and Department of Medicine, School of Medicine, University of California, Los Angeles, Los ..... and its fully automated approach saves technologist time and.
Postexercise Lung Uptake of 99mTc-Sestamibi Determined by a New Automatic Technique: Validation and Application in Detection of Severe and Extensive Coronary Artery Disease and Reduced Left Ventricular Function Claudia Bacher-Stier, Tali Sharir, Paul B. Kavanagh, Howard C. Lewin, John D. Friedman, Romalisa Miranda, Guido Germano, and Daniel S. Berman Departments of Imaging and Medicine and Burns and Allen Research Institute, Cedars-Sinai Medical Center, Los Angeles; and Department of Medicine, School of Medicine, University of California, Los Angeles, Los Angeles, California
increased
lung uptake of 2°'T1after exercise (7,2), as well
This study validated a new automatic algorithm for assessment of lung-to-heart ratio (L/H) of radiotracers in myocardial perfusion as after pharmacologie stress (3), has been shown to SPECT and assessed the diagnostic value of 99mTc-sestamibi correlate with the presence of severe coronary artery disease L/H after exercise. Methods: The new technique extracts a left ventricular region of interest (ROI) from a summed anterior projection image and generates a lung ROI by reshaping and translating the left ventricular ROI. This algorithm was applied to 230 patients who underwent exercise 99mTc-sestamibiSPECT (gated SPECT, n = 88) with first-pass ventriculography. Normal values were established in 26 patients in whom the likelihood of coronary artery disease (CAD) was 5% or less. An abnormality threshold for detecting severe and extensive CAD was defined in a subgroup of 109 patients who underwent coronary angiography and was validated in a prospective group (n = 72). Results: The success rate of the automatic algorithm was 97%. Excellent correlation was found between automatic and manual L/H values (r = 0.95; P < 0.001 ). The mean UH was higher in patients with a peak exercise ejection fraction (EF) less than 40% versus 40% or more (0.51 ±0.07 versus 0.43 ±0.05, P < 0.001 ) and in patients with a poststress EF less than 40% versus 40% or more (0.50 ± 0.07 versus 0.44 ±0.06, P < 0.01). A threshold of L/H greater than 0.44 yielded a sensitivity and specificity of 63% and 81%, respectively, for identifying severe and extensive CAD in the prospective group and a sensitivity of 86% in identifying stenosis of 90% or more in the proximal left anterior descending artery. Conclusion: The new automatic algorithm for assessing L/H correlated well with manually derived L/H for "Tc-sestamibi as well as 201TISPECT. An increased postexercise 99mTc-sestamibi
(CAD) (3,4), sion defects Because of sestamibi is
the extent and severity of scintigraphic perfu (3,5), and left ventricular dysfunction (6-9). its physical and biologic properties, ""Tcnow more commonly used than 2°'T1for
myocardial perfusion imaging. Only a few studies, based on highly selected patient populations (10-12), have evaluated the diagnostic value of "mTc-sestamibi lung uptake. Further more, previously described algorithms for determining lungto-heart ratio (L/H) have required various degrees of opera tor interaction during processing (10-13), and a completely automatic method for determination of abnormal lung uptake has not been defined for either 2°'T1or "Tcsestamibi. We have developed a fully automatic approach for calculating the L/H of radiotracers. The new technique extracts a left ventricular region of interest (ROI) from a summed anterior projection image using heuristics and mathematic operators and generates a lung ROI by reshap ing and translating the left ventricular ROI. This study was undertaken to describe and validate the new automatic algorithm and to assess the diagnostic value of ""Tcsestamibi L/H uptake after exercise.
L/H adds significant diagnostic value to study myocardial perfu sion SPECT as a marker of severe and extensive CAD and reduced ventricular function. Key Words: 99mTc-sestamibi; lung uptake; automatic technique; coronary artery disease; left ventricular function J NucÃ-Med 2000; 41:1190-1197
Study Population Two hundred thirty patients who underwent separate-acquisition dual-isotope resting 20IT1and exercise ""Tc-sestamibi myocardial perfusion SPECT (14) were evaluated for the L/H of ""Tc-
Received Jun. 24,1999; revision accepted Nov. 1,1999. For correspondence or reprints contact: Daniel S. Berman, MD, CedarsSinai Medical Center. 8700 Beverly Blvd.. Rm. A041. Los Angeles. CA 90048.
sestamibi uptake after exercise. The automatic assessment of L/H was successful in 223 of the initial patients (97%). The program failed in 7 patients because of high gut activity, leading to incorrect ROI placement. The final study population, comprising 223 patients, was divided into the following 3 subgroups (Table 1).
1190
MATERIALS AND METHODS
THEJOURNAL OFNUCLEAR MEDICINE • Vol. 41 • No. 7 • July 2000
TABLE 1 Patient Characteristics and Exercise Variables Variable
Group 1 (n = 26)
Group 2 (n = 125)
Group 2A (n = 109)
Group 3 (n = 72)
11.2* 11.2*(81 11.3* ±12.3 (76)Ã)t 14(54) 58 97 (78)t 83 (22)* (36)* 34 (27)* 24 26 0 (43)* (36)* (36)* 54 26 0 39 (63)* (54)* (59)* 1(4) 74 69 39 : 14.4* : 13.3* : 10.4* 56 : 56: 67 ±7.4 53: 64 ±6.465: 54:t : 15.4*65:56:t : 13.4*68:51 ::: 12* GS EF (n = 181) (n = 26) (n = 88) (n = 73) (n = 67)
Age (y) No. of men History of Ml Angina§ ST depression§ FPEF50
*P< 0.001 vs. groupl. fP< 0.01 vs. group 1. ÕP=0.02. §Duringexercise. Ml = myocardial infarction; FP = first-pass radionuclide ventriculography; EF = ejection fraction; GS = gated SPECT. Numbers in parentheses are percentages.
Normal Limits (Group I). Twenty-six patients with a low likelihood of CAD (0.44). The mean L/H was 0.41 ±0.06 in patients with insignificant CAD, and 7 of 22 had an increased L/H. The mean L/H was 0.41 ±0.04 in patients with mild to moderate CAD, and 8 of 42 showed an elevated L/H (insignificant P versus the low-likelihood group). The L/H was abnormal (>0.44) in 17 of 21 patients (81%) with S+E LAD and in 15 of 24 patients (63%) with S+E 2/3 CAD. The specificity for identifying S+E LAD disease was 65% (58/89).
THEJOURNAL OFNUCLEAR MEDICINE • Vol. 41 • No. 7 • July 2000
1n
0.8
y = 0.9x+ 0.0489 R = 0.93 p < 0.001
0.6
0.4 0.2-
0.2
0.4
0.6
0.8
ManualLung/HeartRatio Validation of Threshold for Abnormality. Based on the threshold for L/H abnormality (>0.44) defined in group 2A, overall sensitivity and specificity in identification of S+E CAD were 63% (22/35) and 81 % (29/36), respectively, in the prospective group (group 3). Figure 5 shows the relationship between L/H and angiographie results for that group. The sensitivity for identifying S+E LAD was 87% (13/15), and the sensitivity for identifying S+E 2/3 CAD was only 45% (9/20). The specificity for S+E LAD stenosis was 71% (40/56). Mean L/H was 0.47 ±0.02 in patients with S+E
11
FIGURE 2. Correlationbetween automati cally and manually derived L/H of 201TI.
LAD and 0.46 ±0.06 in patients with S+E 2/3 CAD (P < 0.01 versus 0.40 ±0.03 in the low-likelihood group). The mean L/H was 0.39 ±0.05 in patients without S+E CAD (n = 36) (insignificant P versus the low-likelihood group). Relationship Between L/H and Other Markers of S+E Coronary Disease L/H and Summed Stress Score. Figure 6 displays the relationship between the automatically derived L/H of 99mTc-sestamibi uptake and the extent and severity of
Threshold = 0.44
0.8-
I 1
'53
°-6"
Area = 0.78
e
pO.OOl
£> 0.40.2 O 0.2
0.4
0.6
1-specificity
0.8
FIGURE 3. Receiver operatingcharacter istic analysis for determining optimal thresh old for L/H for detection of S+E CAD in group 2A patients. Point of optimal thresh old is indicated by arrow.
AUTOMATIC"MTC-SESTAMIBILUNG UPTAKE •Bacher-Stier
et al.
1193
SENSITIVITYRATE.• SPECIFICITY
-0,6 -0,5-0,4
t** A:_.
011uffiMC30,7
i-l™ •0,3-0.2-0,1
IT
-_4-L ---¿L-.-. ir \\rJ» ¡lt ! ^^| •I
:96%
FIGURE 4. Relationshipbetween UH and presence and S+E of CAD in group 2A patients. Mean ±SD values are shown at right of individual values of each group. Dashed line is upper normal limit for L/H. LK = likelihood; MOD = moderate; VD = vessel disease.
-NORMALCY 62%LOW
68%
VDCAD LK (< 5%)
NO STENOSIS
81%
81%
MILD/MOD CAD
S+E LAD
S+E 2/3
perfusion defects, measured as the summed stress score in group 2A. The frequency of elevated L/H (>0.44) increased progressively with the summed stress score. Patients with a summed stress score > 13 had a significantly higher frequency of elevated L/H than did patients with normal perfusion (summed stress score < 4). L/H and TID. Forty-seven of 109 patients (43.1%) of group 2A had an elevated L/H, and 20 of 109 (18.3%) had an abnormal TID (> 1.22). Patients with an increased TID had a higher L/H than did those with a normal TID (0.44) (n60)65 =
1248(74)76.1 ± 1149± (y)No. (82)79.3 menWeight of 14.58(12)40 ± 117(12)29 ± (kg)SmokingHypertensionDiabetes
(62)5(8)14(22)141 (48)9(15)21 mellitusPrior infarctionMaximal myocardial duringexerciseChest heart rate
4-8
9-13
1531 ±
FIGURE 6. Percentage of patients with increased L/H accord ing to summed stress score (summed stress score s 4, 2/11; 4-8, 7/22; 9-13,11/13; s 13, 25/46). 'P < 0.05; LHR = LVH.
had a higher frequency of prior myocardial infarction (35% versus 22%, P = 0.005) and diabetes mellitus (15% versus 8%, P = 0.04). No significant differences existed in age; sex; weight; history of smoking or hypertension; and maximal heart rate, chest pain and ST segment depression during exercise between patients with an L/H greater than 0.44 and an L/H of 0.44 or less. DISCUSSION
This study validated a newly developed automatic algo rithm for assessment of L/H from myocardial perfusion SPECT and determined the diagnostic value of postexercise "Tc-sestamibi lung uptake. Validation of Automatic L/H in SPECT
To our knowledge, all previously reported approaches for assessment of lung uptake of 2°'T1or "Tc-sestamibi have been either manual (2,4,10,12,13,23) or semiautomatic, with the operator assisting ROI generation (13). Our newly developed algorithm does not require operator interaction, although a manual ROI definition feature is provided for cases in which the algorithm does not correctly identify the TABLE 2 Relationship Between L/H and EF ¡nGroup 2 Patients (n = 125)
±0.07*
Peak exercise FP EF
103 0.43 ±0.05 0.50 ±0.07t 16 0.44 ±0.06 72L/H0.51
*P< 0.001 vs. EFa40%. tP40%. FP = first-pass ventriculography; GS = gated SPECT.
1623 ±
exerciseST pain during (48)42 (38)35 depression during exerciseNormal (64)Elevated (58)PNSNSNSNSNS0.040.005NS
Summed Stress Score
s40
