214A
JACC Februaryt996
ABSTRACTS-Poster
twesn maximal and minimal diameter of the smallest cress-section were
considered. Finally, a wire frame display of the LVOT was obtained, which
can be rotated on the screen for versatile interpretation. In all patients, the
shape of the LVOT as well as the site of the minimal CSA could be clearly
identified. Quantitative data were as follows:
MinimalCSA(cm2)
Max/rninCSA
Max/raindiameter
HCM
2.3 4-1.O
2.6 4.0.9
1.64. 0.3
Normals
5.0 :E0.9
1.44.0.2
p value
<0.0001
<0.0001
1,2 • 0.1
<0,001
Attenuation Corrections and Automation
Tuesday, March 26, 1996, 9:00 a.m.-11:00 a.m.
Orange County Convention Center, Hall E
Presentation Hour: 9:00 a.m.-10:00 a.m,
Evaluation o ! Attenuation Corrected Gated
Tc-ggm-SestamibI Myocardial Perfusion S P E C T
Edward P. Ficero, Mlchelle L. Burger, Prakash Adiseshan, Petricia A. Rose,
James R. Corbett. University of Michigan, Ann Arbor, MI
T
S
D
A
Y
This study evaluated a clinical protocol for performing measured attenuation
correction on gated sast~mibi parfusion studies to reduce regional perfusion
artifacts due to photon attenuation in the chest. Fifteen normal volunteers
had a transmission scan followed by a gated emission scan using a three detsctor SPECT system (Picker 3000XP). The transmissionscan was acquired
in a 64 x 64 matrix at 6~ steps over 3609for a total of 6 minutes using a 5.6
GBq Am-241 line source opposite a 65 cm fanbeam collimator. The gated
emission study was acquired in 16 frames in a 64 x 64 metdx at 6~ steps
over 1209for an average of 40 seconds per step. Uncorrected (NC) images
were ramp-reconstructedusing filtered-beckprojection,while the attenuation
corrected (AC) emission images were iteratively reconstructed using the
transmission data to correct the emission data for photon attenuation (pro.
cossing time < 20 mitt). Endocardial end epicardlal surfaces were detected
from short axis images and the left vanldcular ejection fraction (LVEF) was
calculated from the change in volume from end diastole (ED) to end systole
(ES). Mean activity distributions were tabulated at ED, ES, and for the time
composite (ungated) study. The AC images showed increased uniformity in
tracer distribution relative to the NC images. Decreased activity in the AC
apex was due to anatomical thinning and the partial volume effect. The difference in apex activity between ungated and ED may necessitate a specific
EO normal database. Estimated LVEFs frem the AC and NC Images were not
statistically different. Based on these results, we anticipate improved accuracy in the assessment of coronary heart disease from AC pedusion images
in addition to assessments of vantricular function from AC gated SPECT.
P
0
E
R
Ungated
ES
ED
MeanAC activity
LAT
INF
86 :l: 7
82:E8
844-7
80+8
84 :t:8
80:t:8
Apex
774-7
79:E7
70±7
MeanNC activity
I.AT
INF
864-6
674-8
85-~-6 65:E9
84~6
66:E9
No Defect
Withoutcorrection
With correction
In conclusion, transthoracic three-dimensional eohocardiography allows
detailed qualitative and quantitative information of the LVOT. Patients with
HCM ere characterized by a smaller minimal cross-section area and by a
highly eccentric and asymmetric shape of the LVOT, compared to normal
subjects.
•
whe,~e ~.fterattenuation corre~ion the activity was class to peak. Defects
were dearly visible before and after attenuation correction, but defect peak
activities were higher when attenuation correction was used. Mean count
activity (% maximal counts) of the antedor segments were as follows:
Apex
82+7
83±e
76±8
•
Attenuation Correction With a Gadolinium-153
(Ga-153) Line Sour¢~ Does Not Jeopardize
Detection o f Small Perfuslon Defects
Matthew D. Scarlett, Zuo-Xlang He, John J. Mahmarian, Mado S. Verani.
Baylor College of Medicine, Houston, TX
This study assessed whether attenuation correction with a Ga-153 llnesoume might obliterate small perfusion defects with mildly reduced tracer
activity. Tomograms were acquired of a cardiac phantom (Caplnte¢, Inc.,
NJ) filled with Tc-ggm, with a 0.5 mm thick attenuating lead sheet placed
over the left vantricular (LV) antedor wall, Tomograms were acquire(l: 1) of
the intact phantom; and 2) after introducing defects involving 10% of the LV
volume with activities ranging from 0 to 75°/=of peak activity. Transmission
and emission data were simultaneously acquired using a Ga-153 line source
and a dual-head tomogrephic (aPECT) system.Tomograms were generated
by an iterative maximum-likelihood reconstruction algodthm and evaluated
both qualitatively and quantitatively. On the tomogramsof the intact phantom,
diminished basal wall activity was noted when attenuation was not corrected,
56%
86%
DefectActivity
(P/,,
25%
12%
20=/0
26%
34%
50%
32%
53%
75%
44%
64%
Thus, in this phantom study, attenuation correction with a Ga-153 linesource provides on improved estimation of true myocardial count activity.
Importantly, small, mild defects were not obliterated by the correction algorithm. Scatter correction may further refine defect characterization,
~
Operator-Less Processing o f Myocardial SPECT
Studies
Guido Germano, Paul B. Kavanagh, Johnny Chen, Parker Waechter,
Hsiao-Te Su, Hosen Kiet, Daniel S. Barman, Cedars-Sinai Medical Center,
Los Angeles, CA
We have developed a completely automated algorithm for the processing of
myocardial pedusion SPECT data. The algorithm consists of three cascaded
software modules. The first module determines the reconstruction limits from
the projection imago set, using iteretive image convolution, feature detection,
and artificial intelligence techniques based on the expected anatomical size,
shape and location of the left ventdcular myocardium. The second module
reconstructs projection images into transaxial tomographic images using
standard filtered backprojection, and the third (previously described in J Nuol
Mad 1995; 36(6):1107-1114) reodants transaxlal into short-axis images.
The global process requires no operator interaction (although quality control
and manual processing capabilities are provided) and no use of prapfleta~
hardware. The software was written in C, ~ns on standard Unix platforms
using X-Windows, and executed in 12 seo per study on a SunSPARC5
computer.
The algo~thm was tested on 530 patient studies acquired on a triple detector (Picker Pdsm 3000, 193 studies), a 90= dual detector (ADAC Vertex,
200 studies) and a single detector camera (Siemens Orbiter, 137 studies).
A~!!~nte,~ and processing parameters were consistent with those used in
clinical routine. Reconstruction limits, the reodentation axis and the final
short axis images were presented simultaneously to, and evaluated by an
experianced operator, who judged the processing accurate in 92.6% of the
patients (177/193+ 185/200+ 129/137). We conclude that automatic processing of myocardial SPECT data is feasible, and may allow faster and more
reproducible analysis than currently achieved with manual approaches.
~
Attenuation Correction f o r Myocardial Peduslon
Imaging: Effect o n Perfuslon Images o f a
Scanning Transmission S o u r c e and Preliminary
Results of s Multicenter Trial
Robert C. Handel, Hosen S. Kiat, Brenda M. Teth, Steve Ballo, Daniel
S. Barman. Northwestern University Medical School, Chic~O, IL;
Cedars.Sinai Medical Center, Los Angeles, CA
A major concern with myocardial perfusion imaging is the effect of nonuniform soft-tissue attenuation and the potential production of imaging artifacts.
Recently, a method utilizing a scanning line source of Gd-153 has been developed which performs attenuation correction (AC) following the construction
of a transmission map (Vantage®, ADAG Lab., Milpitas, CA). The current
study sought to examine if the transmission source altered the standard
emission data, Accordingly, 4010is(20 men, 20 women) undenvant porfusion
imaging (E) after a peak-stress injection of ~25 mCi of Tc-99m sestamibL A
simultaneous emission-transmission iS) study was subsequently obtained.
S, E, and AC images were read independently and in a blinded fashion.
E and uncorrected S images were in agreement for the presence of an
abnormality in 39•40 patients (98°/=;Kappa = 0.90). When analyzed segmentally (sags), concordance was noted in 765/800 sags (96%; Kappa = 0.85).
Virtually identical numbers of segments were scored as normal for the E as
for S (658 vs. 657; p = n.s.). Sags were also evaluated semi-quantitatively on
a 0-4 scale; precise agreement was noted in 93=/0of studies, The summed
stress scores were also equivstent for E and S (5.7 vs. 5.8; p = n.s.). No systemic vadetlons were present based on gender, body size, or the likelihood
of CAD.
Preliminary efficacy data demonstrated that AC, when compared with E,
increased the number of normal segments (690 vs 657; p < 0.0001) and
reduced the summed stress score (4.5 vs. 5.8; p < 0.01). Additionally, more
pt studies were interpreted as r,ormal or probably normal with AC then with
S (28 vs. 21; p < 0.02).
In conclusion, the aforementioned technique for simultaneous emission.
JACC February 1996
ABSTRACTS-Poster 215A
transmission imaging does not adversely effect the emission data. Additionally, AC results in mere individual segments and overall pt studies being
interpreted as normal.
~
Quantitstlonof RegionalEjectionFractionsUsing
Gated Tomographic Imaging With
l¢-ggm-Sastamibi
E t ~ t h Kfodas, Paul J. Rogers, Lawrence J. Sinak, Thomas C. Hammall,
Michael K. O'Connor, Raymond J. Gibbons. Mayo Medical Center,
Rochester, MN
Gated single photon emission computed tomographlc (SPECT) sastamibi
imaging allows eimultaneous assessment of myeoer0~ pedusion and left
ventdcular function. We evaluated a technique based upon edge detection to
quantify regional eJeefionfre~ons (El=)in 15 ~ubjecta without cerdias pathology and 24 patients following myocardial infarction (5/11).After tomngrsnhic
reconstruction o! gated short axis slices, identification of endoeardialborders
was mode by a standard edge datan~on program in systole and diastole in
each of five sel~'~ed slices. Regional EPs were cafouls:ed for five regions
within each slice. Myocardial parfusfon was quantitated in the same reqions.
Ten patients unclatwent echocerdlegraphforegional wall motion (RWM) an~lyals which was compared to corresponding regional EF results. Results: A
high degree of interobserver repredu(~'billty In the assessment of regional
EF's was found with r values ranging from 0.94 to 0.98. In patients with antefior and infedor MI., regional EPs were abnormal in the anterior and septal,
and infedor and lateral regions, respectively. Regional EF's correlated sign a l l y with regional parfusion in anterior (r = 0.63, p < 0.0001), lateral (r
= 0.40, p < 0.0001), and inferior walls (r = 0.38, p < 0.0001). There was
a significant association between regional EF and echoce~iagrephio RWM
assessment at the base (p < 0.0001), mid ventricle (p = 0.004), and apex
(p = 0.0003). Conclusions: Gated tomographlc images with technatium-99mssalamibi can provide reproducible quantitative segmental regional EF's for
muif@e left ventdcular slices that are significantly associated with subjective
RWM assessment by echosardingrsphy.
~
Artificial Neural N e t w o r k f o r Automatic
interpretation o f Myocardial Perfusion $PECT
Jacob Erel, Parker Waschter, Joseph Areeda, Kenneth Van Train,
Hossn Kiat, John Friedman, Daniel Berman, Guido Gem~no. Cedars-Sine/
Med'~al Center, Los Angeles, CA
Interpretation of myocardial perfualon SPECT (MPS) is mainly subjective,
whereas quantitative programs (QP) are currently still not in widaspresd
clinical use. We developed a novel, flexible, non-rule based artificial neural
network (NN) which can uniquely "learn" from experience. A group of 140 pts
who underwent exercise sestsndbi SPEGT comprised our training set (TRG),
and a different group of 97 pts was used for tasting {TSG). Defeot extent and
severity data were dedvod from raw circumferential profiles, produced by an
optimized QP (CEqual), and ussd as NN inputs. Exercise scans were visually
scored (:2 experts) in 20 myocardial segments (5 point score: 0 to 4 ==normal
to no uptake) for use as NN and QP targets. Abnl scan was defined by a
summed score of >_2 to," >__2 segments. NN output scores were compared
to visual scores. The NN results were also compared to CEqual outputs for
the same TSG. The overall sensitivity/specificity for the NN and CEqual were
85/83% and 92/77%, respectively (p = no). Overall acoureo/results are as
follows:
Accu,rncy (%). Total ant
NN
B3
77
CC--quat
81
eS
ant.sept,, Inf.sept int
80
e2
77
7s
e6
ss
Inf-lat ant-lot apx
85
88
76
7s
77
7s
ROC curves for the NNand for CEqual show comparably highvelees.
Conclusion: these initial results suggest that our NN issultable for accurate
ldantiltcation of MPS defects.
Accuracy and Reprodualblllty o f Automated
Tomographle Venfficular F u n n i e s Measurements
Kenneth Nichols, Alan Rozanski, Helena Salenaky, E. Gordon DePuey. St.
Luke's.Roosevelt Hospital and Columbia University, New York, NY
Both manual and automated methods for comping left vsntflcuiar (LV) enddiastolic (ED) volume, end-systolic (ES) volume and ei.,~ctionfrection (EF)
from s¢lntigraphle gated tomogrems have been developed but reproducibility
and ~¢curacy of measurements versua other methods are not well known,
particularly since automated programs =an be confounded by conditions
such 8s reduced regional myocardial perfualon. In automated processing of
hodzontBI ~ vertical long axis Tc-99m sestamlbi gated myocardial perfusion
tomogrems of 145 patients (pts), experienced observers judged it necessary
to alter: (I} ED or ES frames in 7% of pts, (2) endocerdlal borders in 14%,
and (3) LV cavity center locations !n 28%o.Agreement among independent
observers using the software demonstrated correlation by linear regression of
r = 0.92, and compared to manual EF determinations as r = 0.90. Regression
analysis yielded r = 0.87 for tomegrephic EF vs. first pass EF in 67 pts.
Unear regression for tomegraphic EF vs. gated equilibrium EF in 77 pts was
r = 0.87, and similar correlation coefficients were computed for these pts
when subgrouped according to whether changes were or were not needed
for LV Center location (r = 0.91 vs. r = 0.881, or .~.~Xlocardialborders (r =
0.87 vs. r = 0.89); arid were likewise similar for k~ ~,rouperl according to
ED volume greater or less than the median value of 87.5 mt (r = 0,89 vs.
r = 0.87). We conclude that our automated algorithms are reproducible and
accurate comparedto independent EF methods, and that changes needed in
LV center location for over 25% of lots cause no degradation in EF accuracy.
~
Abselute Right and Loft Ventricular Volume a n d
_Ejection Fraction Measurements b y T o m o g r a p h i c
MUGA
Bennett B. Chin, Daniel C. eloomgardan, Wetshi Xia, Hee-Joung Kim, Zabi
A. Fayed, VictOrA. Fen'aft, Leon Axal, Abass Alevi. University of
Pennsylvania, Philadellohia"PA
To validate tomogrephic radionucllde gated equilibrium blood pool seintigraphy (TMUGA) absolutevolume arid ejection fraction measurements, gradientecho MRI, conventional first pass radionuclide (FP) and planar gated equiI~fium blood pool (PMUGA) scintigrephy were performed. Ten patients with
normal and abnormalventricular funo~an hod TMUGA and MRI. Ten pa~ents
also had TMUGA and FP. TMUGA was acquired using a 3-heeded camem and lb gated intervals. TMUGA ventricular volumes ware calculated by
Simpson's role with regions defined by 1) phase analysis to separate atria
and ventricles 2) threshold to define ventrleular free borders and 3) visual
inspection to define Interventdcular septum. Cardiac phantom studies were
performed to obtain appropriate volume threshold for patient TMUGA. As
previously validated, MRI studies were acquired on a 1.5 Teala system with a
multiphase, breath-hold, segmented k-space technique. MRI volumes were
calculated by Stmpson's rata using 12--16 6 mm thick parallel short axis slices
and regions defined by a semi-automatic contour tracking routine. Gated FP
rad'ronuslideasgiogrephy and PMUGA ware perfonoed for RVEF and LVEE
Volume measurementsat end-syatole and end-diastoleby TMUGA showed
good correlation with MRI for RV (r = 0.91, slope = 0.90, SEE = 15.7) and LV
(r = 0.96, slope = 0.88, SEE = 18.2). TMUGA also showed good correlation
for both RVEF and LVER
MRi
FP
TMUGARVEF
r
slope
0.88 0.79
0.88 1.2
SEE
6.0
7.9
MRI
PMUGA
TMUGALVEF
r
slope
0.94 1.10
0.97 1.23
SEE
9.0
6.2
TMUGA almws good accuracy in simultaneous measurements of absolute
fight and left ventficuiar volumes and election kactions.
Hypertension: Basic and Clinical
Tuesday, March 26, 1996, 9:00 a.m.--11:00 a.m.
Orange County Convention Center, Hall E
PresentationHour: 10:00a.m.--11:00aom.
~'~-~
Outseme in MedJcsily Managed Second O p i n i o n
Patients With C o r o n a r y Artery Disease and Short
Exercise Duration
Chdstion S. Bmburda, Cling Yao, Thomas B. Oreboys. Lown Card/ovascular
Center, Harvard University, Boston, MA
An exercise ~eadmi, test (ETT) duration of less than 6 mins (standard Bruce
protocol) is viewed as a poor sunMal predictor in patients with coronary
arrow disease (CAD). To determine the prognosis of short exercise duration
(SED) in patients with documented CAD, mfaned for a second opinion prior
to bypass grafting, we analyzed El-r and clinical data with a mean follow up
period of 6.1 yrs (SD 5.7) in SO consecutive pts (mean age was 69.3 yrs,
46 M), who received medical treatment. Clinical and ETT data revealed ador
myocardial infarct (MI) n ,, 25, a mean EF of 45%, a peak HR of 113.3, mean
resting BP of 134/75, peak BP of 139/74 and a mean ST segment depression
of I ram. A cardiac event was defined as cardiac death, nonfatalmyocardial
infamt, congestive heart failure (CHF) or revascula~zatianprocedure: PTCA
or CABG. Results: Cardiac events occurred in 22 pts, with nonfatal MI (n =
7), CHF (n = 6) and revascul~dzation (n = 15). Cardiac mortality was 12%
(n = 6), with ~tal MI (n = 1), 2 death ped-angloplasty and 3 peri-uperetivety.
Multtvefiale significant (p < 0.05) adverse prognostic predictors (Cox) were
presence of rest angina, b-blocker therapy end pro ETT use of s.L nitro.