GB2329313A

GB2329313A - Selecting region of interest in renal scintigraphy

Info

Publication number: GB2329313A
Application number: GB9814174A
Authority: GB
Inventors: Yumi Tomaru
Original assignee: Individual
Current assignee: Individual
Priority date: 1997-07-02
Filing date: 1998-07-01
Publication date: 1999-03-17
Anticipated expiration: 2018-07-01
Also published as: JP3249946B2; GB2329313B; GB9814174D0; JPH1172568A

Abstract

To assist in calculation of radiopharmaceutical uptake, a gamma camera image of a renal area is displayed on a computer. An operator then "clicks" on a pixel of the image at or near the centre of the image of the kidney. A first area around the pixel is automatically selected, and a maximum count within this area is used to calculate a threshold which is then used to select pixels within a second area around the central pixel to form a central renal area. If there is continuity between two parts of this central area due to a single pixel, this pixel is disregarded, and only the part of the area containing the original central pixel is taken as the renal central area. Finally, a peripheral area is determined either as a region of predetermined width around the central area or by using a second threshold, and the central and peripheral areas are added together to define the region of interest.

Description

METHOD OF SEMI-AUTOMATED SELECTING RENAL REGION OF INTEREST IN SCINTIGRAPHY BACKGROUND OF THE INVENTION Field of the Invention The present invention relates in general to the quantitative analysis of renal function using renal scintigraphy (one field of clinical nuclear medicine) in which the area in a scintigram corresponding to the kidney (renal region of interest; renal ROI) is assigned in order to calculate renal uptake per injected dose of radiopharinaceutical (% injected dose; %ID), and more particularly, to the assignment of the renal ROI.

DESCRWTION OF RELATED ART Renal scintigraphy can be divided into static imaging and dynamic imagine. The former is usually obtained some three hours following intravenous injection of a radiopharmaceutical which is taken up by renal parenchyma and remains there. Static imaging can reveal anatomical inforrnation, e.g., renal position, size, shape, and distribution of functional tissue. Dynamic imaging is performed using a radiopharmaceutical excreted by the kidney. Following an intravenous bolus injection of a radiophnnmaceutical, dynarnic imaging is accomplished by sequential renal images obtained over a few second interval. These images provide not only anatornical information but also data concerning renal function such as, secretion and excretion.

The quantitative methods for investigating renal function using dynamic renal scintigraphy (camera-based method) are widely used because neither urine collection nor blood sampling is necessary and split renal function can be calculated.

Camera-based methods have already been developed to calculate the glomemlar filtration rate, effective renal plasma flow and technetium-99m mercaptoacetyltriglycine (Tc 99m-MAG3) clearance based on 7cID of technetium-99m diethylene triarnine pentaacetic acid (Tc-99m-DTPA), iodine-l 31 onho-iodohippurate (I-131-OIE) or Tc-99m-MAG3 in the kidneys, respectively.

The quantitative values obtained using the camera-based methods are influenced by several factors. Porential improvements have been suggested by several investigators. Taylor, et. al. described a camera-based method to calculate Tc-99m-MAG3 clearance which contained a number of optirnized features, including a background region of interest, better estimation of renal depth, a more appropriate attenuation coefficient, correction for dose inhltration and table attenuation and correction for disparities between starting the camera and injecting the dose (J Nucl Med 1995;36:1689-1695).

In nuclear medicine the area in the scintigram corresponding to the target organ for qualitative or quantitative analysis is defined as a region of interest (ROI). Most commercial software programs make the aforementioned measurements based on total count in renal ROl per injected dose of radiopharmaceutical (% injected dose; %ID) of Tc-99m-DTPA, I-131 OIH or Tc-99m-MAG3 in the kidneys at 1-2 or 2-3 min. post-injection. Therefore, interoperator variability in the assignment of the renal ROI is a critical factor.

Conventionally, the operator assigns the renal ROI by tracing renal contour manually with a track-ball, light-pen or mouse on a displayed computer image. Therefore, there is a possibility that the %ID obtained by different operators can differ significantly Itoh, et al. reported that the %ID calculated by three operators did not differ significantly if there was general agreement arnong them in advance with respect to assignment of renal ROI. By contrast, Halker, et al. reported that inter-operator variability in assignment of the renal ROI represents a potential source of error in calculating the relative and absolute renal uptake of Tc-99m-MAG3. They indicated that the difference in %ID calculated by two technologists was significant, although there was a good correlation. They also reported that inter-operator variability was improved when using a semi-automated renal ROI selection method with a single-threshold technique (single-threshold method) (J Nucl Med l996;37(suppl):293).

Halker's single-threshold method was as follows. The operator set a large elliptical area around the kidney on a scintigram displayed on a computer monitor. The following steps were then performed automatically. The maximal count in this area was calculated.

The renal ROI was composed of the pixels with a count exceeding the threshold based on the maximal count (for example, twenty percent of the rnaximal count). The pixel value in the scintigram indicates radioactivity of the corresponding position.

However, even if there is general agreement among the operators in advance with respect to manual assignment of renal ROI, there is a possibility that the %lD obtained by different operators would differ significantly. Moreover, it is difficult to delineate manually the kidney with poor function due to poor contrast between the kidney itself and the adjacent organs. In such cases, differences of %ID among operators would be even more pronounced.

In our study, the operator with the least experience tended to strictly trace the renal contour for assignment of renal ROI. In the process, the peripheral renal count was lost because the commercial ROI program in our system worked by adding the values of pixels completely inside the traced contour to obtain the ROI's value. A larger ROI setting is necessary to prevent an underestimation of the entire renal count or the renal function.

Halker, et al. reported that inter-operator variability was improved when using the single-threshold method. However, obtained renal ROI is underestimated when a pixel with a value exceeding the mairnal renal count is present in the set ellipse. On the other hand, when the threshhold is set low enough to include the entire kidney, renal detectability is low.

Taking the above problems into consideration, the inventors herein developed a new serni-automated renal ROI selection method. This method improves the inter-operator variability in assignment of the scintigraphic renal ROI, and it can more effectively detect kidneys with poor renal function, SUMMARY OF THE INVENTION The potential problems described above can be solved by using the semi-automated renal ROI selection method of the present invention which consists of steps one to six as follows: Photons from the radiopharmaceutical administered in vivo is collected by a gamma camera. Obtained images (scintigram) are transferred to the computer where each renal ROl is set by the serni-automated renal ROl selection method based on pixel values on a displayed computer image. First, the operator clicks around the center of the kidney on a displayed computer image. (All the following steps are performed automatically). Second, the maximal renal count for each kidney is defined as the maximal count in the first area, with the clicked pixel as the center. The subsequent threshold technique is based on this maximal renal count.

The following calculation is based on the second area, where again the clicked pixel represents the center. Third, pixels with values higher than the first threshold in the second area are selected as the renal central areas. Fourth, if there is continuity between two areas in the renal central area due to one pixel, the pixel responsible (bridged pixel) is deleted to break the continuity. Fifth, only the area including the clicked pixel is newly selected as the renal central area. Sixth, an area 5-20 mm in width outside the renal central area is selected.

The pixels in this area with values exceeding the second threshold are selected as the renal peripheral area, and together with the renal central area they constitute the renal ROI.

BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 shows a flow diagrarn of the semi-automated renal ROT selection method of the present invention.

Fig. 2 shows the areas selected in steps one and two of the semi-automated renal ROT selection method of the present invention in a 1-2 min. image of a patient.

Fig. 3 shows the renal central areas selected in step three of the semi-automated renal ROI selection method of the present invention in a 1-2 min. image of a patient.

Fig. 4 shows the renal central areas newly selected in step four of the semi-automated renal ROI selection method of the present invention in a 1-2 min. image of a patient.

Fig. 5 shows the renal central areas newly selected in step five of the semi-automated renal ROI selection method of the present invention in a 1-2 min. image of a patient.

Fig. 6 shows the renal ROIs selected in step six of the semi-automated renal ROI selection method of the present invention in a 1-2 min. image of a patient.

Fig. 7 shows the method of the present invention to break the continuity between two areas in the renal central area due to one pixel.

Fig.8 shows the relationship of %ID obtained by operators A and B using manual ROI with monochromatic display in 1-2 min. images (n=59).

Fig. 9 shows the relationship of %ID obtained by operators A and C using manual ROI with monochromatic display in 1-2 min. images (n-59).

Fig. 10 shows the relationship of ID obtained by operators B and C using manual ROI with monochromatic display in 1-2 rniri. images (n=59).

Fig. 1 1 shows the relationship of %ID obtained by operator A using manual ROI with monochromatic display in 1-2 min. images on the first and the second trials (n=59).

Fig. 12 shows the relationship of 70D obtained by operator A using manual ROI with monochromatic and color displays in 1-2 rnin. images (n=59).

Fig. 13 shows the relationship of %ID obtained by operator A in 1-2 min. images using manual ROI witb. color display and semi-automated ROr (the first and second thresholds are 60% and 30%, respectively) (n=59).

Fig. 14 shows the relationship of %ID obtained by operator A in 1-2 min. images using manual ROI with color display and semi-automated ROI (the first and second thresholds are 60% and 20%, respectively).

Fig. 15 shows the relationship of %ID obtained by operator A in 1-2 min. images using manual ROI with color display and semi-automated ROI (the first and second thresholds are 60% and 10%, respectively).

Fig. 16 shows the relationship of %ID obtained by operator A in 2-3 min. images using manual ROI with color display and semi-automated ROI (the first and second thresholds are 50% and 30%, respectively) (n=62).

Fig. 17 shows the relationship of %ID obtained by operator A in 2-3 min. images using manual ROI with color display and semi-automated ROI (the first and second thresholds are 50% and 20%, respectively) (n=62).

Fig. 18 illustrates the relationship of %ID obtained by operator A in 2-3 min. images using manual ROI with color display and semi-automated ROl (the first and second thresholds are 50% and 10%, respectively) (n=62).

Fig. 19 illustrates the relationships of %ID obtained by operators; A and B or C using the semi-automated renal ROI selection method (the first and second thresholds are 60% and 20%, respectively) in 1-2 min. images. The data obtained by operators B and C are identical (n=59).

Fig. 20(a) shows the unsuccessful ROI of a normal volunteer using the conventional single-threshold method with a 20% threshold value.

Fig. 20(b) shows the successful ROI of a said normal volunteer shown in Fig. 20(a) using the semi-automated renal ROI selection method of the present invention with 60 and 20% threshold values.

Fig. 20(c) shows the unsuccessful ROI of a patient with chronic renal failure using the conventional single-threshold method with a 20% threshold value.

Fig. 20(d) shows the successful ROl of said patient with chronic renal failure using the semi-automated renal ROI selection method of the present invention with 60 and 20% threshold values.

Legends in the Figures Pl: The first area consisting of pixels.

P2: The second area consisting of pixels.

Ml and M2: The renal central areas.

R: The renal peripheral area.

S: The renal ROI.

K: The bridged pixel.

a, : Clicked positions.

DETAILED DESCRIPTION OF THE INVENTION When utilizing the semi-automa.ted scintigaphic renal ROI selection method described above, the first area is the rectangular area 30-70 mm in width and 60-120 mm in height, and the second is the rectanguJar area 100-250 mm in width and 150-300 mm in height. The problems described above can be solved by using semi-automated renal ROI selection method which has the above features.

When utilizing the semi-automated scintigraphic renal ROI selection method of the present invention, in a preferred embodiment instead of step six, all pixels which are completely or partially included in the area 5-20 mm in width outside the renal central area described above are defined as the renal peripheral area, and together with the renal central area they constitute the renal ROI. The problems described above can be solved by using semi-automated renal ROI selection method which has this preferred feature.

In another preferred embodiment of the semi-automated scintigraphic renal ROI selection method, the first and second thresholds are fixed at 50-70% and 1-40%, respectively, of the maximal renal count on 1-2 min. image generated by the summation of images from 60 to 120 seconds after administration of Tc-99m-MAG3. The problems described above can be solved by using semi-automated renal ROI se]ection method which has the above features.

In a further preferred embodiment of the semi-automated scintigraphic renal ROI selection method, the first and second thresholds are fixed at 40-70=0 and 1-40S, respectively, of the maxinnal renal count on 2-3 min. image generated by the summation of images from 120 to 180 seconds after administration of Tc-99m-MAG3. The problems described above can be solved by musing semi-automated renal ROI selection method which has the above features.

In yet a further preferred embodiment of the semi-automated scintigraphic renal ROI selection method, instead of step four, steps one to three followed by five and six are performed. The problerns described above can be solved by using semi-automated ROI selection method which has the above feature.

When utilizing the serni-automated scintigraphic renal ROI selection method, it is preferred that the operator only clicks around the center of each kidney on displayed scintigram in step one. The following steps are performed automatically and assignment of the renal ROI can be completed in a very short time, compared with the manual method. The kidney is separated from the surrounding organs by completing steps two to five. Step six prevents underestimation of the renal ROI by including the counts of the renal peripheral area in ROt's value.

Implementation of the Invention The algorithrn of this invention, the semi-automated renal ROT selection method, is explained in detail using a patient's data as follows: The flow diagram for this invention, the semi-automated renal ROI selection method, is shown in Fig. 1. The selected areas for each step of this invention, the semi-automated renal ROI selection method, are indicated in Figs. 2-6 using 1-2 min. image of a patient.

Fig. 7 shows how to break the continuity between two areas in the renal central area due to one pixel.

Linear regression analyses of 70ID were performed by least squares method with several combinations of conditions. Relationships between them are indicated in Figs. 8-19.

When performing a quantitative analysis of the kidney using scintigraphy, photons frorn the radiopharmaceutical administered in vivo is collected by a gamma-camera. Obtained images (scintigram) are transferred to the computer, and quantitative analysis is performed based on the total pixel value in the renal ROI. Conventionally, each renal contour is traced manually on a displayed computer image with a track-ball, light-pen or mouse, and the area consisting of the pixels inside the traced contour is defined as the renal ROI. However, when using this invention, the semi-automated renal ROI selection method, which makes use of computer (hardware) consisting of a central processor unit, memory or display, only a single click for each kidney on displayed digital image on the computer monitor is necessary for assigning the renal ROT. The renal ROI is detennined from the pixel values of digital image.

The semi-automated renal ROI selection method of the present invention is based on a novel idea described above and consists of automated sequence of steps by computer. The flow diagrgn is shown in Fig. 1. Each step is explained in detail as follows.

Step one: The operator clicks around the center of the kidney on a displayed computer image.

Step two: the maximal renal count for each kidney is defined as the maximal count in the first area, with the clicked pixel as the center.

Step three: Pixels with values higher than the first threshold based on the maximal renal count in the second area (for example, whose size is twenty times that of the first area) where again the clicked pixel represents the center are selected as renal central areas.

Step four: If there is continuity between two areas in the renal central areas due to one pixel, the pixel responsible (bridged pixel) is deleted to break the continuity.

Step five: Only the area including the clicked pixel is newly selected as the renal central area.

Step six: An area 5-20 rrrrn in width outside the renal central area is selected. The pixels in this area with values exceeding the second threshold are selected as the renal peripheral area and together with the renal central area constitute the renal ROT.

(See Fig. 2) In performing step one, the operator only clicks around the center of each kidney (points a and ). The rectangular area 5 pixels in width and 9 pixels in height was considered appropriate as the first area (P1) in step two in order to store the maximal renal count correctly, when the size of each pixel was 9.3 rnm x 9.3 mm. The reason being that when its center is on the clicked pixel, this area must be large enough to include the pixel with the maximal renal count and at the same time small enough to exclude pixels having a higher count of the other organs. Considering the size of an average Japanese adult's kidney, the rectangular area 30-70 mm in width and 60-120 mm in height was considered appropriate as the first area.

(See Fig. 3) When performing step three the second area (P2; dotted line) must be large enough to include the kidney. Of course, its size may be large enough to include the whole body, but for practical purposes the size 1-15 times that of the first area may be sufficient. In any case, a larger than necessary P2 will not affect adversely the assignment of the renal ROT. Besides, since the second area is set for each kidney and calculation is done separately, it does not matter whether the two areas overlap or not (their central parts overlap).

The pixels in the second area (P2) (see Fig. 3) with values exceeding the first threshold based on the maximal renal count (for example, 60% of the maximal renal count) are selected as the renal central areas (M1; solid lines in Fig. 3). This procedure establishes the approximate contours of the renal ROI minimizing the connection with the neighboring organs.

When performing step four, (see Figs. 3-4) if there is continuity between two areas in the renal central areas (M1), due to one pixel, the pixel responsible (the bridged pixel (K)) is deleted frorn the renal central areas to break the continuity. If either of the two pattems on the left of Fig. 7 (a) and (b) is present in the bitmap storing the position of the renal central area, the value of the center pixel is changed to 0. The center pixels of the two patterns on the left are bridged pixels. The value of 1 denotes "selected as the renal central area", 0 denotes "not selected" and a space denotes either "selected" or "not selected". The real renal central area is separated from the surrounding organs by step four.

Step four is rarely performed. In our study, only in one case out of fifty nine, this step proved to be necessary. This experiment is explained in detail below. It is this data which is used in Figs. 2-6. Thus, even if step four is left out, there is usually no inconvenience for assignment of the renal ROI except in few isolated cases.

When performing step five (see Fig. 5), only the area including the clicked pixel in the renal central area (M1), is newly selected as the renal central area (M2). When the pixels are adjacent to each other sideways, they are supposed to crcate continuity. This procedure helps distinguish the kidney from any other organs.

When performing step six (see Fig. 6), since the second threshold was set high in order to separate the kidney from the surrounding organs, the obtained area is smaller than the appropriate renal ROI. Therefore, the area consisting of the pixels with values exceeding the second threshold based on the maximal renal count (for example, 20% of the rnaxi:mal renal count and lower than the first threshold) in the pixels adjacent to the renal central area (M2), is added as renal peripheral area (R) to the renal central area and together form the renal ROI (S). The size of each pixel is 9.3 mm x 9.3 mm. In the case used in Figs. 2-6, all the pixels adjacent to each renal central area (M2), have higher values than each second threshold and are added to each renal central areas as the renal peripheral area.

In our study, almost all the pixels around the renal central area (M2), have the values higher than 10-20% of the rnaximal renal count. Therefore, the area consisting of all these pixels may become the renal peripheral area unconditionally, and is included in renal ROI (S).

In other words, instead of step six, all pixels which are completely or partially included in the area 5-20 mm in width outside the renal central area are defined as the renal peripheral area (R) and together with the renal central area constitute the renal ROI.

Next, inter-operator variability in calculating %ID using our semi-automated method versus manual method was evaluated with the following results.

Subjects: Eleven normal volunteers, 22 patients with nephro-urological diseases (18 men, 15 women; 32 left kidneys and 30 right kidneys) were examined. Their ages ranged from 18 to 83 years (mean, 46 years)7 and their serum creatinine levels ranged from 0.5 to 3.5 mgldl (mean, 1.0 mgidl).

Thirty minutes after the intake of 300 ml of water, a rapid bolus injection of 79-314 tvlBq (2-9 mCi) Tc-99m-MAG3 was administered with subject in the supine position above a large field of view gamma camera equipped with a low-energy all-purpose collimator. Renal irnages were collected in 10-s frames for 20 min. by interfaced computer. The image matrix was 62x64. The size of each pixel was 9.3 mm x 9.3mum. The photo peak was selected at 140 KeV with a 1570 window. The pre-injection and post-injection syringes were placed on a camera and were counted for 10 sec. The data was stored in 128x128 rnatrix images.

The 1-2 min. image and the 2-3 min. image were generated by the summation of images from 60 to 120 s, and frorn 120 to 180 s, respectively. %TD was defined as 100 times the total renal count in each ROI divided by the 60-s count of the decay-corrected injected dose which was obtained from the counts of pre-injection and post-injection syringes.

Utilizing the semi-automated renal ROI selection method, the first area (P 1), was the rectangular area 5 pixels (46.5 mm) wide and 9 pixels (83.7 mm) high. The second area (P2), was fixed at the rectangular area 21 pixels (195.3 mm) wide and 27 pixels (251.1 mm) high.

Experiment 1 The 1-2 min images were displayed using the monochrornatic colourmap with 256 grey scale; the upper level of the display window was fixed at the maximal count of the displayed image and the lower level was set at 1070 of the upper level. This display condition was named monochromatic display. ROLLS were drawn manually by three operators (A, B and C), with 3, 20 and I year experience in nuclear medicine, respectively. The interoperator variability among these three operators in calculating the roID waS evaluated.

Experiment 2 The operator A assigned renal ROT manually on the 1-2 min. images once more using monochromatic display. The intra-operator variability was evaluated by comparing the %TD of this trial with that of the first trial.

Experiment 3 The 1-2 min. images were displayed using the color colourmap with 256 grey scale; the upper level of the display window was fixed a.t the maximal renal count and the lower level was set at 10% of the upper level. This display condition was named color display.

ROIs were dra.wn manually again by operator A using this display condition in order to evaluate if display condition influences 70ID, and the %ID of this trial was compared with that obtained prior by operator A using manual method with monochromatic display.

Experiment 4 The operator A processed ROIs on 1-2 min. images and 2-3 min. images using our serni-autornated renal ROI selection method, with various combinations of thresholds. %ID obtained by this method was compared with that of manual ROI obtained by operator A using color display.

Experiment 5 The three operators A, B and C processed ROIs on 1-2 min. images using the new technique, the semi-automated renal ROt selection method of the present invention, with appropriate parameters in the experrrnent 4. The inter-operator variability among them in calculating the 7oID was evaluated.

RESULTS %ID was expressed as mean standard deviation. Evaluation of inter-operator variability among the three operators was performed by a one-way repeated measures ANOVA, and comparison between two operators in this group was made by post-hoc tests (Fisher's PLSD, Scheffe and Bonferroni/Dunn tests). The intra-operator variability and comparison between manual ROI and semi-automated ROt were made by Student's t-test. Ps 0.05 was considered significant. Linear regression analyses were done by the least squares method.

For Experiment I, the %ID was 4.14+1.27 for A, 4.32to.28 for B and 3.281.07 for C, and these values were significantly different (Ps0.0001). The relationships between them were indicated in Figs. S-10. Although there was good correlation among them, the 7OID was significantly different between operator C and operator A or B. (The term "r" in the figures represents correlation coefficient.) In Experiment 2, the %ID of the second trial obtained by operator A using manual method with monochromatic display was 4.12+1.26. The %ID was significantly different between the first and second trials (P=0.37). The relationship of %ID between them was indicated in Fig. 11.

In Experirnent 3, the %ID obtained by operator A using manual method with color display was 4.63 1.33. Compared with %ID obtained prior by operator A using manual method with monochromatic display, these values were significantly different (psO.0001).

The relationship between them was tndicated in Fig, 12.

In Experiment 4, the detectability of the kidney on the semi-automated renal ROI selection method depends only on the first threshold. On 1-2 min. images, 8 of 62 kidneys (four left and four right kidneys) could not be delineated correctly when the first threshold was set at 50%. When it was set at 60%, 3 kidneys (two left and one right kidneys could not be delineated. When it was set at 70%, obtained areas did not reflect the renal shapes.

Therefore, we accepted 605 as the first threshold. Next, the second threshold was set at 10, 20 and 30% and the values obtained using these parameters were compared with those obtained by operator A using manual method with color display. There was significant difference, when the second threshold wa.s set at 30% (ps0.0001). No significant difference was observed, when it was set at 20% (p=0.10) and 10% (p=0.77). The 20% was thought to be appropriate for the second threshold, because of ROI's shape. These relationships are indicated in Figs. 13-15.

The three kidneys not successfully detected using the semi-automated renal ROT selection method with the first threshold of 60% had %ID values of 1.8770 (left), 1.83% (right) and 2.95% (left) obtained by operator A using the manual method with color display.

The semi-automated method failed in their cases because they had very markedly reduced function and overlapped the other organs.

On 2-3 min. images, obtained renal central area did not reflect the renal shape in one of 62 kidneys when the first threshold was set at 60%. When it was set at 50%, all the kidney could be delineated correctly. When it was set at 40%, it failed in three kidneys (two left and one right kidneys), owing to superimposition by liver or spleen.

Therefore, 50% was accepted as the first threshold.

Next, the second threshold was set at 10, 20 and 30% and the %ID obtained using these parameters was compared with that obtained by operator A using manual method with color display. There was significant difference, when it was set at 30% and 10% (p < 0.0001). No significant difference was found, when it was set at 20% (p=0.06). These relationships are shown in Figs. 16-18.

Experiment 5 The first and second thresholds of the semi-automated renal ROI selection method were set at 60% and 20%, respectively, on 1-2 min images. The %ID was 4.60+1.31 for three operators. No significant difference was observed (p10.9999).

Complete reproducibility was shown in 58 of 59 kidneys; one kidney showed %ID of 4.87% for A and 4.76% for B and C. This difference was minor. The relationships between them are shown in Fig. 19, as well as Figs. 20(b) and 20(d).

CONCLUSION From the reasons described above, 60% and 20% were appropriate for the first and second thresholds, respectively, of the semi-automated renal ROI selection method on 1-2 min. images using Tc-99m-MAG3. Moreover, 50% and 20% were appropriate on 2-3 min. images, respectively.

Almost complete reproducibility of %ID was demonstrated on 1-2 min, images using the semi-automated renal ROI selection method with these appropriate parameters (see Fig. 19).

In other words, an operator with little experience in nuclear medicine could assign the renal ROI as accurately as an experienced one.

Detectability performance of our semi-automated method was as follows: It failed only with 3 of 62 kidneys on 1-2 min, images, and was completely accurate on 2-3 min. images. The lowest %ID on 1-2 min. images that was accurately calculated using this method was 0.72%. On the other hand, the single-threshold method using the threshold of 20% failed in 51 of 62 kidneys on 1-2 min, images and in 20 of 62 kidneys on 2-3 min. images.

Compared with the single-threshold method, our semi-automated method demonstrated superior renal detectability. Figures 20(a) and (c) show unsuccessful ROI selection using the single-threshold method in a normal subject and a patient with chronic renal failure. To the contrary, figures 20(b) and (d) show successful ROls in the same subjects obtained using our method.

In addition, although we used Tc-99m-MAG3 as a radiopharmaceutical in this study, our new technique, the semi-automated renal ROI selection method, is also applicable for the images obtained using Tc-99m-DTPA or l-131-OlH.

Moreover, our new technique, the semi-automated renal ROI selection method, is applicable for the assignment of the other organ's ROI on scintigram using, the following procedure.

Photons from the radiopharmaceutical administered in vivo is collected by a gamma camera. Obtained images (scintigram) are transferred to the computer, where ROI for the target organ is set by the semi-automated renal ROI selection method based on pixel values on displayed computer image. First, the operator clicks around the center of the organ on displayed computer image. (All the following, steps are performed automatically.) Second, the maximal count for the organ is defined as the maximal count in the first area, with the clicked pixel as the center. The subsequent threshold technique is based on this maximal count for the organ. The following calculation is based on the second area which is larger than than the first area, and where again clicked pixel represents the center. Third, the pixels in this area with values exceeding the first threshold are selected as the central areas for the organ.

Fourth, if there is continuity between two areas in the central areas due to one pixel, the pixel responsible (bridged pixel) is deleted to break the continuity. Fifth, only the area including the clicked pixel is newly selected as the central area. Sixth, an area i20 mm in width outside the central area is selected. The pixels in this area with values exceeding the second threshold are selected as the peripheral area and together with the central area they constitute the ROI.

These, six steps were performed using computer equipment.

Of course, the values for the first and second thresholds and the sizes and shapes of the first and second areas, and the width of the peripheral area must be fixed based on clinical data taking into account the organ, purpose or radiopharmaceutical.

As described above, the semi-automated scintigraphic renal ROI selection method of the present invention is a very useful method for assignment of renal ROI, because it guarantees almost complete reproducibility of renal ROI, can successfully assign renal ROI for the kidney with reduced function, and is applicable to the assignment of the other organ's ROI on scintigram.

ADVANTAGES OF THE INVENTION Because this invention, the semi-automated renal ROI selection method, involves the procedures described above, it has the following advantages: (1) Inter-operator variability in the assignment of the renal ROT can be improved.

(2) Assignment of the renal ROI can be completed in a very short time.

(3) Experienced and inexperienced operators alike can assign renal ROT with accuracy.

(4) This method demonstrates superior renal detectability.

(5) The method can be utilized for assigrrrnent of ROT of the other organs on scintigram.

Claims

HAT IS CLAIMED IS:

l. In a semi-automated scintigraphic renal region of interest selection method in which photons from a radiopharmaceutical administered in vivo are collected by a garrrna- camera, obtained images (scintignm) are transferred to a computer, and each renal region of interest is set by a semi-autornated renal region of interest selection method based on pixel values on a displayed computer image, the improvement comprising the steps of: (a) cliclcing around the center of a kidney on a displayed computer image to store the position of the kidney; (b) setting the first area with the clicked pixel as the center, calculating maximal renal count for each kidney, which is defined as the rnaximal count in the first area, and setting first and second thresholds based on maximal renal count; (c) setting the second area with the clicked pixel as the center, and selecting the renal central areas by determining which pixels in the second area have values which exceed the first threshold; (d) deleting any bridged pixel between two areas in the renal central area to break continuity between them; (e) distinguishing the kidney from other organs by newly selecting only the area including the clicked pixel as the renal central area; and (f) selecting, as the renal peripheral area, pixels with values exceeding the second threshold in the pixels adjacent the renal central area, and adding the renal central area to the renal peripheral area to determine the renal ROI.
2. The method of claim l, wherein the f t area is set as a rectanaular are 30-70 nun in width and 60-120 mm in height, and the second area is set as a rectangular area 100250 mm in width and 150-300 mm in height.
3. The method of claim 1, wherein instead of step (f), the renal region of interest is calculated by including all pixels which are completely or partially in an area 5-20 my in width outside the renal central area together with the renal central area.
4. The method of claim 1, wherein first and second thresholds are fixed at 50-70% and 1-40ió, respectively, of the rnnximal renal count on 1-2 min. image generated by the summation of images from 60 to 120 seconds after administration of Tc-99m-MAG3.
5. The method of claim 1, wherein first and second thresholds are fixed at 40-705 and 1-40%, respectively, of the maximal renal count on 2-3 min. image generated by the surnrnation of images from 120 to 180 seconds after administration of Tc-99m-MAG3.
6. The method of claim 1, wherein steps (a) to (c) followed by (e) and (f) are carried out without step (d).
7. A method of deriving information from pixel values on a displayed computer image of a kidney, said method comprising the steps of: 1) clicking around the centre of the image of the kidney to store the position of the kidney, 2) setting a first area with the clicked pixel as the Centre, calculating a maximal renal count for the kidney, which is defined as the maximal count in the first area, and setting first and second thresholds based on the maximal renal count, 3) setting a second area with the clicked pixel as the centre, and selecting the renal central area(s) by determining which pixels in the second area have values which exceed the first threshold, 4) distinguishing the kidney from other organs by selecting only the area including the clicked pixel as the renal central area, and 5) selecting, as the renal peripheral area, pixels with values exceeding the second threshold in the pixels adjacent the renal central area, and 6) adding the renal central area to the renal peripheral area to determine the renal region of interest.
8. A method of deriving information from pixel values on a displayed computer image of a kidney, said method comprising the steps of: i) clicking around the centre of the image of the kidney to store the position of the kidney, ii) setting a first area with the clicked pixel as the centre, calculating a maximal renal count for the kidney, which is defined as the maximal count in the first area, and setting first and second thresholds based on the maximal renal count, iii) setting a second area with the clicked pixel as the centre and selecting the renal central area(s) by determining which pixels in the second area have values which exceed the first threshold, iv) distinguishing the kidney from other organs by selecting only the area including the clicked pixel as the renal central area, v) selecting, as the renal peripheral area, all pixels which are completely or partially in an area 5 - 20 mm. in width around the renal central area, and vi) adding the renal central area to the renal peripheral area to determine the regional area of interest.
9. A method as claimed in any one of the preceding claims, substantially as hereinbefore described with reference to the accompanying drawings.