup.u /e.sub.1 provided that not more than P.sub.e % of the pixels are mapped to the maximum code value of the system (e.g. 4095 in a 12-bits/pixel system) in the final output, ELSE the gain is set such that g=Q.sub.e.sup.u /e where e (e<e.sub.1) is determined such that 1% of the pixels are mapped to the maximum code value of the system in the final output.
IF the x-rays are collimated, THEN the gain is set such that e.sub.1 is mapped to code value Q.sub.e.sup.u,c : g=Q.sub.e.sup.u,c /e.sub.1
2. IF the histogram is not unimodal, THEN
IF the x-rays are not collimated
(a) IF the histogram is bimodal, THEN
i. If the slope of the CDF computed between e.sub.1 and s.sub.2 is greater than the predetermined threshold S.sub.e, THEN the gain is set as g=Q.sub.e /e.sub.2 provided that not more than P.sub.e % of the pixels are mapped to the maximum code value of the system (e.g. 4095 in a 12-bits/pixel system) in the final output, ELSE the gain is set such that g=Q.sub.e /e where e (e<e.sub.2) is determined such that 1% of the pixels are mapped to the maximum code value of the system in the final output.
ii. IF the slope of the CDF computed between e.sub.1 and s.sub.2 is less than or equal to the predetermined threshold S.sub.e, THEN the gain is set such that a code value between e.sub.1 and s.sub.2, determined from a convex combination of e.sub.1 and s.sub.2, i.e., (L.sub.e)e.sub.1 +(1-L.sub.e)s.sub.2, is mapped to code value Q.sub.e. (0.ltoreq.L.sub.e .ltoreq.1. )
(b) IF the histogram has more than two clusters, THEN
i. IF the slope of the CDF computed between e.sub.2 and s.sub.3 is greater than the predetermined threshold S.sub.e, THEN the gain is set as g=Q.sub.e /e.sub.3 provided that not more than P.sub.e % of the pixels lie in [e.sub.3, q.sub.max -1], ELSE the gain is set such that the percentage of the pixels that are mapped to the maximum code value of the system in the final output is 1%.
ii. IF the slope of the CDF computed between e.sub.2 and s.sub.3 is less than or equal to the predetermined threshold S.sub.e, THEN
IF the x-rays are collimated, THEN the gain is set such that the end point of the last cluster is mapped to code value Q.sub.e.sup.c.
B. Recommended Values of Parameters
Q.sub.e.sup.u =3500; Q.sub.e.sup.u,c =1500 Q.sub.e =4095; Q.sub.e.sup.c =3500
S.sub.e =2.0.times.10.sup.-4 ; P.sub.e =2.0; L.sub.e =0.4.
ABDOMEN EXAMS
A. The Rule Base
1. IF the histogram is unimodal, THEN
The gain is set as g=Q.sub.a.sup.u /e.sub.1 provided that not more than P.sub.a % and not less than 0.5% of the pixels are mapped to the maximum code value of the system in the final output, ELSE the gain is set such that g=Q.sub.a.sup.u /e where e is determined such that 0.5% of the pixels are mapped to the maximum code value of the system in the final output.
2. IF the histogram is not unimodal, THEN
IF the x-rays are not collimated
(a) IF the histogram is bimodal, THEN
i. IF the slope of the CDF computed between e.sub.1 and s.sub.2 is greater than the predetermined threshold S.sub.a, THEN the gain is set as g=Q.sub.a.sup.u /e.sub.2 provided that not more than P.sub.a % of the pixels are mapped to the maximum code value of the system (e.g. 4095 in a 12-bits/pixel system) in the final output, ELSE the gain is set such that g=Q.sub.a.sup.u /e where e (e<e.sub.2) is determined such that 0.5% of the pixels are mapped to the maximum code value of the system in the final output.
ii. IF the slope of the CDF computed between e.sub.1 and s.sub.2 is less than or equal to the predetermined threshold S.sub.a, THEN the gain is set such that a code value between e.sub.1 and s.sub.2, determined from a convex combination of e.sub.1 and s.sub.2, i.e., (L.sub.a)e.sub.1 +(1-L.sub.a)s.sub.2, is mapped to code value Q.sub.a. (0.ltoreq.L.sub.a .ltoreq.1.)
(b) IF the histogram has more than two clusters, THEN
i. IF the slope of the CDF computed between e.sub.2 and s.sub.3 is greater than the predetermined threshold S.sub.a, THEN the gain is set as g=Q.sub.a.sup.u /e.sub.3 provided that not more than P.sub.a % of the pixels lie in [e.sub.3, q.sub.max -1], ELSE the gain is set such that the percentage of the pixels that are mapped to the maximum code value of the system in the final output is 0.5%.
ii. IF the slope of the CDF computed between e.sub.2 and s.sub.3 is less than or equal to the predetermined threshold S.sub.a, THEN
IF the x-rays are collimated, THEN the gain is set such that the end point of the last cluster is mapped to code value Q.sub.a.sup.c.
B. Recommended Values of Parameters
In Rules 2(a)(ii) and 2(b)(ii), the coefficients used in forming the convex combinations of peak parameters vary with the slope of the CDF as follows: ##EQU7## STEP V (54): The final scan gain g.sub.f is obtained by simply scaling the gain factor g by a coefficient that reflects the system parameters 56, such as the the laser power, photomultiplier tube settings and the pre-scan gain. If we denote the gain introduced by the increase in laser power during the final scan by g.sub.l, and the pre-scan gain by g.sub.p, then the final scan electronic gain g.sub.f is given by
where g denotes the gain factor determined in the previous step.
EXAMPLES
In the following we present three examples. The corresponding pre-scan image histograms, h(n) 58, and CDFs, c(n) 60, are illustrated in FIGS. 5-7, respectively for three examples. The peak information and the gain factors calculated according to the method of the present invention are given below. These results are obtained by using the recommended values of the parameters noted above.
EXAMPLE NO. 1
Exam Type: Chest (lateral)
Image Recording Condition: x-rays are not collimated
Exam Category Used in Gain Setting: Chest
A.sub.1 ={[0,386], [1665,3850]}
A.sub.2 ={[0,261], [2099,3495]}
A={[0,386], [1665,3850]}
Rule Used in Gain Setting: CHEST: 2.(c)(i)(A)(2)
Gain Factor(g): 3.99
EXAMPLE NO. 2
Exam Type: Lumbar Spine (AP)
Image Recording Condition: x-rays are not collimated
Exam Category Used in Gain Setting: Abdomen
A.sub.1 ={[0,1395]}
A.sub.2 ={[0,1015]}
A={[0,1395]}
Rule Used in Gain Setting: ABDOMEN: 1
Gain Factor(g): 1.16
EXAMPLE NO. 3
Exam Type: Hands
Image Recording Condition: x-rays are not collimated
Exam Category Used in Gain Setting: Extremity
A.sub.1 ={[0,1279]}
A.sub.2 ={[0,667], [706,1249]}
A={[0,667], [706,1249]}
Rule Used in Gain Setting: EXTREMITY: 2.(a)(i)
Gain Factor(g): 3.17
A Fortran program for implementing the gain setting method according to the present invention is included in Appendix A.
APPENDIX A ##SPC1##
Claims
We claim:
1. A method of adjusting final scan gain to produce a final image in a storage phosphor radiography system, comprising the steps of: preliminarily reading out an exposed storage phosphor by stimulating the storage phosphor with a lower intensity stimulation to produce a preliminary image signal, obtaining a histogram of the preliminary image signal, finally reading out said exposed storage phosphor by stimulating said storage phosphor with a higher intensity stimulation based on the histogram to produce a final image signal, and in a digital computer, a. generating a histogram from the results of the preliminary read-out; b. detecting peaks and/or clusters of peaks A.sub.1 and A.sub.2 in the histogram; c. determining one major set of peaks or clusters, A, representing structures of interest; and d. determining a gain factor g that will locate the major set of peaks at desired gray level locations in the final image based on a set of exam dependent rules and image recording conditions.
2. The method claimed in claim 1, wherein said finally reading out step includes using a laser for conducting final scan stimulation of the storage phosphor, detecting a final image signal with a photomultiplier tube, and amplifying the final image signal by means of a variable gain amplifier, and further comprising the step of: employing system parameters including pre-scan gain, laser power, photomultiplier tube setting, and the gain factor g to compute an electronic gain factor g.sub.f which is applied to the variable gain amplifier.
3. The method claimed in claim 1, wherein the steps of detecting peaks and/or clusters of peaks comprises the steps of: a. forming a cumulative distribution function (CDF) from the histogram; b. smoothing the cumulative distribution function with a sliding window average of size N.sub.1, to produce a smoothed cumulative distribution function; c. subtracting the smoothed cumulative distribution function from the cumulative distribution function to produce a peak detection function; d. employing the peak detection function to locate a set of peaks, A.sub.1 in the histogram by 1). identifying positive to negative zero crossings of the peak detection function as the start of a detected peak and 2). identifying a maximum following such a positive to negative zero crossing as the end of a detected histogram peak; and e. repeating steps a-c with a window of size N.sub.2 where N.sub.1 >N.sub.2 to locate a second set of peaks A.sub.2.
4. The method claimed in claim 3 wherein the step of determining one major set of peaks, A, is performed according to the following rules: R1. an A.sub.1 -peak qualifies for the set A if (i) it is an independent peak, or (ii) it is not an independent peak but the total number of the significant overlaps is less than t, where t is an empirically predetermined parameter. (If an A.sub.2 peak overlaps significantly with an A.sub.1 -peak then the overlap is said to be a `major overlap` if the ratio of the number of pixels contained in the overlap to the total number of pixels contained in the A.sub.1 -peak exceeds the value R.sub.maj.); and R2. an A.sub.2 -peak qualifies for the set A if (i) it is an independent peak, or (ii) it is not independent, but its overlap with the A.sub.1 -peak is a major overlap and the total number of A.sub.2 -peaks that have major overlaps with the A.sub.1 peak is at least t, or (iii) it is not independent and its overlap with the A.sub.1 -peak is not a major one, but there exist at least t other A.sub.2 -peaks with major overlaps with that A.sub.1 -peak, in which case, adjacent peaks that do not have major overlaps with the A.sub.1 -peak are combined into single peaks.
5. The method claimed in claim 1, wherein the sets of exam dependent rules include a set of rules for chest exams comprising the following rules: a. IF the histogram is unimodal, THEN the gain is set such that e.sub.1 is mapped to gray level (or code value) Q.sub.c.sup.u : g=Q.sub.c.sup.u /e.sub.1 wherein, the superscript denotes the modality of the histogram and the subscript denotes the exam category of chest exam; b. IF the histogram is not unimodal, THEN (1). IF at least a predetermined percentage, P2%, of the total number of pixels attain values in the interval [e.sub.2, q.sub.max -1] (where q.sub.max denotes the largest gray level present in the pre-scan image), THEN the gain is determined by the following rule: First, the local maximum of the histogram in the interval [s.sub.2, e.sub.2 ] is determined, let m.sub.2 denote the code value at which the local maximum occurs, then the interval [m.sub.2, q.sub.max -1] is searched for the smallest code value at which the histogram attains a value less than or equal to K1h(m.sub.2) (K1<1 is a predetermined coefficient), (2). IF the percentage of pixels that attain values in the interval [e.sub.2, q.sub.max -1] is greater than or equal to a predetermined percentage, P1, but smaller than P2 THEN the previous rule is used with K2h(m.sub.2) (K1<K2<1 is a predetermined coefficient), (3). IF the percentage of pixels that attain values in the interval [e.sub.2, q.sub.max -1] is less than P1 THEN (a). IF the x-rays are not collimated i. IF the histogram is bimodal, THEN ((1)). IF the percentage of the total number of pixels that attain the value q.sub.max is less than or equal to P.sub.c %, THEN, ((a)). IF the slope of the CDF between e.sub.1 and s.sub.2 is greater than the predetermined threshold S.sub.c, THEN the gain is set such that s.sub.2 is mapped to code value Q.sub.c : g=Q.sub.c /s.sub.2, ((b)). IF the slope of the CDF computed between e.sub.1 and s.sub.2 is less than or equal to the predetermined threshold S.sub.c, THEN the gain is set such that a code value between e.sub.1 and s.sub.2, determined from a convex combination of e.sub.1 and s.sub.2, i.e., (L.sub.c)e.sub.1 +(1-L.sub.c)s.sub.2, is mapped to code value Q.sub.c, ((2)). IF the percentage of the total number of pixels that attain the value q.sub.max is greater than P.sub.c %, THEN the gain is set such that e.sub.2 is mapped to code value Q.sub.c : g=Q.sub.c /e.sub.2, ii. IF the histogram is not bimodal, THEN the gain is set such that e.sub.2 is mapped to code value Q.sub.c : g=Q.sub.c /e.sub.2, (b). IF the x-rays are collimated, THEN the gain is set as in b.(1).
6. The method claimed in claim 1, wherein the sets of exam dependent rules include a set of rules for extremity exams comprising the following rules: a. IF the histogram is unimodal, THEN, 1). IF the x-rays are not collimated, THEN the gain is set as g=Q.sub.e.sup.u /e.sub.1 provided that not more than P.sub.e % of the pixels are mapped to the maximum code value of the system (e.g. 4095 in a 12-bits/pixel system) in the final output, ELSE the gain is set such that g=Q.sub.e.sup.u /e where e (e<e.sub.1) is determined such that 1% of the pixels are mapped to the maximum code value of the system in the final output, where the superscript denotes the modality of the histogram and the subscript denotes the exam category of extremity, 2). IF the x-rays are collimated, THEN the gain is set such that e.sub.1 is mapped to code value Q.sub.e.sup.u,c ; g=Q.sub.e.sup.u,c /e.sub.1, where the superscript c reflects the fact that the x-rays are collimated, b. IF the histogram is not unimodal, THEN 1). IF the x-rays are not collimated, (a). IF the histogram is bimodal, THEN i. IF the slope of the CDF computed between e.sub.1 and s.sub.2 is greater than the predetermined threshold S.sub.3, THEN the gain is set as g=Q.sub.e /e.sub.2 provided that not more than P.sub.e % of the pixels are mapped to the maximum code value of the system (e.g. 4095 in a 12-bits/pixel system) in the final output, ELSE the gain is set such that g=Q.sub.e /e where e (e<e.sub.2) is determined such that 1% of the pixels are mapped to the maximum code value of the system in the final output, ii. IF the slope of the CDF computed between e.sub.1 and s.sub.2 is less than or equal to the predetermined threshold S.sub.e, THEN the gain is set such that a code value between e.sub.1 and s.sub.2, determined from a convex combination of e.sub.1 and s.sub.2, i.e., (L.sub.e)e.sub.1 +(1-L.sub.e)s.sub.2, is mapped to code value Q.sub.e, (b). IF the histogram has more than two clusters, THEN i. IF the slope of the CDF computed between e.sub.2 and s.sub.3 is greater than the predetermined threshold S.sub.e, THEN the gain is set as g=Q.sub.e /e.sub.3 provided that not more than P.sub.e % of the pixels lie in [e.sub.3,q.sub.max -1], ELSE the gain is set such that the percentage of the pixels that are mapped to the maximum code value of the system in the final output is 1%, ii. IF the slope of the CDF computed between e.sub.2 and s.sub.3 is less than or equal to the predetermined threshold S.sub.e, THEN ((a)). IF the slope of the CDF computed between e.sub.1 and s.sub.2 is greater than the predetermined threshold S.sub.e, THEN the gain is set such that a code value between e.sub.2 and s.sub.3, determined from a convex combination of e.sub.2 and s.sub.3, i.e., (L.sub.e)e.sub.2 +(1-L.sub.e)s.sub.3, is mapped to code value Q.sub.e, ((b)). IF the slope of the CDF computed between e.sub.1 and s.sub.2 is less than or equal to the predetermined threshold S.sub.e, and the second cluster is closer to the third, THEN ((1)). IF the histogram has three clusters, THEN the gain is set such that a code value between e.sub.1 and s.sub.2, determined from a convex combination of e.sub.1 and s.sub.2, i.e., (L.sub.e)e.sub.1 +(1-L.sub.e)s.sub.2, is mapped to code value Q.sub.e, ((2)). IF the histogram has four clusters, THEN the gain is set such that s.sub.3 is mapped to Q.sub.e : g=Q.sub.e /s.sub.3, ((3)). IF the histogram has more than four cluster, THEN the gain is set such that e.sub.3 is mapped to Q.sub.e : g=Q.sub.e /e.sub.3, ((c)). IF the slope of the CDF computed between e.sub.1 and s.sub.2 is less than or equal to the predetermined threshold S.sub.e, but the second cluster is closer to the first, THEN the gain is set such that a code value between e.sub.2 and s.sub.3, determined from a convex combination of e.sub.2 and s.sub.3, (L.sub.e)e.sub.2 +(1-L.sub.e)s.sub.3, is mapped to code value Q.sub.e, 2). IF the x-rays are collimated, THEN the gain is set such that the end point of the last cluster is mapped to code value Q.sub.e.sup.c, where the superscript c reflects the fact that the x-rays are collimated.
7. The method claimed in claim 1, wherein the sets of exam dependent rules include a set of rules for abdomen exams, comprising the following rules: a. IF the histogram is unimodal, THEN 1). The gain is set as g=Q.sub.a.sup.u /e.sub.1 provided that not more than P.sub.a % and not less than 0.5% of the pixels are mapped to the maximum code value of the system in the final output, ELSE the gain is set such that g=Q.sub.a.sup.u /e where e is determined such that 0.5% of the pixels are mapped to the maximum code value of the system in the final output, where the superscript denotes the modality of the histogram, and the subscript denotes the exam category abdomen, and b. IF the histogram is not unimodal, THEN 1). IF the x-rays are not collimated, (a). IF the histogram is bimodal, THEN i. IF the slope of the CDF computed between e.sub.1 and s.sub.2 is greater than the predetermined threshold S.sub.a, THEN the gain is set as g=Q.sub.a.sup.u /e.sub.2 provided that not more than P.sub.a % of the pixels are mapped to the maximum code value of the system (e.g. 4095 in a 12-bits/pixel system) in the final output, ELSE the gain is set such that g=Q.sub.a.sup.u /e where e (e<e.sub.2) is determined such that 0.5% of the pixels are mapped to the maximum code value of the system in the final output, ii. IF the slope of the CDF computed between e.sub.1 and s.sub.2 is less than or equal to the predetermined threshold S.sub.a, THEN the gain is set such that a code value between e.sub.1 and s.sub.2, determined from a convex combination of e.sub.1 and s.sub.2, i.e., (L.sub.a)e.sub.1 +(1-L.sub.a)s.sub.2, is mapped to code value Q.sub.a, (b). IF the histogram has more than two clusters, THEN i. IF the slope of the CDF computed between e.sub.2 and s.sub.3 is greater than the predetermined threshold S.sub.a, THEN the gain is set as g=Q.sub.a.sup.u /e.sub.3 provided that not more than P.sub.a % of the pixels lie in [e.sub.3, q.sub.max -1], ELSE the gain in set such that the percentage of the pixels that are mapped to the maximum code value of the system in the final output is 0.5%, ii. IF the slope of the CDF computed between e.sub.2 and s.sub.2 is greater than equal to the predetermined threshold S.sub.a, THEN, ((a)). IF the slope of the CDF computed between e.sub.1 and s.sub.2 is greater than the predetermined threshold S.sub.a, THEN the gain is set such that a code value between e.sub.2 and s.sub.3, determined from a convex combination of e.sub.2 and s.sub.3, i.e., (L.sub.a)e.sub.2 +(1-L.sub.a)s.sub.3, is mapped to code value Q.sub.a, ((b)). IF the slope of the CDF computed between e.sub.1 and s.sub.2 is less than or equal to the predetermined threshold S.sub.a, and the second cluster is closer to the third, THEN the gain is set such that a code value between e.sub.1 and s.sub.2, determined from a convex combination of e.sub.1 and s.sub.2, i.e., (L.sub.a)e.sub.1 +(1-L.sub.a)s.sub.2, is mapped to code value Q.sub.a, ((c)). IF the slope of the CDF computed between e.sub.2 and e.sub.3 is less than or equal to the predetermined threshold S.sub.a, but the second cluster is closer to first, THEN the gain is set as g=Q.sub.a /e.sub.3, 2). IF the x-rays are collimated, THEN the gain is set such that the end point of the last cluster is mapped to code value Q.sub.a.sup.c, where the superscript c reflects the fact that the x-rays are collimated.
8. The method claimed in claim 1, wherein the storage phosphor radiography system includes a quality control station having a display monitor, further comprising the steps of: a. scaling the preliminary-scan image pixel-by-pixel to produce a quality control image, and b. displaying the quality control image on the monitor prior to final read out of the image.
9. A method of adjusting final scan gain to produce a final image in a storage phosphor radiography system, comprising the steps of: preliminarily reading out a storage phosphor by stimulating the storage phosphor with a lower intensity stimulation to produce a preliminary image signal, obtaining a histogram of the preliminary image signal, finally reading out said storage phosphor by stimulating said storage phosphor with a higher intensity stimulation based on the histogram to produce a final image signal, and comprising the steps of: a. computing a gain factor g for the final scan stimulation, and b. scaling a preliminary scan image pixel-by-pixel to produce a quality control image signal, and c. displaying the quality control image signal on a monitor prior to conducting the final scan stimulation, whereby an operator viewing the quality control image may order a re-take of the image prior to final read-out in the event that the quality control image is unsatisfactory.