A New Washout Rate Display Method for Detection of Endocardial and Pericardial Abnormalities Using Thallium-201 SPECT

A New Washout Rate Display Method for Detection of Endocardial and Pericardial Abnormalities Using Thallium-201 SPECT Mansour Jamzad, Hajime Murata, Hinako Toyama, Yuji Takao, and Akihiko Uchiyama School of Science and Engineering, Waseda University; Toranomon Hospital; and Tokyo Metropolitan Institute of Gerontology, Tokyo, Japan Using thallium-201 single-photon emission computed tomography images, we have developed a new washout rate display method, which can demonstrate abnormalities in the inner and outer sections of myocardial slices. The washout rate is calculated on a pixel basis: the count of a pixel in an early slice is compared with the mean count of a set of pixels in the corresponding slice of the delayed images. The output of this method is a set of 10 images showing the washout rate for Slices 1 to 10 of the left ventricle. A low washout rate can be visualized in detail on both the inner and outer parts of the slices. We have applied this method on several patients who had a non-q-wave infarction and compared these results with those of the bull's-eye method. Our method was able to distinguish the location of abnormalities in the endocardial and pericardia/ parts of myocardial slices. The shape and the color of the output images provide information that is helpful in the diagnosis of non-q-wave infarction. The number of patients who suffer from non-q-wave (subendocardial) infarction is much lower than the number who suffer transmural myocardial infarction (MI), but the precise diagnosis of non-q-wave infarction is important in clinical medicine. The bull's-eye display ( 1 ), circumferential profile analysis (2), and the newly developed Stereo-View (3) are the methods most commonly used to detect myocardial abnormalities, using thallium-20 I e 0 'Tl) single-photon emission computed tomography (SPECT) images. However, these methods cannot distinguish abnormalities in the inner and outer halves of myocardium. In the bull's-eye method, each myocardial slice is divided into 40 to 60 sectors. The maximum or mean count in a sector is taken to represent that sector in the washout rate calculation. Although this method can demonstrate the abnormality in a sector, in some cases it cannot detect the exact location and severity of a sector's abnormality because the endocardial and pericardia! portions are not treated separately. The aim of our method is to calculate the washout rate in For reprints contact: Mansour Jamzad. PhD. Department of Communication and Electronics, School of Science and Engineering. Waseda University. 3-4-1 Shin Ohkubo, Shinjuku-Ku. Tokyo, Japan 169. VOLUME 20, NUMBER 2, JUNE 1992 such a way that the displayed washout rate images can show the abnormal areas in the inner and outer halves of myocardial slices. Using our display method, subendocardial and transmural abnormalities can be distinguished from each other. In order to evaluate our method we have compared it with the bull's-eye method. MATERIALS AND METHODS For all normal volunteers and patients that we have studied, the exercise test was performed with a supine bicycle ergometer: the graded workload started at I W /kg body weight and increased every 3 min by 15 W. Exercise was stopped at the end point, according to the Michigan Standards. One min before the end of exercise, Ill MBq (3 mci) of 201 Tl was injected intravenously ( 4). The SPECT images were acquired at 5 min (stress scans) and at -4 hr (rest scans) after the thallium injection, by a rotatable gamma camera, with a low-energy general-purpose collimator interfaced with a dedicated computer (Maxi camera 400 AC/T Maxi Star system, General Electric, Milwaukee, WI). Images were reconstructed by a filtered back projection method with attenuation correction (4). For each of the early and delayed studies, ten short axial SPECT images (matrices 64 x 64 x 16 bits in size) were extracted from the apex to the base of the left ventricle (LV). These short axial images are referred to as slices. For each slice of the early and delayed studies, the region of myocardium is determined automatically (5). The endocardial and pericardia! parts of a slice are distinguished by automatically drawing a circle in the middle of the inner and outer myocardial rings, as seen in Figure I. The inner and outer halves of a myocardial slice are divided into 16 and 32 equiangular sectors, respectively (Fig. 2). Our method is programmed in FORTRAN-5 on the SCIN TIPAC-2400 system, which uses a NOVA-4 computer. Normal Limit Calculation In order to evaluate the patient's washout rate we needed to compare it with a normal limit. To obtain these normal limits, we used 10 normal healthy volunteers whose average age was 51-yr. For these volunteers the washout rate, WR, of 75

sector kin slice n, (n = 1-10) is calculated as the mean minus 2 s.d. of washout rate of the same sector in all I 0 normal volunteers. In the above formula we used Emean k and Dmean k, which are the mean count of pixels in sector k, instead of Emax k and Dmax k, to obtain another set of normal washout rate limits. The washout rate of a pixel can be compared with either of the above two normal limits. In our experiments, we used the normal limit based on the maximum count. However, for the normal volunteers that we examined, there was not a significant difference between the normal limits calculated in either way. The inner half of a myocardial slice The outer half FIG. 1. Separation of the endocardial and pericardia! parts in a myocardial slice. ANT Patient Washout Rate Calculation The corresponding slices on the early and delayed images usually have different sizes and shapes physiologically. Therefore, they cannot be aligned, and it is not possible to relate one pixel in an early slice with one pixel of the same slice in the delayed stage. To overcome this problem, we take one pixel from the early image and compare its count with the mean count of a set of pixels located approximately on the same location of the same slice in the delayed image. For example, assume a pixel (i, j) is located in sector k of slice number n in the early stage (Fig. 2). The count of this pixel is compared with the mean count of all pixels located in the same sector k of the same slice n in the delayed stage. We have calculated the pixel base washout rate, WR,.. J> according to the following formula: wo. - E,,.i> - Dk 100 l'{!.j)- E X (l.j) Eq.2 LAT INF FIG. 2. Dividing the endocardial and pericardia! parts into 16 and 32 equiangular sectors, respectively. a sector, k, was calculated as follows: WRk = Emax k - Dmax k x 100 Emax k Eq. I For the sectors in the inner and outer halves of a myocardial slice, k equals 1-48, as shown in Figure 2. Emax k and Dmax k are the maximum count of pixels in sector k of the early and delayed slices, respectively. The normal washout rate limit for 76 E 1 ;.J 1 is the count of pixel (i, j) in the early slice and D, is the mean count of pixels located in sector k of the same slice in the delayed stage. The above formula is applied to all 10 slices and the washout rate is calculated for all pixels in the myocardium. Output Image Calculation The output of this method is a set of 10 images representing the washout rate of Slices I through 10. The output images have the same shape as the early slices because the washout rate calculation is based on the pixels in the early slices. For a pixel (i, j), located in sector k of slice n, c(i.j) is the value of that pixel in the output image. If W~i.j> is greater than the normal limit of sector k in slice n, then C 1 ;.J 1 equals a constant greater than the overall maximum washout rate. Otherwise, C 1;.i 1 equals W~ -i> Using these calculations, all normal pixels will have a unique color (we have selected red), and the color of abnormal pixels will vary from yellow (near normal) to black (a very low washout rate, indicating severe abnormality). In the output images, the color of a pixel is an indicator of the abnormality's severity. In each of the ten output images. three white circles are displayed to separate the endocardial and pericardia! portions of the slice. JOURNAL OF NUCLEAR MEDICINE TECHNOLOGY

RESULTS We have applied our method and the bull's-eye method on 12 patients. The prolonged ST, T abnormalities on an ECG, abnormal enzyme value, and the clinical features have shown non-q-wave infarction for all of these patients. The results are summarized in Table l. Image Display Figures 3A, 3B, and 3C show the SPECT images, the bull'seye method output, and our method output, respectively, for Patient 5 in Table I. In Figures 3A and 3B, although the bull's-eye and SPECT images showed the existence of a low washout rate in the anteroseptal wall, they were unable to single out the non-q-wave abnormality from transmural abnormality. In our method (Fig. 3C), the washout rate abnormality located at the subendocardial part of the anteroseptal wall is clearly visualized; this abnormality could be due to the non-q-wave infarction. Figures 4A and 4B demonstrate the output of the bull's-eye method and our method for Patient 7 in Table l. The bull'seye image shows the existence of a low washout rate in the anteroseptal wall, but it cannot localize it in the endocardial part of the slices. Our method shows the transmural abnormality in Slices 3 through 7 and non-q-wave infarction in Slices 8 through 10 and in some parts of the septal area in Slices 6 and 7. Detectablllty of Myocardial Abnormality The ability of SPECT, the bull's-eye method, and our method to detect myocardial abnormalities, for the 12 patients examined, is summarized in Table I. In this table, the positive and negative signs indicate each method's ability and inability, respectively, to detect abnormality. Columns 2 and 3 show the ability of the bull's-eye method and the visual interpretation of SPECT images to detect abnormalities. Columns 4 and 5 show the performance of our method in detection of non-q-wave and transmural abnormalities. For Patients 2, 8, and 9, none of the methods detected any abnormality. DISCUSSION The washout rate image is well accepted as a tool for the detection of ischemic heart disease. In transmural myocardial infarction, the washout rate does not necessarily decrease due to the lack of viable myocardium. However, in the case of non-q-wave infarction, the washout rate decreases because there is a high possibility of existence of viable tissue in the endocardial portion of the myocardium. Therefore, the decrease of washout rate in the endocardial portion suggests non-q-wave infarction. Clinical detection of non-q-wave infarction is sometimes difficult. Myocardial scintigraphy with 201 Tl, visual interpretation of SPECT images, and use of the bull's-eye method all seem to be unable to demonstrate subendocardial abnormalities (see Figs. 3A, 3B, and 4A). The bull's-eye method does not separate the endocardial and pericardia! parts on a sector. It calculates only one washout rate value for the whole area of a sector and paints it with one unique color. In visual interpretation of SPECT images, calculation of the washout rate and comparison with the normal limit and the washout rate is not performed. Therefore, it is difficult to confirm the existence of a subendocardial abnormality. In our method, the washout rate can be calculated and displayed for each pixel of a slice. This gives detailed information about the washout rate in all parts of the LV. Thus, our method is capable of demonstrating transmural and nontransmural abnormalities. For a normal person, all parts of the washout rate images are displayed in one unique color (red in our method). The only input parameter required for our method is the patient's file name. The rest of the calculations are performed automatically. Three of the twelve patients that we examined, showed no TABLE 1. Comparison of Washout Rate Display Methods Our Method Bull's-eye SPECT Endocardial Pericardia! Region of Extent of Patient No. Method Image Abnormality Abnormality Abnormality Abnormality 1 + + + Anterior Small 2 3 + + + Anterior Medium 4 + + + Anteroseptal Large 5 + + + Anteroseptal Medium 6 + + + + Anteroseptal Large 7 + + + + Anteroseptal Large 8 9 10 + + Anteroseptal Small 11 + + + Anteroseptal Small 12 + + + + Anteroseptal Medium The plus and minus signs denote each method's ability or inability, respectively, to detect abnormality in the region indicated. VOLUME 20, NUMBER 2, JUNE 1992 77

FIG. 3. (Upper left) The early and delayed SPECT images for Patient 5 of Table 1, (57 yr, male) with anteroseptal subendocardial abnormality. CAG: LAD(7) 75%-90% stenosis. Subendocardial abnormality cannot be visualized in these images. (Upper right) The bull's-eye washout rate images for the same patient. Although the abnormality in the anteroseptal wall is shown, the subendocardial abnormality cannot be clearly identified. (Lower left) Pixel base washout rate image by our method for the same patient. The red color shows the normal area; the dark color in Slices 3 through 1 0 shows the existence of subendocardial abnormality in most of the anteroseptal wall. abnormality, whether using SPECT images, the bull's-eye method, or our method. Our method was able to detect subendocardial and transmural abnormalities in the other nine patients. The bull's-eye method showed the abnormality and its relative location in eight patients. But it could not single out the subendocardial and transmural abnormalities in any of them. For Patient lo in Table l, the bull's-eye method could not detect any abnormality. The wall motion that occurs during the data acquisition period for ordinary myocardial SPECT images can be a limitation for our method. As a result of this motion, the myocardial images that we use are superimposed images of all phases between diastole and systole. For these kind of images, separation of the inner and outer parts of myocardial slices cannot be done properly. However, we believe that we can obtain the best results from our method if we use the multi-gated method in myocardial SPECT imaging and take images from the diastolic phase. Using the diastolic phase images, the endocardial and pericardia! parts of the myocardium can be separated precisely, and our method is then able to show the endocardial and pericardia! abnormalities with more accuracy. Due to the relatively long data acquisition time (about l hr) for the multi-gated method, it is not practiced in most clinics. However, it has been reported that "this data acqui- 78 sition time can be reduced to 30 to 40 min by increasing the 201 Tl dose from 111 MBq to 185 MBq (3 mci to 5 mci) and by acquiring fewer steps" (6). Moreover, it is likely that the problem of a long data acquisition time could be overcome by the new technetium-99m labeled radiopharmaceuticals (7,8). CONCLUSION Our experiments showed that the bull's-eye display and the visual interpretation of SPECT images-two very common methods for detecting myocardial abnormalities-could not distinguish the abnormalities in the inner and outer halves of the myocardium. In our method, the washout rate is calculated and displayed on the pixel basis. It has the advantage of being able to visualize in detail the abnormal washout rate areas in the endocardial and pericardia! parts on each myocardial slice. Our method can provide supporting data for the diagnosis of non-q-wave infarction. ACKNOWLEDGMENTS We would like to thank Mr. M. Onoguchi and the entire staff of the nuclear medicine division at Toranomon Hospital for their support and cooperation during this research. JOURNAL OF NUCLEAR MEDICINE TECHNOLOGY

FIG. 4. (Left) The bull's-eye washout rate image for Patient 7 in Table 1, (75 yr, male), with anteroseptal subendocardial abnormality, CAG: LAD(6) 90, RCA(3) 90% stenosis. Abnormality can be seen on the anteroseptal wall, but the transmural and subendocardial abnormality cannot be distinguished. (Right) Pixel base washout rate image by our method for the same patient. A transmural abnormality is seen in Slices 3 through 7 and a subendocardial abnormality is seen in Slices 8 through 1 0 and in some septal areas of Slices 6 and 7. REFERENCES I. Garcia EV, Train KV, Madahi J, eta!. Quantification of rotational thallium-201 myocardial tomography. J Nucl Med 1985;26:17-26. 2. Burow RD, Pond M, Schafer W, et a!. "Circumferential proftles~: a new method for computer analysis of thallium-20 I myocardial perfusion images. J Nuc/ Med 1979;20:771-777. 3. Toyama H, Matsuda H, Murata H, et a!. A method for quantitative evaluation of myocardial SPECT images with three-dimensional display "STEREO-VIEW.~ Japanese! Nucl Med 1987;24:983-990. 4. Takao Y, Murata H, Katoh K. Availability and limitations of thallium- 20 I myocardial SPECT quantitative analysis: assessment as daily routine procedure for ischemic heart disease. Ann Nuc/ Med 1991;5:11-18. 5. Jamzad M, Uchiyama A, Toyama H, Murata H. Analysis ofthallium-201 SPECT images using fuzzy set theory. Ann Nuc/ Med 1988;2:63-71. 6. Mochizuki T, Murase K, Fukiwara Y, et a!. ECG-gated thallium-201 myocardial SPECT in patients with old myocardial infarction compared with ECG-gated blood pool SPECT. Ann Nucl Med 1991;5:47-51. 7. Okada RD, Glover D, Gaffney T, eta!. Myocardial kinetics oftechnetium- 99m-hexakis-methoxy-2-methylpropyl-isonitrile. Circulation 1988;77: 491-498. 8. Gerundini P, Maffioli L. Cationic complexes of technetium for myocardial imaging. (Abstract.) J Nucl Med 1989;30:80 I. VOLUME 20, NUMBER 2, JUNE 1992 79