NL7909276A - RADIATION IMAGE PROCESSING. - Google Patents

Description

"* ^ ·. * 70 8393 3etr .: Method and device -for processing a straline image .___

The invention relates to a method for processing a radiation healing in a radiation image recording system used for medical diagnosis, and to a device therefor. The invention relates in particular to a method and an apparatus for processing an image in a radiation healing recording system, wherein a stimulant phosphor is used for recording radiation image information. as an intermediate recording medium, and the image recorded therein is read out for display-1C and finally recording on a recording medium.

It is known in the art, as disclosed in U.S. Pat. No. 3,859,527, to record radiation image information in a phosphor to be stimulated by exposing the phosphor to radiation passing through an object, such as the human body, than the read out recorded information using a photodetector by stimulating the phosphor with a scanning laser beam and the like, and recording the readout information on a recording medium by modulating a recording laser beam and the like with the photodetector selected information.

The previously described radiation healing recording system is much more advantageous than conventional radiography using a silver halide photographic film in that the image is recorded over a wide range of radiation exposure. This system is therefore particularly valuable when used in the medical diagnosis, registering the human body in the manner of conventional radiography. By using this system for such a purpose, it becomes possible to obtain the information which cannot be obtained with the conventional radiography due to the limited exposure range, which can be recorded and observed on the conventional X-ray film.

7909275

V

2

Furthermore, because the X-rays are harmful to the human body, it is impossible from a safety point of view to expose the human body to X-rays of a large dose. Therefore, it is desirable that the necessary information in the radiography can be obtained by exposing the human body to the X-rays at a relatively low dose only once. On the other hand, radiograms should preferably have a large exposure clearance and a high image quality with a high contrast, a high sharpness, low noise, etc. for consideration and diagnosis. Unfortunately, since the conventional radiography is designed to meet all the required conditions to some extent, the range of recording density or the possibility of recording different levels of information and the image quality are both inadequate, none of which are fully met.

The aforementioned radiation image recording system as disclosed in U.S. Pat. 3,859,527, does not solve the foregoing difficulties in conventional radiography, although the system is new in itself.

The main object of the invention is to provide a method and an apparatus for processing a radiation image in a radiation image recording system, using a stimulant phosphor, and meeting both the capability of a wide exposure clearance '25 as high image quality.

Another object of the invention is to provide a method and an apparatus for processing a radiation image in a system for recording a radiation image using a stimulant phosphor, whereby a radiation image can be obtained with great diagnostic efficiency and accuracy.

Yet another object of the invention is to provide a method and apparatus for processing a radiation image in a radiation image recording system using a stimulant phosphor, with high speed 7909276 3 and low cost a radiation image can be provided with great diagnostic efficiency and accuracy.

Research and experiments have shown that the spatial frequency components of the radiation image of a human body, which radiation image is important for diagnosis, are in the region of very low frequencies (hereinafter referred to as "ultra-low frequency1"), although few difference is in the main frequency between the parts of the human body to be diagnosed, and it has also been found that more pronounced high-frequency components do not improve diagnostic efficiency and accuracy, but emphasize noise components and reduce diagnostic efficiency and accuracy. , on the other hand, the reduction in highlighting of the high-frequency components reduces the noise and provides a legible image from the viewpoint of diagnosis. The invention is based on the foregoing findings.

The present method of processing a radiation image is characterized in that, during scanning of a stimulable phosphor with a stimulation beam, reading out the radiation image information registered thereon, converting the information read out into a electrical signal and recording a viewable image on a recording medium using the electrical signal, an unsharp masking signal with a Sus corresponding to the ultra-low frequency is obtained for each scan point, as well as a signal conversion represented by the formula:

Sf = Sorg + ft (Sorg - Sus) (l) wherein Sorg is an original image signal read from the phosphor to be stimulated, and ^ 3 is a coefficient for highlighting, used for highlighting the frequency component above ultra-low frequency.

According to the invention, more than one unsharp mask can be used as long as the signal conversion according to the previous formula is performed. For example, when two blurry masks of different sizes are used, the formula can be represented as 7909276 "1 * follows: S" = Sorg + β (Sorg - Sus 1) + c ((Sorg - Sus 2).

However, this formula can be rewritten in the form: S '+ Sorg + (ƒ> + <Λ) £ Sorg - (/ ^ Sus1 + # Sus2) j 5 This rewritten formula means that the previous operation using the two blurry masks, can be considered an equivalent of said ground action using only one unsharp mask. When the size of the unsharp mask Sus2 is smaller than that of the unsharp mask Sus1, and the coefficient o (for highlighting is positive, the graph representing the modulation transfer function has a shape that has an additional peak in the higher frequency component in the range of the brought-out frequency When the coefficient o (for decongesting is negative, the graph has a done-15 te, which has a stepped lower portion in the higher frequency component in the range of the to frequency brought out.

The former is suitable for recording an image of bones, blood vessels (angiography) and the stomach (double contrast), the latter is suitable for recording an image of breast tomography, cholecystography, liver, the lower abdomen and head.

Furthermore, the present method encompasses any method in which the results of the signal conversion are the same as those of the aforementioned formula (1), wherein there is no limit or limitation on the sequence.

The unsharp masking signal Sus means a signal representing each scanning point made by unsharping the original image signal to contain only the frequency component lower than the ultra-low frequency. In other words, the unsharp mask signal Sus is a signal displaying an unsharp image obtained by blurring the original image to such an extent that the unsharp mask signal contains only the ultra-low frequency. In the unsharp mask, corresponding to the unsharp image, the modulation transfer function is not less than 0.5 at the spatial frequency 79 0 9 2 7 6 5 of 0.5 periods / mm. In order to further improve the diagnostic efficiency and accuracy clearly, it is desirable to use an unsharp mask, the modulation transfer function being not less than 0.5 at the spatial frequency of 0.02 pe-5 riodes / mm, and not more than 0.5 at the spatial frequency of 0.15 periods / mo.

In other words, the unsharp mask to be used according to the invention can be defined as the mask, the spatial frequency fc, wherein the modulation transfer function becomes 0.5 10, ranges from 0.01 to 0.5 periods / mm. , and preferably in the range of 0.02 to 0.15 periods / mm.

Furthermore, it should be noted that the original signal contains a signal processed with a well known signal processing means used in the field of optical instruments, ie, that the signal is not rectilinearly amplified by logarithmic amplification and the like for band narrowing or non-linear compensation. It is often useful to logarithmically narrow the signal to narrow the signal's fingertips when processing a signal representing the strength of light and the like.

According to the invention, the hatching coefficient β may be fixed or changed as a function of the original image signal Sorg or the unsharp mask signal Sus. By changing the coefficient of expression as a function of the original image signal Sorg or the blurry mask signal Sus, the diagnostic efficiency and accuracy are further improved.

By further selecting the highlighting coefficient β and the unsharp masking signal Sus, the ratio of the maximum value B of the modulation transfer function of the system, which provides the visible image on the final recording medium, can be based on the signals brought together, up to the limit value A of the modulation transfer function, which is a limit value at which the spatial frequency is infinitely close to .35, ie B / A. Under the condition 7909275 6 • of B / A <1.5, diagnostic efficiency and accuracy have not improved much compared to conventional radiography. When the coefficient f5 for hatching is fixed, the ratio B / A should not exceed 6, because if the ratio is over 6, the image becomes partly unnatural due to over-hatching, for example the area whose density is saturated to white (veil level of the recording medium) or black (maximum density of the recording medium) appears in the image. On the other hand, when the co-10 efficient p before hatching is changed according to the original image signal Sorg or the blurry mask signal Sus, the desired ratio range B / A is increased, where it may be greater than 6, if not greater than 10. In this case, the maximum value of the B / A ratio is considered the 15 B / A value, because the B / A ratio itself changes when the Sorg or Sus value changes. It is therefore necessary that the ratio B / A is 1.5 to 6 when the coefficient (¾ for hatching is determined) and 1.5 to 10 when this coefficient is changed with Sorg or Sus. it has been found that the diag-20 nose efficiency and accuracy are markedly improved when the B / A ratio is selected in the range of 2 to 5.5 and 2 to 8 in the cases involved.

The highlighting coefficient β is chosen so that the B / A ratio is within the desired range. The ratio. However, B / A also changes somewhat with the shape of the unsharp mask or the unsharp mask signal Sus. Therefore, the value of 33 cannot be readily determined unless the shape of the unsharp mask, i.e., the signal Sus, has been determined.

The unsharp mask can be obtained, for example, by the following ways.

(1) The original image signal at each scan point is stored in a memory, the stored original image signals being read out together with the environmental signals according to the size of the unsharp mask to obtain an average value as the unsharp mask signal.

79 0 9 2 7 6 ...............: .......

7

Sus. The mean value is obtained as a simple arithmetic mean or different types of weighted mean. In this way you make them blurry mask in the form of analog signals or in the form of digital signals after A / D conversion.

Furthermore, it is also possible to make the unsharp mask by sending the analog signal through an underpass screen in the main scanning direction, and processing the signal in digital form in the minor scanning direction.

(2) After the original image signal is read by using a small diameter light beam and the like, the unsharp masking signal is read by using a large diameter light beam. This is possible if the phosphor to be stimulated can still be stimulated after the first stimulation.

(3) Widening the diameter of the stimulating light beam, which takes place during the beam passing through the stimulating phosphor layer by scattering, is used. When the stimulating light beam scans the phosphor to be stimulated, the original image signal Sorg is obtained on the incident side of the phosphor layer, whereby the unsharp masking signal Sus is obtained on the opposite side of the phosphor layer. In this case, the size of the unsharp mask can be controlled by changing the degree of light scattering action by the phosphor layer or changing the size of the aperture used to receive the scattered light.

In the previous three ways, the first way is most preferable from the viewpoint of being able to provide versatility to the image processing.

To perform the first way, the following arithmetic mean operation is performed for each scan point to obtain the unsharp mask signal Sus.

Sus = Σ * (X. Sorg (i, j) (2) i »j§0 ^ where i and ^ are the coordinates of the circular area, which has the scan point in the middle of it, and CK - a weight co. 7909275 ή 8 is efficient, which preferably has a gradual isotropic change in all radial directions, and conforms to the formula Σ, <5 <- = 1. This circular area contains an I number of images.

i, j 6 0 J

elements in the direction of their diameter.

However, in order to easily perform the above operation, • / 7 2, / 7 * 2, it is necessary to perform the -ψ multiplications and IT additions. Therefore, when N is a large number, it takes a very long time to perform the operation, which is impractical. Since it is necessary to scan the phosphor plate to be stimulated at a key speed of 5 to 20 pixels / mm (50 to 200 µm with respect to the size of the picture element) in order to preserve the required frequency components of the picture, the number of pixels I contained in the unsharp mask corresponding to the ultra-low frequency is inevitably large, so that it takes a long time to perform the previous operation. For example, in the case of applying an unsharp mask with a weighting coefficient, provided with a Gaussian distribution, N becomes about 50 when the size of the picture element is 100 µm x 100 µm, and fc = 0, 1 period / mm, and about 250 if fc = 0.02 20 cycles / mm. This means that the time to perform the previous operation is considerably long.

Furthermore, in order to obtain the arithmetic mean for the circular area, the range in which the addition must be performed must be changed for each scan line, which makes the processing mechanism very complicated and expensive.

It is therefore desirable to simplify the operation to reduce the time to perform the operation to obtain the unsharp masking signal. An example of such simplified ways is to obtain the 30-fold arithmetic mean (non-weighted arithmetic mean) over a rectangular area enclosed by two lines parallel to the main scanning direction and two lines parallel to the subordinate scan. In other words, the unsharp mask signal Sus is obtained by calculating simple arithmetic mean of the original image signals Sorg 7909276 9 within the rectangular region. Another example of such simplifications is making an unsharp signal in the main scanning direction by transmitting the analog signal of the original image signal through a lower pass screen, and then obtaining the arithmetic mean of the A / D converted digital signals in the sub scanning direction.

In the former way, in which the unsharp mask signal Sus is obtained by a simple arithmetic mean within a rectangular area, it has been found that the results in the diagnostic inaccuracy and accuracy are as good as those obtained using the ideal circular unsharp mask , which has a Gaussian distribution in its weighting coefficient, although the foregoing has disadvantages, in that the degree of blur differs in the direction, and further the transfer function has an undesired fluctuation compared to the mask provided with a gradually changing weighting coefficient in the form of the Gaussian distribution. This manner is furthermore much more advantageous in that the processing is very simple, and accordingly does not take long, resulting in a high speed and low cost of the image processing apparatus.

These benefits apply to both analog and digital signals. It becomes more detailed in case the weighting coefficient

.. must be multiplied by the original signal —J

Sorg (i, j) at each scan point (i, j), the blurry mask-sew Sew Sus (IJ) obtained by the operation, represented by the formula: y

Sus (IJ) =. Sorg (i, j) (3)

ij j Z

wherein i, j are the numbers, which indicate the coordinates of the scanning point or pixel, and I, J are the numbers, which indicate the coordinates of the obscure mask. (· 2 · © (.. = 1).

i, j. 2

The number of calculations is therefore approximately 3J multiplications 2 and approximately 1 additions, II being the number of pixels contained in an unsharp mask applied in one direction. Therefore, when the number of picture elements in the blur 7900278 * • 4

TO

mask is large, it will take a very long time to obtain the blurry Mas signal Sus. Therefore, for example, when the size of the unsharp mask is 6 mm x 6 mm, and 3 ^ 00 pixels (0.1 mm x 0.1 mm) are contained in the unsharp mask, 3600 multiplications and 3600 additions must be repeated . When an eight-bit microcomputer is used to perform these calculations with 3 ms for multiplication and 5 / Us for one addition, it takes about 3 ms x 3600 + 5 ^ us x 366 s 11 s to get one unsharp mask signal.

In contrast, in accordance with the above, as; first-mentioned way, using the simple arithmetic mean, the time to obtain the blurry masking signal. For example, it takes only 18 ms to obtain a blurry mask signal. Furthermore, by using these tens mentioned below, the number of calculations can be reduced to only U, resulting in a greatly reduced time of only a few tenths of a second to obtain one unsharp mask signal Sus. The blurry mask signal Sus (IJ) 2q. in other words can be obtained by:

Sus -5- (2 Sij) (b)

IST

2 which means only F additions and division. When more detailed, the unsharp mask has a rectangular shape with a size of F. in the main scanning direction and F * in the underside. . aft such a suitable subtract, with regard to the number of picture elements, becomes the blurry mask signal. Sus tide) represented by the formula:

Sus (lJ) = · γ "- (2 * Sij) (5) 1 2 Fj-1 Fj-1 where i is a number in the range of I - —to I + —r— V1 V1 30 and j is a number is in the range of J - —until J + - ^ - and F ^ and F ^ are positive odd numbers. This means that the unsharp, masking signal can be obtained by F ^ x F ^ additions and only one division. improve the operation 7909276 11 kar--further reduce the number of calculations for obtaining an unsharp mask signal to an average of k.

Because the modulation function of the rectangular, unsharp mask with regular sweeps, a sinc function becomes [. . g · * r * ΊΓ x 1 ...

sm'x) = - ^ - J, the said definition is that the spatial frequency fc, where the nodulation transfer function is 0.5. equals 0.01 to 0.5 periods / mm, preferably 0.02 to 0.15 periods / am, equivalent to having a size of βθ mm to 1.2 mm through the rectangular, unsharp mask , preferably 10 30 mm to b ma. Furthermore, in case the shape of the unsharp mask is an elongated rectangle, each side of the rectangle corner has a length within the previous range. In the case of processing an image with a linear tomography, it is preferred that the unsharp mask have an elongated rectangular shape.

In the latter way, using an underpass sieve to obtain an unsharp masking signal Sus, it has also been found that the results in diagnostic inefficiency and accuracy are as good as those obtained by using the ideal, circular, unsharp mask with a variable sweeping coefficient, although the previous way does not have a regular out of balance weighting coefficient. This manner is further more advantageous in that the operation is very simple, and therefore does not take long by only conducting the passing of the analog signal in the main scanning through a lower pass screen to make the processing of the digital signals such that no unnecessary time is consumed. This results in high speed and low cost of the image processing apparatus. In addition, if addition of the digital signals in the subordinate scanning direction is made a simple arithmetic addition to obtain an arithmetic mean, it is not necessary to perform multiplication, which also results in simplification of the device and acceleration of operation. It has also been found that even in such a highly simplified manner, the resulting diagnostic efficiency and accuracy is not less noticeably less than due to the manner in which the unsharp mask is obtained by an ideal operation, which is time consuming.

According to the invention, it is possible to perform a smoothing in addition to said blurry masking. Since noise generally appears, especially in the high frequency range, diagnostic efficiency and accuracy are usually improved by performing a smoothing. As for this smoothing, it is desirable to make the modulation transfer function not less than 0.5 at the spatial frequency of 0.5 cycles / mm, and not more than 0.5 at the spatial frequency of 5 cycles / mm . The desired degree of smoothing depends on the types of the radiation part. For example, in the case of breast tomography where the pattern is to be observed at a relatively low frequency, it is desirable that the noise be removed as much as possible. In contrast, in the case of angiography, where the fine patterns containing the fine blood vessels are to be observed at a high frequency, the excess flattening damages the fine patterns, thereby reducing the quality of the image. According to research conducted, however, diagnostic efficiency and accuracy are improved for virtually all types of radiation images when smoothing is performed within this range. Furthermore, it has also been established that smoothing is effective not only when it is applied to the signal S 'after unsharp masking, but also when it is applied directly to the original image signal Sorg.

Furthermore, according to the invention, a grading can be performed in addition to the foregoing unsharp masking. Grading (such as contrast enhancement using a non-linear or linear signal conversion) is particularly effective for the radiation image, gradually changing the density or strength of emitted light over a wide range, such as the image of lung or breast cancer . The grading that 7909276 13

A progress used in a radiation image recording system is described in Japanese Patent Applications 53 (1916) -162573, 5 ^ (1979) -23091 and 5 ^ (1979) -23092. Grading can be performed before or after unsharp masking.

The phosphor to be solubilized means a phosphor, which can store radiant energy when exposed to a radiation, such as light or high energy radiation, and then emit light according to the stored energy in optical, thermal, mechanical, chemical or electrical stimulate 'that. It is desirable that the phosphor to be stimulated emit light with a wavelength in the range of 300 to 500 nm. For example, rare earth-activated, alkaline earth metal fluorocrhalide phosphorus is preferred. An example of this phosphorus, as described in Japanese Patent Application 53 (1973) —BUT1 + 2, is represented by the formula (3a., Με, Ca) EX : aEu, where X consists of Cl and / or Ί — x — y jC y ^

Br, x and y are numbers that satisfy 0 <x + y = 0.6 and xyrO, and -6 <<-2 a are a number xs, which satisfy 10 = a = 5x10. Another example of this phosphor, as described in Japanese Patent Application 53 (1973) -8U7kU, is represented by the formula 20 (Ba ^ ^ M'j ^ FXiyA, which consists of Mg, Ca, Sr, Zn and / or

Cd, X consists of Cl, 3r and / or J, A consists of Eu, Tb, Ce, Ta, Dy, <<.

Pr, Ho, Ud, Yb and / or Er, x is a number satisfying 0 = x = 0.6 and y is a number satisfying 0 = y = 0.2. Furthermore, as the phosphor to be stimulated according to the invention, use can be made of SnC: Cu, Pb; 3a0.xAl? 0_: Eu, where 0.8 * = x = 10, and M ^ Q.xSiO ,,: A, where Μ1 * is equal to Mg, Ca, Sr, Zn, Cd or Ba, A is equal to Ce, Tb, Eu, Tm, Pb, Tl, Bi or Mn and x is a number satisfying 0.5 = x = 1.5, as described in Japanese patent application 53 (1978} -8U7UO. stimulating phosphorus can furthermore be used LnOX: xA, wherein Ln equals La, Y, C-, Gd and / or Lu, X equals Cl and / or Br, A equals Ce and / or Tb , and x is a number satisfying 0 x 0.1 as described in Japanese Patent Application 53 (1973) -8UtU3. For the aforementioned phosphors, the rare earth-activated, alkaline earth metal fluorine halide phosphorus deserves most preferred, particularly using 7903275

1U

is made from barium fluoride halides, given the high intensity of the emitted light,

Furthermore, it is desirable to color the phosphor layer of the phosphor plate to be stimulated, made of the aforementioned phosphor, by using pigments or paints to improve the sharpness of the image obtained therewith, as described in Japanese Patent Application 5 ^. (1-979) -716 ° U.

As for the stimulating rays for stimulating the phosphor, after it has been activated with the radiant energy according to an image pattern, a high-directivity laser beam is used. As the laser beam, a laser with a wavelength in the range of 500 to 800 nm, preferably 600 to 400 nm, is preferred. For example, a He-Ee laser (633 nm) or a Kr laser (6 ^ 7 nm) can be used. If a color screen, which cuts off the light 15 with a wavelength outside the range of 500 to 800 nm, is used together with a light source, it may have a wavelength distribution beyond the intended range.

The radiation image information read by using the present device is used to display a radiation image on a recording medium, such as a silver halide photographic film, a diazo film or an electrophotographic recording material. Furthermore, it is possible to display the radiation image on a cathode ray tube.

The invention is further elucidated with reference to the drawing, in which: Fig. 1 shows an example of the radiation image recording system, in which the present method of processing the image is applied, Fig. 2A-D graphs 3A - D are graphs for showing the different ways of changing the coefficient Jb for highlighting the brightness, shown by the original image signal Sorg or the blurry mask signal, Sus, 79 0 9 27 5 15 * Fig. k is a graph, illustrating an example of the change of the coefficient p * before bursting with respect to the value of the original image signal Sorg, FIGS. 5 and 6 are graphs for showing examples 5 of the change of the coefficient ƒ3 to highlight just with respect to the image signal, FIGS. 7 and 5 are block diagrams for illustrating circuitry for obtaining the unsharp masking signal, FIG. 9 shows a ten-part system for calculating the unsharp masking signal, FIGS. 10A-C the capacitance showing the memories used to perform the ten-part system of FIG. 9, FIG. 11 is a block diagram for showing an example of the circuit for executing the ten-part system of FIG. 9, FIG. 12, and 13 show the changes of the memories used to perform the previous ten-part system, FIG. 1¾ illustrates another ten-part system for calculating the unsharp mask signal, FIGS. 15A-D show the capacity of the memories used for the Fig. 16 shows the ten-part system according to Fig. 1 in more detail, the ten-part system according to Fig. 1U, Fig. 1JA-B are graphs for showing the relationship between the hatched frequency and the determination of the value of the acquired images for the diagnostic efficiency and accuracy, Fig. 18 is a graph for showing the relationship between the degree of hatching and the determination of the value of the obtained images for the diagnosis efficiency and accuracy, Fig. 19 is a graph for showing the relationship between the hatched frequency and the valuation of the obtained images for the diagnosis efficiency and accuracy, 7909276 16 and Fig. 20 a graph is to represent the relationship between the degree of hatching and the valuation of the acquired images for diagnostic efficiency and accuracy.

The invention is now described in more detail with reference to its particular embodiments used in an X-ray image recording system using a stimulant phosphor.

Fig. 1 shows an example of the radiation image recording system 10 using the present image processing method. With reference to fig.

1, the stimulating phosphor 1 is mounted on a drum 2. The scanning mirror 3a deflects the stimulating beam from the laser source 3 in the main scanning direction B, the drum 2 being rotatable about its axis to move the stimulating phosphor 1 into the subordinate scanning direction A. The phosphor to be stimulated is thus scanned two-dimensionally by the stimulating beam. A stimulating beam source (laser) 3 is provided for transmitting a stimulating beam to the drum 2. The stimulating beam emitted by the laser source 3. collides with the stimulus 1, mounted on the drum 2, to sense and stimulate the stimulus 1.

The light emitted by the phosphor 1 to be stimulated when it is stimulated is received by a photodetector, such as a photoyer multiplier k., By a light guide Ua. The output of the photodetector k is amplified by an amplifier 53 and then converted into a digital signal by an A / D converter 6. The digital signal is recorded on a magnetic ribbon 7 ·

The digital signal stored in the magnetic ribbon 7S 30 is read by an operating means 8, such as a minicomputer, wherein the unsharp masking is performed after obtaining the unsharp mask signal Sus. The processing of the image is mainly to make the ultra-low frequency, i.e. blurry, mask out. In addition to unsharp masking, it changes the coefficient β for highlighting, grading, 79 0 9 2 7 6 η the image. reduction, flattening and the like performed to further improve the diagnostic inertia and accuracy of the final resulting image.

The unsharp masking is performed by performing the operation represented by the formula: S '= Sorg + β (Sorg - Sus}. (6}

The unsharp mask signal Sus obtained by the method described below must have a modulation transfer function of not less than 0.5 at the spatial frequency of 0.01 periods / 10 ma and not more than 0.5 at the spatial frequency of 0.5 periods / month, preferably nien less than 0.5 at the spatial frequency of 0.02 periods / mm and no more than 0.5 at the spatial frequency of 0.15 periods / mm. Rather, to perform the operation of the preceding formula, the highlighting coefficient must be determined. These values are predetermined for the different parts of the human body or object to be diagnosed or determined case by case by external processing. When these values are predetermined for the different objects, these values 2C are stored in the memory of the unit used for signal processing.

The signal S 'obtained by the unsharp masking, as mentioned above, is further subjected to a smoothing to reduce the high frequency component. Smoothing reduces noise without damaging the information needed for diagnosis.

The unsharp mask operation is described in detail below with reference to Figures 2A-D.

Fig. 2A shows the frequency response when the image, recorded on the phosphor to be stimulated, is tested with ten pixels pixel / mm. It is known that the frequency response or modulation transfer function (MTF) is represented by a siae function when a rectangular weight aperture is used as the photodetector aperture, and a Gauss 35 function when an aperture with a Gaussian function weight is used 7909275 38.

Fig. 2B shows the modulation transfer function of a rectangular, unsharp mask I, and a Gauss unsharp mask II, which is not less than 0.5 at 0.01 periods / mm and not more than 0.5 at 0, 5 5 periods / mm. In the case of the unsharp mask of curve I, the unsharp mask signal was calculated by obtaining an arithmetic mean of about 63 pixels x 63 pixels (represented by the dimension N = 63) on the phosphor to be stimulated, tested with ten pixels. (pixel) / mm. This is equivalent to the case where the image on the phosphor to be stimulated is scanned with a light beam having a cross-sectional size of 6.3 mm x 6.3 mm. In the case of the unsharp mask of curve II, the unsharp mask signal was calculated by obtaining a weighted average with a Gaussian distribution weighting coefficient. The other factors were all equal to those, ge. in the case of the curve I. Tests showed that, in terms of diagnostic efficiency and accuracy, the results were almost the same for the two unsharp masks I and II with a different shape of MTF in its high frequency range.

Fig. 2C shows the modulation transfer function of the processed signal from Sorg-Sus.

Fig. 2D shows the result of the operation at I, corresponding to the signal S ', wherein the coefficient for punching out is fixed at 3 · As a result, the maximum value B, as shown, of the MTF of the output image signal is , about k, 6 times larger than the value A of MTF, which value is a limit value, the spatial frequency being infinitely close to zero. The broken line II in Fig. 2D shows the MTF in case the 5 pixel x 5 pixel smoothing is applied to the signal S 'shown in Fig. 2D.

Fig. 3A-D show the change of the coefficient ft for highlighting as a function of the brightness, represented by the original image signal Sorg or the unsharp mask signal 35 Sus. The brightness represents the image signal, i.e., the amount of light 7909276 19 emitted by the phosphor when stimulated.

Fig. 3A shows a flat graph with a constant value. Fig. 33 shows a gradually increasing graph icb "" = 0;, FIG. 3C shows a gradually decreasing graph (β '= 0}, and FIG. 3D a graph with an average brightness output. The graph, such as shown in fig.

3C, can also be called a graph with a low output brightness, where these two graphs contain a range where-10 at β '^ 3. The werandring. of the highlighting coefficient β, shown by FIGS. 3B-D, has a stepped graph, as indicated by curve a, and a gradually curved graph, as indicated by curve b.

By changing β as a gradually increasing function, as shown in Fig. 3E, it is possible to prevent the formation of an art image that may appear when the frequency is decayed. As an example of this, when the X-ray image of a stomach obtained using barium sulfate as a contrast medium is subjected to frequency decay (amplification of certain spatial frequency components) or hex masking, the coefficient for doing has been recorded, the boundary of the region with a regular low brightness over a wide range, corresponding to the portion containing the barium sulfate contrast medium, has been over-emphasized, whereby an art image with a double contour appears. If the coagulation coefficient ƒ3 is changed to be made smaller in the low brightness region for the portion filled with the contrast medium, and made large in the high brightness region for the stomach details and the like , the occurrence of the double contour art image can be prevented. In the case of the chest front image, if β is captured, further the noise increases in the low brightness area, such as the spine and heart, in an extreme case the fine portions are saturated only in white (the veil level of the record 7 9 Ö 9 2 7 δ ♦ 20 recorder & ium), which seriously interferes with visual observation and significantly reduces diagnostic efficiency and accuracy. On the other hand, if β is made small in the low brightness areas, such as the spine or heart, and large is made in the high brightness area, such as the lung, said noise and the saturated white areas can be reduced.

The brought out low brightness, as shown in Fig. 3C, is suitable for the object, the diagnosis of the low brightness portion is of particular importance, and the area of the low brightness portion is not an main body of the whole image. For example, the angiography or the lymphography is preferably subjected to frequency enhancement, because in these objects it is desirable that the sharpness of a given portion be greatly improved, even if noise is slightly increased. Thus, the diagnostic efficiency and accuracy are greatly improved in such objects by highlighting the low brightness.

Highlighting the moderate brightness, as shown in Fig. 3D, is suitable for the object, the diagnosis of the medium brightness portion being of particular importance, and the low and high brightness portions making up a large portion of the whole picture, and are not important for diagnosis. For example, the cholecystography or liver is subjected to doing. frequency of this kind, because in these objects it is desirable that only the moderate brightness portion be brought out, and the noise and air portion, which reduce diagnostic efficiency and accuracy, are not brought out .

In any previous example, if the coefficient β for 30 is set to a small value for frequency boiling, the diagnostic efficiency and accuracy are not improved because the contrast of the important parts, such as the stomach details, the blood vessels of the lung and veins is not improved, although different art images can be avoided. Thus, by continuously changing the 7909276 21 coefficient β * to reflect the brightness of the image on the phosphor to be stimulated, it is possible to obtain a radiation image with high diagnostic efficiency and accuracy, and arrange for.

Λ 5 Fig. ^ shows another example of the change of p according to the original image signal Sorg. In this example, jb is changed almost linearly between the maximum brightness and the minimum brightness SQ, which are obtained from a histogram of the image on the phosphor to be stimulated. The maximum and minimum 10 values S, and S_. be processed according to the type of X-ray image to be refreshed. For example, the maximum and minimum brightnesses can be determined as the brightness, with the integrated histogram becoming 90 to 100 $, and 0 to 1Q%.

Figures 5 and 6 are graphs for showing examples of the change of jb to highlight the low brightness and highlight the moderate brightness.

In fig. 5 β decreases from the maximum value β max to the. minimum value jb min, with the brightness changing from A to B. In the low brightness range (from Smin to A), the coefficient ƒ3 for highlighting has been maximized (y3 max) and area with the high brightness (from B to Smax), the coefficient ƒ3 for highlighting is minimized (ƒ5 min). The brightness A is preferably the sum of the minimum brightness (Smin) and 0.2 tet 0.5 times the difference (Δ S) between the maximum brightness (Smax) and the minimum brightness (Smin), i.e.

j ^ Smin + (0.2 .... 0.5) x Δ. s]. The brightness B is preferably the sum of Smin and 0.7 to 1 times AS, i.e.

£ smin + (0.7 .... 1) x Δ S J.

In Figure 6, as shown by the solid line a, 30 ƒ3 increases from the first minimum (ft mini) to the maximum (ƒ5 max) between the brightnesses A and B, and decreases from the maximum (ƒ3 max) to the second minimum (β min2) between the brightnesses C and D. In the low brightness area (Smin to A) and the high brightness area (D to Smax), in other words, the coefficient ƒ3 for the highlighting (ƒ3 m £ nj s ƒ3 min2), 7909275 22 where in the area of moderate brightness (B to C), the highlighting coefficient is increased (ƒ3 max). The first minimum value (| 3min1) and the second minimum value (min2) can be equal to each other. In the case of the change, 5 as shown by the dashed-dotted line b, which change is different from the aforementioned change, shown by the solid line a, the highlighting coefficient ƒ3 increases between A and E, and decreases between E and D. In Fig. 6, the brightness A, B, C, D and E is preferably the minimum brightness (Smin) plus 0 to 0.2 times the difference S) between the maximum brightness (Smax) and the minimum brightness (Smin), ie Smin + (0 ... 0.2) x ^ S, the mean brightness (S g '. Or a statistical average) minus 0 to 0.2 times the difference (Λ S), ie S ~ (0 ... 0.2) x Δ S, the mean brightness (S) plus 0 to 0.2 times the difference (Δ S), 15 i.e. S + (0 ... 0.2) x LS, the maximum brightness (Smax) minus 0 to 0.2 times the difference (Δ S), ie Smax- (0 .. .0.2 | xAS, and the mean brightness S, respectively.

In the foregoing operation using changing the highlighting coefficient, as shown in FIGS. 6 and 6, the maximum and minimum brightnesses (Smax,

Smin) both the maximum and minimum brightnesses in the image required for diagnosis, i.e., there may be brighter or less bright areas outside the significant image in the phosphor region to be stimulated. If desired, the maximum and minimum brightnesses can be selected as the maximum and minimum brightnesses throughout the range of the phosphor to be stimulated.

In accordance with investigations, it has further been found that the results are substantially equal between the coefficient f3 for decay, as changed with the original image signal, as changed with the unsharp mask signal.

In addition to highlighting the aforementioned frequency enhancement by using the unsharp mask, it is possible to provide a grading for changing the grading of the image. When grading is performed prior to unsharp masking, A / D converting is performed after 7909278 23 the signal is graded with a non-linear analog chain. When the grading is carried out after the unsharp masking, the grading can be further carried out in digital form or in analog form after the D / A conversion. Furthermore, it is possible to perform grading in the digital form after A / D amping, prior to unsharp masking.

The data which has been subjected to hatching the frequency and further grading, if required, are recorded on a magnetic ribbon 7. The data recorded on the magnetic ribbon 7 is read out and converted into a analog signal through a B / A converter 9, each analog signal is used to modulate a recording light source 11 after amplification by an amplifier 10. The recording light emitted from the light source 11 exposes a recording film 13 mounted on a drum 1¼ through a lens 12. The drum 1¼ is rotatable about its axis and movable in the axial direction A radiation image, subjected to the expression of the frequency of the unsharp masking, is thus recorded on the film 13 · The image finally recorded on the film 13 is used to perform the diagnosis.

Finally, when the image is reproduced on the photographic film, a reduced size image can be obtained by recording the image at a higher key frequency than the frequency at the time of the input scan. For example, if the input scan system has a test frequency of 10 pixels / mm and the output scan system has a test frequency of 20 pixels / mm, the finally obtained image has a half-reduced size with respect to the original image size.

The resized image by 1/2 to 1/3 y size is indicative of further improving the diagnostic efficiency and accuracy because the frequency component required for diagnosis comes close to the frequency with the greatest 35 visibility, and it therefore seems as if the contrast for the 7905276 2k viewer has increased.

The preferred methods of performing the operation to obtain the unsharp masking signal are now described in detail.

FIG. 7 shows an example of a method of performing the operation of obtaining the unsharp mask signal Sus ·. Referring to Fig. 7, the output of the photo detector 21, which measures the light emitted by the phosphor to be stimulated after its stimulation, is amplified by an amplifier 22, 10 which performs a gain linear correction or band narrowing, such as a logarithmic conversion, to obtain an original image signal Sorg. The original image signal Sorg is supplied to a processing unit 23 for performing the unsharp masking of formula 15 (1) on the one hand, and on the other hand sent to an underpass screen 2k to obtain the unsharp mask signal Sus. In the lower pass sieve 2 ^, the analog value of Sorg is sieved, passing only the ultra-low frequency component thereof and then converted into a digital signal Si by an A / D 20 converter 25. The converted digital signal is used for calculating of an arithmetic mean value Sus = ^ a ^ * Si by a digital arithmetic circuit 26. The value obtained is applied to the processing unit 23 as the unsharp mask signal Sus. In this formula, a. Is a sweeping coefficient for the signal Si.

25 from the A / D converter 15. In the case of a simple * arithmetic mean, a ^ is made equal to 1 / U, where IJ is the number of scan lines counted in the subordinate scan direction over an unsharp mask. cover range.

As shown in FIG. 7, the original image 30 signal Sorg is supplied to the processing unit 23 in the form of an analog signal. Since this signal Sorg is obtained prior to supplying the unsharp masking signal Sus to unit 23, it is necessary to delay the input of the original image signal Sorg so that both signals Sorg and Sus 35 are simultaneously supplied to unit 23 Also, the initial signal signal Sorg may continue further, stored in a hung. after being converted to a digital value, and read from memory when used together with the blurry masking signal Sus. In any case, it is necessary to delay the input of the original image signal Sorg in the unit 23 by the time required to calculate the unsharp mask signal Sus through the lower pass screen 2b, the A / D converter 25 and the circuit 26, so that the signals Sorg and Sus are simultaneously applied to the beacon unit 23.

In the preceding circuit, the output of the photo detector 21 represents the strength of the light emitted by the phosphor to be stimulated when it is stimulated, changes the strength over a wide dynamic range. Accordingly, it is impractical to directly process the output signal that changes over such a wide dynamic range. It is therefore a concern that the dynamic range of the signal is narrowed. For example, the signal should preferably be converted to the value corresponding to the final optical density. In order to perform the narrowing of the output signal on this mode, 20. , giving the output signal narrowing logically, as is well known, for a light measuring circuit included in a camera. The logic narrowing can be performed by the amplifier 22, which, if necessary, performs nonlinear amplification or conversion of the signal.

The cutoff frequency of the lower pass screen 2b is selected by a relationship between the spatial frequency of 0.01 to 0.5 periods / mm or preferably 0.02 to 0.15 periods / mm, the pixel size (mm / pixel ) and the pixel frequency (pixel / s).

The cutoff frequency (6 dB down) of the lower pass screen 30 2b, denoted by f, (periods / s) is determined by other advantages hz by the formula: f (periods / s) = (periods / mm) xd (mm / pixel ) xn (pixel / s) (7) where the given spatial frequency equals f (perio-35 den / mm), the pixel size equals d (mm / pixel) and the 73 0 9 2 7 5 2 6 pixel frequency in the main scan equals n (pixel / s).

It is to be noted that the output Tan the photodetector 21 described as amplified prior to passing through the lower pass screen in the previous example shown in FIG. 7 may be amplified after passing through the lower pass screen as shown. in FIG. 8. In the example shown in FIG. 8, the output of the photo detector 21 is divided into two outputs, one of which is supplied to the lower pass screen 2b, and the other to. an amplifier 22a, which, if necessary, performs a non-linear correction, such as the logarithmic narrowing. The output of the amplifier 22a is taken as an original image signal Sorg, on the other hand, the output of the lower pass screen 2V is fed to another amplifier 22b, equivalent to the amplifier 22a. The output of the amplifier 22b is supplied to the calculation circuit 26 to obtain an arithmetic mean value Sus * £ a.Si through an A / D converter 25. The output of the calculation circuit 26 is the unsharp mask signal Sus and is supplied to an unsharp masking processing unit 23 along with the original image signal Sorg using the formula: S '= Sorg + ft (Sorg - Sus).

Fig. 9 shows the pixels and the use of the ten-digit array, which is preferred, in calculating the blurry masking signal, in accordance with another preferred embodiment of the invention.

Referring to FIG. 9, it is assumed that an unsharp mask is rectangular, as indicated by a solid line enclosed by two parallel lines extending in the main scanning direction and two parallel lines extending in the subordinate extend scan direction. In the figure, the main scanning direction means the horizontal scanning direction. The subordinate scan means, of course, the vertical scan. In order to simplify the following explanation, the unsharp mask is assumed to be square. The length 7909278 of one side of the square mask is N with respect to the number of pixels, where H is a positive odd number. The on-mask mask M * T is calculated for the signal S'_ based on all the original image signals enter the image elements contained in the mask M-S 'T is the final signal X yV X jd sew to be obtained by the formula [sT = Sorg + β (Sorg-Sus)] for a scan point (pixel) in the center of the mask. ST_ is the original image signal J.,0 for the pixel P_T at the top of the mask Na

I, J J-, cJ

10 getting 5, _, it becomes possible to calculate the

x, tT

know blurry mask. T_ m is the total sem of all signals of i »u 2 the picture elements within the mask MT T with a number of d, d.v.Z. : j t TIJ = Σ Σ s ... (8)

*. j = Y-N + 1, 1 = 1-11 + 1 J

r

First, the signal S_ of the relevant picture element becomes PTT

Xti XJ

stored in the corresponding address of the sum S in the memory.

Each address must have a number of bits, which can indicate the signal value of the picture element, for example 8 bits.

Then the sum C__, IJ of the signals of the H pixels m 20 becomes the direction of the main daft ast and, represented by the formula

T

Cu - r s. . o) ^ i = I-N + 1 ^ obtained. This can be obtained by the formula: CIJ = CI-1, J + SI, J “SI-IT, J ^ (10)

by application of the sum C, .- '_ of the signals of the II image "i", <J

2 ^ elements, located for the picture element Pjj in the line of the element P ^ j, the signal Sj_jj j of the picture element, which is located on H picture elements for the relevant picture element PTT, and the signal S_j of the picture element P ^ j . The sum is stored in the corresponding address of the sum C in the memory.

Each address of this memory needs a number of bits, required to prevent overflow, which depends on the number N.

2

Then the total sum of the signal from the H image 7909278 28 elements within the mask 'Mjj-j' is obtained. This can be obtained by the formula:

TIJ = T1, J-1 + CI, J ~ CI, tf-N

by applying the total sum TT _ Λ when the signal of the 2 i5U — Ί 5 Ir pixels within the mask j -] »which is a line backward in the subordinate scan direction from the mask jt containing the pixels j, the sum ^ ^ of the signals of the H pixels in the last line of the mask M_ _, which is not located in the mask MTT, and the sum C__ of X ^ J ^ j XJ Xd 10 the signals of the pixels in the upper line, which contains the picture element Pjj-. The obtained value T ^ j is stored in the corresponding address of the total sum T in the memory. Since the 2 total sum T ^ j has the value, which is H times greater than the unsharp masking value, the unsharp masking can be performed using this value with the formula: ^ -I - ¥ · j- ¥: * p <ν ¥ · '- ¥ - ^> · <12>

The capacity of the memory required for the previous operation is described below. Fig. 10A shows a memory for the H + 1 'signal S of S1, which must have words in the subordinate scan direction, and Ito words in the main scan direction. Hm is the number that is equal to or close to the number of all pixels in the main scanning direction. For example, a word from this memory can have 8 bits. Fig. 10B shows a memory for the sum of CTT, which memory must have H + 1 words in the appropriate scanning direction, and Hm words in the main scanning direction.

A word from this memory must have two or three times as many bits as the previous spokesperson S. ^. Fig. 10C shows a memory for the total sum T ^ j., Which memory must have only two words in the subordinate scan direction, and Hm words in the main index.

Fig. 11 shows an example of the block diagram circuit for performing the previous operation. The original image input signal S1, supplied to a gate 31, is passed to the memory 32, provided with said capacity, and unbeaten therein. On the basis of the stored information, a processing unit 33 performs the processing. The gate 31, the memory 32 and the processing unit 33 are controlled by a control circuit 3¼.

The results of the processing by the processing unit 33 pass as output of the gate 31 through the memory 32 as an image output signal S1.

In accordance with the prior processing method, the operation for obtaining the unsharp mask signal Sus is considerably reduced, so that the device for stirring the operation is also greatly simplified. This simplification is based on the method of applying the rectangular mask and obtaining a simple arithmetic mean of the signals within the rectangular mask. In accordance with this method of calculating the simple arithmetic mean of the signals within a rectangular mask, in other words, a considerably simplified ten-digit system can be used, as explained above, whereby the operation can be considerably simplified. The processing of the radiation image can therefore be applied very simply.

Furthermore, in the foregoing explanation, the memories for the three types of information 35, 36 and 37 may be divided in the manner shown in Fig. 12, so that the address main and data main lines are divided into three groups, and three types of information, can be obtained simultaneously. Furthermore, as shown in Fig. 13, it is possible to daisy chain the three memories so that the addresses in the three memories are continued in series. In the example shown in Figure 12, the processing time is further shortened.

The control circuit and the processing unit can be designed as a specific logic circuit, such as a PDA. (logic circuit to be programmed} or a logic circuit with random access. A microcomputer or a minicomputer can also be used for the control circuit and the processing unit. For example, a bit-part microcomputer can be used for the control circuit, and one in particular designed circuit for the processing unit 7909276. In the practical chain, a suitable circuit is selected according to the required connection speed.

Another ten-digit system, which allows for a further reduction in the capacity of the memory, is described below with reference to Figures 1Us 15 and 16.

According to this ten-part system, after the signal STT

of the involved upper picture element PT in the unsharp mask

XJ

MIJ is stored in the corresponding address in the memory for 10 S, the sum E ^ j of the signals of the N pixels in the subordinate scanning direction, i.e.

E1> J = ('3) calculated and stored in the corresponding address in the memory for E. This calculation is performed using the formula: EI, J = EI, J-1 + SI, J “SI, J-H. (14)

By applying the stored values, the total sum becomes. . 2 T_ _, equal to IT times the unsharp masking signal value,

X 9 V.

got by applying the formula: 20 TI, J "TI-1-J + EI, J" EI-IT, J "

By this method, it is impossible to perform the calculation of the formula ("15") when the main axis starts and returns from the right end to the left end. Therefore, the sum of the

I signals ST, from the left side of the main scan line, which becomes X 5 J.

__ indicated by R_, represented by the formula: O «

ÏT

HT = t s. . . (16) J i = 3 1.0 first calculated, and stored .in the corresponding address in. the memory for R. De H. is shown, as shown in Fig. 6

O

give as R ^, i.e., the sum of up to ^ and R ^, i.e., the sum of S. _ to S_ ς, when ï 1 = 5, for example. When the pixel concerned changes from S_ to Sr, the 7909278 3α does not change.

therefore, when the main scan from the right end returns to the line end, the TT_ is obtained by incrementing the R_ by the formula; V '+ -P-H. fl7)

5 I, vT “J U-IT

By applying the obtained "- the blurry masks are performed by the fornnile:

This decade system needs a memory for the signal S1 which has N + 1 words in the minor 10-µ scan direction and Sm words in the main scan direction, as shown in FIG. 15A. However, this decade system requires only very small memories for Η, E and T, as shown in Figs. 15B, 15'C and 1 $ D. The memory for R and E requires N + 1 words, the memory for T needing only two words. For example, one word of the memory used for S needs only 8 bits, but one word of the memory used for R, E and T needs, for example, 16 bits, which depends on the size N. The memory capacity The 22 gene, which has a longer bit length, is therefore reduced, the ten-digit system having the great advantage that the total memory capacity is very small. Thus, the capacity of the memory shown in Figs. T5A-D is much smaller than the capacity of the memory shown in Figs. 10A-C, which is effective to simplify the entire device for performing the method of processing a radiation image.

The previous two dozen systems are of the digitally processing type, the signals being processed in digital ^ 2 form. However, it is possible to integrate the analog signal in the main scan direction and store the integrated value in the memory, and then perform a numerical integration of the stored values in the minor scan direction to obtain the unsharp masking signal Sus .. Because in this case the analog value is integrated, for each picture element, N analog integration chains are needed. However, the number of analog integrating chains can be reduced to only one by using the following method.

That is, the analog output Sorg from the scan point is divided into two signals, one of which is delayed by a delay circuit. The delayed signal and the other signal are applied to a signal difference enhancement circuit which provides an output 10 representing the difference between two input signals (Sorg-TSorg). The delayed signal TSorg is a signal having a delay time T which is represented by the product of a scanning time £ of a picture element and the number of picture elements H in the unsharp mask counted in the main scanning direction, d.v. z. T = F Ή. The output of the signal difference enhancement circuit is integrated to obtain the total sum of Sorg by the formula: t t tf-H. Γ t ƒ (Sorg-TSorg) = J Sorg - J Sorg = ƒ Sorg (19) ~ OO ~ S3 t-H. Γ

The integrated value corresponds to the value C shown in the figure. 9 and 10, which are added in the subordinate scanning direction by a digital operation to obtain the skill T_ T. Then, using the skill T_ T, the on-X 9 VX 9 scherpe sharp mask Sus · is obtained, as mentioned above . This is also a method, which can easily calculate the desired value Sus at a high speed and, i.e., a preferred mode of operation as an analog operation.

Furthermore, the blurry mask Sus (IJ) is a skill obtained from the signals S-. inside the mask, provided with a scanning point (ij), in the center thereof, the scanning point and areas being covered in a region of VI <i <\ (20a) 2 2 * -Il-1 .l-1 "V <S <- - (20b) 7909276 33 where hst is the number of pixels in the main scan dating, and 17 is the number of pixels in the sub scan 7 direction, no therefore it is impossible to obtain the blurry masking signal from a scan point at the edge of the image, because certain signals around the scanning point at the edge of the image have not been determined.

In order to obtain the blurry mask signal Sus for the scanning point at the edge of the image, it is the simplest and most economical method to store the signals of the outer image elements (i.e. image elements at the edge of the image} and the stored use signals for the imaginary picture elements around the image, assuming that the signal of the outer picture elements is the same for the imaginary picture elements around the image, or it is possible to assume that the 15 imaginary picture elements around the image are shown as black or white, or as an intermediate value between black and white.

In the foregoing embodiments, the original operating signal Sorg further includes the signal which has been subjected to the band narrowing and / or a nonlinear correction such as the logic narrowing. In practical use, it is desirable that the original image signal is a signal subjected to a band dimming, such as logarithmic narrowing and the like, because the signal is the output of the photodetector, which output represents the brightness. It is of course possible to use the output signal of the photo detector without any processing directly as Sorg. Theoretically, the calculation of the unsharp masking signal can also take place on the basis of the signal, which represents the energy itself. According to investigations, however, it has been found that the mean value obtained from the logarithmically narrowed value, corresponding to the density and not to energy, shows the same results in terms of diagnostic efficiency and accuracy.

In practice, this is very easy and advantageous when carrying out the operation.

The invention is not limited to the previous embodiments, 7909275 3k, but may be embodied in various variations.

The readout of the image in the phosphor can be performed by using a rotatable drum to which the phosphor plate is mounted or by using a flat support movable for scanning and on which the phosphor plate is mounted. The phosphor can also be scanned optically by scanning the laser beam. Also, the reading can be performed using a beam scanning system, such as a light spot scanner. Although, in the foregoing embodiments, the digital output of the A / D converter 6 is first stored on a magnetic ribbon, and the aforementioned processing is performed on the basis of the stored output, it is further possible that the signal is at the real time and send the processed signal directly to the display station. Furthermore, the processing of the unsharp mask signal can be performed separately after recording the necessary information on a magnetic ribbon or in line with the information temporarily stored in a core memory.

The invention is further elucidated by means of a number of examples.

EXAMPLE I

More than 100 samples of conventional radiation images from different parts of the human body were examined in both a conventional radiogram and a radiation image form recorded on a recording medium by the present radiation image processing method. More specifically, the diagnostic efficiency and accuracy between these two types of images were compared. To investigate the various present factors, the coefficient of expression β and the spatial frequency fc, at which the modulation transfer function equals 0.5, were changed in different measures. As the unsharp mask, a circular region in which the image signals were averaged using a Gauss weighted average was used.

The results were evaluated by four radiologists, 35 because it was impossible to evaluate the diagnostic effectiveness and accuracy of the 79 0 9 27 5 35 by physically assessing the objective using sharpness, contrast and grain.

The nom of the assessed tests was as follows.

+2: The diagnostic efficiency and accuracy are greatly increased and improved. The diseased parts, which were not recognizable in a usual radiogram, had become recognizable or the diseased parts, which were very difficult to recognize, had become very clearly recognizable.

: Φΐ: Diagnosis efficiency and accuracy were improved. 10 For example, the diseased parts, which were difficult to recognize, had become recognizable.

0: The diagnostic efficiency and accuracy were not greatly improved, although the image was slightly clearer.

-1: Diagnosis efficiency and accuracy were reduced in certain sections, and improved in other sections.

-2: Diagnosis efficiency and accuracy were reduced without parts improving.

Fig. ITA-3 show the results of the relationship between the valuation of the obtained image and the frequency fc, whereby the modulation transfer function became 0.5. Fig. 17A shows the examples of the image of the front of the breast, FIG. 17B showing the examples of the bones. The thin, solid line (1) shows the results where the coefficient | $ for the hatching was fixed at p = 3. When comparing Fig. 1JA with Fig. 25 173, it can be seen that the frequency range, where the valuation is high, has shifted down in the examples of the front of the chest compared to the leg examples.

Thus, it has been found that the frequency components to be brought out depend on the type of disease or the portion of the human body. The broken line (il) shows the results, where j *? was continuously changed according to the original image signal. In both examples, the range of the high value determination was extended to both the lower and higher frequency ranges. This is because in Fig. 17A with 35 white saturated areas (saturated to the veil level of the 7902276 3 6 recording medium) at the heart and leg portion (including the spine) disappeared, with Fig. 17B showing the increase in noise appearance.

In the example of the breast, the coefficient ^ for highlighting 5 was changed so that it was set to zero at the brightness Sq, with the integrated histogram becoming 10% (equivalent to the maximum brightness at the spine) and set to 3 at the brightness, where it became 50% (equivalent to the minimum brightness at the lung), and changed linearly between them.

The dashed-dotted line III shows the results, adding the grading to the previous method so that the contrast of the heart was lowered, and the contrast of the lung was increased in Fig. 17A, and the contrast was increased as a whole. up to 1.5 times the original contrast in Fig. 17B.

15.

The solid, solid line (IV) shows the results, with the size of the image reduced to one-half or one-third in addition to the previous methods.

When graded, the disease, which shows a gradual change in contrast over a wide area, such as a lung-20 cancer or a muscle tumor, was clarified. As the image size was reduced, the ultra-low frequency components important for diagnosis were brought closer to the optimum frequency of modulation transfer function for human eye sensitivity (one to two periods / mm), where the contrast was found to have improved, as did the diagnostic efficiency and accuracy.

Further, when smoothing was performed to make not less than 0.5 of the modulation transfer function at the spatial frequency of 0.5 periods / mm, and not more than 0.5 at the spatial frequency of 5 periods / mm In addition to bursting the ultra-low frequency components described above, the noise (graininess) of the image was removed, and the diagnostic efficiency and accuracy improved.

Fig. 1'8 shows the relationship between the valuation and the degree of hatching, represented by the ratio B / A in a breast. In this case, the frequency range to be extracted is fixed at fc = 0.1, the coefficient β for hatching being changed in various ways. Curve a in Fig. 18 shows the results where β was recorded regardless of the original image signal, curve b shows the results where ƒ3 was continuously changed with the original image signal. The B / A ratio is the maximum ratio of B / A.

In the curve a, where β remains the same, the value determination falls below zero as a result of. an art image when the ratio 3 / A becomes more than 6 or 7. In the curve b, where β is changed, the art image disappears, the valuation being above zero over a wx ^ d range of 1.5 = B / A = 10. For the other examples, essentially the same results were observed.

Table I shows the range of fc where the rating was improved or above zero for other samples. The frequency fc is from the spatial frequency, measured on the phosphor plate.

TABLE I

Portion of sample Fr equent i eb er e ik (fc: period / mm) Front of chest 0.01 - 0.2

Chest side 0.01 - 0.05

Leg (incl. Muscles) 0.05 - 0.5 3 Sores (calcification) 0.1 - 0.5 3 Sores (cancer) 0.01 - 0.1

Blood vessel 0.1 - 0.5

Stomach 0.1 - 0.5 25

As shown in the previous table, the frequency range relevant to diagnosis was found to be divided into the x * -c, ultra-low frequency range around the range of 0.01 = fc = 0.5 periods / mm.

It was further confirmed that the diagnosis was further improved by the hatchery assembly. ultra-low frequency and other operations, such as changing the coefficient β for scaling, grading, image size reduction and smoothing in all previous parts or diseases.

EXAMPLE II

- e: · 3 01 5 g *, V hm 2 w 38 200 samples for the parts, as shown in Table II, were examined both in the form of a conventional radiogram and in the form of the radiation image obtained according to the invention . In particular, the diagnostic efficiency and accuracy between these two types of images were compared.

TABLE II

Normal picture: Front of chest, side of chest, lower abdomen, leg, head, breasts

Contrast image: Stomach with double contrast, blood vessel (angiography), 10 veins, lymphography.

Tamographic image: Chest, lower abdomen.

In the display image processing operation, the highlighting coefficient β was fixed at 3, using a rectangular area as the unsharp mask to obtain a simple arithmetic mean of the image signals of the picture elements therein. The valuation was performed at six different spatial frequencies, at which the modulation transfer function became 0.5 (fc). The results were valued by four radiologists, twelve hospital dock-20 tower and four radiological technicians. These specialists subjectively assessed the images obtained on value. The standard of the valuation was the same as in example I. "

Fig. 19 shows the results of the valuation by the twenty specialists for the 200 samples, averaged to a simple curve in a graph of valuation, plotted against the spatial frequency fc, whereby the modulation transfer function became 0.5.

As shown in Figure 1.9, the range of fc, with particularly improved diagnostic efficiency and accuracy, was 0.02 to 0.15 period / mm. This investigation further revealed that the range of fc, where the results of the present processing were recognized as 'Enhanced' *, did not differ significantly for different radiation images, although the value of fc, with the value determination being highest, ie the peak of the 35 valuation, was slightly different, depending on the person who performed the valuation, just on value determined portion (part of the human body) or the disease, and the purpose of the study for the radiation image, i.e. a general mapping or an investigation.

EXAMPLE III

Conventional samples of the portions, as shown in Table II, were graded with fc set at 0.05 period / mm and B / A changed in various ways. Using the same method as in Example II, the present radiation image 10 was obtained and valued by 20 specialists, as in Example II. The average values of the valuation are shown in Fig. 20.

When, as shown in Fig. 20, β was recorded (curve a), the diagnostic efficiency and accuracy was improved in the 1.5 to 6 range of B / A, and in particular improved in the 2 range to 5.5- When β was changed (curve b), the diagnostic efficiency and accuracy was improved in the range of 1.5 to 10, and particularly improved in the range of 2 to 8.

EXAMPLE IV

The usual hundred samples of the portions, as shown in the following Table II, were valued with fb changing according to the original image signal or the unsharp masking signal, as shown in Fig. 3A-D. When the unsharp mask was used made from a simple arithmetic mean of image signals within a rectangular area.

As the frequency fe, the optimum frequency for all samples was chosen within a range of 0.01 to 0.5 period / mm by experimental approach. The evaluation of the obtained images was done by the same method as used in Example I.

Valuation results are shown in Table III.

In Table III, A, B, C, D mean the results of the valuation of the image in the case that β was changed in the form of Fig. 3A-D. When the value determination in the case of j3 7802273 l + o was changed, as shown in Fig. 3B, was better than the value determination in case ƒ3 was changed, as shown in Fig. 3C, the designation of the value determination is shown as C ^ B as an example.

As shown in Table III, it was found that the value determination was higher in the case where ƒ3 was changed as shown in Figs. 3B, 3C or 3D than that in the case when () was recorded as shown in Fig. 3B. 3A.

TABLE III

10 Sample Rank Value determination value (DE: diagnostic efficiency and precision determination)

Head A <C ^ B B: DE of brain tumor and facial muscles was improved.

Ordinary X-ray A-dB <C C: DE of cancer overlapped by spine

Chest Sf heart was improved.

A <C <B B: DE of cancer in the lung field and pneumonia was improved.

Breast A <B <C C: DE of cancer in mediastinum and

Tomography bronchitis was improved.

15 A <C ^ B B: DE of cancer in the lung field section was improved:

Leg, A <B <C. C: DE of fracture epiphysis line was improved.

muscle A <C <B B: DE of muscle tumor was improved.

Breasts A <B <C C: DE of cancer and calcification of breast glandular tissue was improved.

A <C <B B: DE of deviation from fat and skin was improved.

Blood vessel A ^ B <C C: Fine details of blood vessels (angiography) were clearly observed.

Lymphography A <B ^ C C: Lymph node and nodule were clearly observed and well reflected the anatomical structure.

Stomach A <B <D D: DE of stomach details was improved.

Double contrast

Cholecyst o- A <B <D D: DE of common bile duct and gallstone engraving was improved.

Liver A <B ^ D D: The internal structure of the liver was clearly observed.

'79 0 9 2 7 8 '

Onion

EXAMPLE V

Four samples for both the chest and the bones were evaluated for comparison of the ideal unsharp mask with the unsharp mask provided with a rectangular shape.

Sampling of the image in the phosphor to be stimulated was performed to the extent of 10 pixels / mm, using a 6 mm diameter circular mask to calculate the ideal unsharp mask signal by weighing the original image signal with a weighting coefficient, with a Ganss distribution over the mask. Another unsharp mask was obtained by scanning the phosphor in the main direction at a rate of 20 x 10 pixel / s, with the output of the image signal passed through an underpass screen with a cut-off frequency of 0.2 x 10 periods. / s. The signals passed through the lower pass screen were simply added to the subordinate scanning direction by a digital calculation (i.e., a simple arithmetic mean). The coefficient 0) for highlighting j was set to 2.

The results were evaluated by the same procedure as used in Example I by four radiologists. The results showed that there was no difference between the previous two types of unsharp masks in terms of their diagnostic effectiveness and accuracy.

25 - EXAMPLE VI

The hatching coefficient p was changed to U with all other conditions being the same as in Example V. The results obtained were essentially the same as those in Example V.

30 ^ -O C 9 7 λ

Claims (14)

A method for processing a radiation image in a radiation image recording system, wherein a stimulation phosphor is scanned with a stimulation beam, and the radiation image information recorded in the phosphor to be stimulated is read and converted into an electrical signal at stimulation thereof, after which a visible image is recorded on a recording material using. this electrical signal characterized by the step of performing an operation, represented by the formula: 10 S '= Sorg + fl · (Sorg - Sus), where Sorg is the original image signal read from the stimulus phosphor, is a highlight coefficient and Sus is an unsharp masking signal corresponding to an ultra-low spatial frequency at each scan point, bringing the frequency-15 component above the ultra-low spatial frequency. Method according to claim 1, characterized in that the unsharp masking signal corresponds to the modulation transfer function, which is not less than 0.5 at the spatial frequency of 0.01 period / mm, and not more than 0.5 at the spatial frequency of 0.5 period / mm. Method according to claim 1, characterized in that the unsharp masking signal corresponds to the modulation transfer function, which is not less than 0.5 at the spatial frequency of 0.02 period / mm, and not more than 0.5 at the spatial frequency of 0.15 period / mm. 1 *. A method according to any one of the preceding claims, characterized in that the coefficient β for hatching has a constant value. Method according to claim k, characterized in that the maximum value of the modulation transfer function of the finally recorded radiation image, brought out by the formula, is 1.5 to 6 times greater than the limit value of the modulation transfer, 909276 2 * 3 · bring function, where 'the spatial frequency is infinitely close to zero. o. A method according to any one of claims 1-3, characterized in that the coefficient for decomposition is changed with the original image signal or the unsharp mask signal. A method according to claim 6, characterized in that the maximum medulation transfer function of the finally registered radiation image, brought out by the formula, 1.5 to 10 times, is greater than the limit value of the modulation transfer -10 function thereof when the spatial frequency is infinitely close to zero. B. A method according to any one of the preceding claims, characterized by the step for performing a smoothing, wherein the microdulation transfer function is not less than 0.5 at the spatial frequency of 0.5 period / mm, and not more than 0.5 at the spatial frequency of 5 period / mm. Method according to any one of the preceding claims, characterized in that the unsharp mask signal Sus is obtained by passing the original image signal through a lower pass screen 20 Sorg in the analog form in the main scanning direction, and calculating an arithmetic; average of the signals passed through the lower pass screen in the digital form after A / D conversion in the subordinate scanning direction. Method according to claim 9, characterized in that the arithmetic mean is a simple arithmetic mean. Method according to any one of claims 1 to 3, characterized in that the unsharp mask signal Sus is obtained by calculating a simple arithmetic mean of the original image signals Sorg of the scanning points within a rectangular 3 ° region enclosed by two parallel lines in the direction of the main scan, and two parallel lines in the direction of the subordinate scan. 12. A method according to any one of the preceding claims, characterized in that the finally recorded image is reduced in size compared to the image on the phosphor to be stimulated. 7902275. Λ ΙΑ Apparatus for processing a radiation hero in a radiation hero recording system, which system comprises a stimulating beam source for emitting a stimulating beam, which scans a stimulating phosphor and thereby stimulates it by passing through the causing the phosphor to expand light according to a radiation world registered therein, further a photodetector for detecting the emitted light and converting it into an electrical signal, and a processing unit for processing the electrical signal, characterized, that the control unit performs an operation, represented by the formula S '= Sorg + β (Sorg - Sus) where Sorg the original image signal! when observed by the photodetector, ƒ3 is a highlight coefficient and Sus 15 is an unsharp masking signal corresponding to an ultra-low spatial frequency at each sensing point. A. Device according to claim 1U, characterized in that the processing unit comprises a lower pass screen, which sieves in the main scanning direction at the output of the photodetector in the form of an analog value, further an A / D converter for the A / D converting the output of the lower pass screen, and a digital calculator for calculating the arithmetic mean of the output of the A / D converter in the subordinate scanning direction to thereby obtain the unsharp masking signal Sus, corresponding to the ultra-low frequency at every observation point. Device according to claim 13, characterized in that the processing unit comprises a circuit for calculating a simple arithmetic mean of the original image signals within a rectangular area enclosed by two parallel lines in the main scanning direction and two parallel lines in the subordinate scan direction, to obtain the blurry mask signal Sus. 16. Device according to any one of claims 13-15, characterized in that the processing unit comprises a means for changing the coefficient / 3 for hatching in dependence on 7909276 ast original image signal Sorg or the unsharp masking signal Sus. 7909276