Preoperative volume prediction in adult live donor liver transplantation: 3-D CT volumetry approach to prevent miscalculations.

Abstract

The precise preoperative calculation of functional liver volumes for both donor and recipient is a crucial part of the evaluation process in adult living donor liver transplantation. The purpose of this study was to describe and validate our modus 3-D CT volumetry.
Native (unenhanced), arterial, and venous phase CT images from 62 consecutive live liver donors were subjected to 3-D CT liver volume calculations and virtual 3-D liver partitioning. Graft-volume estimates based on our modus 3-D volumetry, which subtracted intrahepatic vascular volume from the "smallest" (native) unenhanced CT phase, were subsequently compared to the intraoperative graft-weights obtained in all 62 cases. Calculated (preoperative) liver-volume-body-weight-ratios and measured (intraoperative) liver-weight-body-weight-ratios of liver grafts were analyzed.
Preoperative calculations of graft-volume according to our modus 3-D CT volumetry did not yield statistically significant over- or under-estimations when compared to the intraoperative findings independent of their age or gender.
Our modus 3-D volumetry, when based on the "smallest" (native) unenhanced CT phase, accurately accounted for intrahepatic vascular volumes and offered a precise virtual model of individualized operative conditions for each potential live liver donor.

Full text

Page 1

EU RO PE AN JOUR NAL OF MED I CAL RE SEARCH319
Abstract
Background: The precise preoperative calculation of
functional liver volumes for both donor and recipient
is a crucial part of the evaluation process in adult liv-
ing donor liver transplantation. The purpose of this
study was to describe and validate our modus 3-D CT
volumetry.
Patients and Methods: Native (unenhanced), arterial,
and venous phase CT images from 62 consecutive live
liver donors were subjected to 3-D CT liver volume
calculations and virtual 3-D liver partitioning. Graft-
volume estimates based on our modus 3-D volumetry,
which subtracted intrahepatic vascular volume from
the “smallest” (native) unenhanced CT phase, were
subsequently compared to the intraoperative graft-
weights obtained in all 62 cases. Calculated (preopera-
tive) liver-volume-body-weight-ratios and measured
(intraoperative) liver-weight-body-weight-ratios of liv-
er grafts were analyzed.
Results: Preoperative calculations of graft-volume ac-
cording to our modus 3-D CT volumetry did not yield
statistically significant over- or under-estimations
when compared to the intraoperative findings inde-
pendent of their age or gender.
Conclusion: Our modus 3-D volumetry, when based
on the “smallest” (native) unenhanced CT phase, accu-
rately accounted for intrahepatic vascular volumes and
offered a precise virtual model of individualized oper-
ative conditions for each potential live liver donor.
Key words: Liver surgery; living donor liver transplan-
tation; liver volume; surgery planning in LDLT; 3-di-
mensional reconstruction; liver imaging
Abbreviations: 3-D: three dimensional; Phase: CT im-
age phase: native = unenhanced (N), arterial (HA), ve-
nous (V); ALDLT: adult living donor liver transplanta-
tion; CT: computed tomography; LVBWR: liver vol-
ume body weight ratio; LWBWR: liver weight body
weight ratio; MHV: middle hepatic vein; MRI: magnet-
ic resonance imaging; SD: standard deviation; SFS:
Small for size; TLV: total liver volume
INTRODUCTION
Accurate preoperative prediction of functional graft
and remnant volumes in adult live donor liver trans-
plantation (ALDLT) is essential in ensuring donor
safety and preventing postoperative graft failure. Two-
dimensional CT or MRI imaging has become the “cur-
rent standard” for total liver volume (TLV) and
graft/remnant volume estimations [1-3]. However, it is
well known that computer systems overestimate real
graft-volumes, and have an error ratio when compared
to the actual graft weight obtained at the time of
surgery [2-7]. The marked discrepancies in graft vol-
umes among pre- and intra- operative values urged
many groups to introduce “conversion” factors [8] or
modified formula-derived estimates [4, 9-10].
The purpose of this study was to describe and vali-
date our modus 3-Dimensional CT volumetry
PATIENTS AND METHODS
1. STUDY POPULATION
Between January 2003 and June 2006, sixty two (f:m =
29:33) of the 103 potential donors evaluated according
to our routine protocol [11-13] ultimately underwent
graft hepatectomy for transplantation. Mean age was
36 ± 10 years. Forty nine of the 62 grafts obtained
were right lobes that included the middle hepatic vein
(MHV). Of the remaining thirteen grafts, 6 were right
lobes without MHV and 7 were left grafts that includ-
ed the MHV. Biopsy results in all resected donors
showed less than 10% steatosis and no evidence of
histopathologic changes.
2. STUDY DESIGN
Multiphasic CT image data from 62 consequtive live
liver donors were prospectively analyzed by 3-D CT
for graft and remnant liver volume calculations, by
utilising the software assistant HepaVision (MeVis,
July 28, 2008
Eur J Med Res (2008) 13: 319-326 © I. Holzapfel Publishers 2008
PREOPERATIVE VOLUME PREDICTION IN ADULT LIVE DONOR LIVER
TRANSPLANTATION: 3-D CT VOLUMETRY APPROACH TO PREVENT
MISCALCULATIONS*
A. Radtke1, G. C. Sotiropoulos1, S. Nadalin1, E. P. Molmenti1, T. Schroeder2, F. H. Saner1, G. Sgourakis1,3,
V. R. Cicinnati1, C. Valentin-Gamazo1, C. E. Broelsch1, M. Malagó1, H. Lang1
1Department of General, Visceral and Transplantation Surgery, University Hospital Essen, Essen, Germany,
2Department of Diagnostic and Interventional Radiology, University Hospital Essen, Essen, Germany,
32nd Surgical Department, Korgialenio-Benakio Red Cross Hospital, Athens, Greece
*This study was supported by the Grant Nr: 117/1-1:A2.2
from the German Society for Research.

Page 2

Germany), and subject to virtual 3-D liver partitions
by two expert surgeons (A.R. and M.M.) and an expe-
rienced radiologist (T.S.).
Step one: The calculation of total liver volume
(TLV) was carried out in a stepwise analysis. Liver
parenchymal segmentation was evaluated independently
in three separate axial 2-D CT imaging phases: phase
N (native = unenhanced), phase HA (arterial), and
phase V (venous). The data obtained enabled a multi-
phasic CT imaging comparison of TLV-estimates for
each individual donor candidate (Fig. 1a-c). The vol-
ume of the intrahepatic vascular tree was calculated
after the segmentation of hepatic vessels phase.
Step two: A virtual 3-D liver partition simulating the
“carving” technique, which exactly follows the course
of middle hepatic vein (MHV), routinely employed at
our institution [14]. This calculation was performed in
all cases using CT phase V images, usually considered
the “standard” CT imaging for donor evaluation in
LDLT [3, 14] (Fig. 2a-b). This step allowed for an ini-
tial estimation of graft- and remnant- liver volumes.
Step three: Subsequently the TLV was calculated ac-
cording to our modus 3-D volumetry, by subtracting
the intrahepatic vascular volume from the “smallest”
TLV (derived from the “smallest” CT phase). The
“smallest” TLV constituted the baseline for the addi-
tional estimation of “smallest” graft/remnant vol-
umes, which were calculated based on the volume per-
centage values for graft- and remnant livers obtained
during virtual 3-D liver partition for “standard” CT
phase V (step two). Finally the definitive preoperative
-graft- and remnant-LVBWRs for each live liver donor
candidate were calculated.
Step four: The intraoperative graft weight was mea-
sured in 62 live liver donors who underwent graft he-
patectomy. All resections were performed according to
the “carving” transection of the virtual 3-D preopera-
tive simulation.
The exact transposition of the virtual transection
plane onto the operative field was confirmed in all cas-
es by photographic documentation and doppler-scan-
ning (presenting the MHV and it`s tributaries on the
resection surface). This allowed for the retrospective
calculation of “actual” intraoperative graft-liver-
weight-body weight- ratios (LWBWR`s).
3. CT PROTOCOL
CT imaging as originally published by Schroeder et al.
[3, 15] was performed using a 16-row-Multidetector-
CT-Scanner (Sensation16®, Siemens, Germany) using
the following parameters: kVp 120, mAs 140-170, slice
collimation 0.75mm, feed/rotation 12mm, and rotation
time 0.5 sec. Reconstruction increments were 1mm for
the native = unenhanced, arterial and venous scans.
4. IMAGE ANALYSIS AND VIRTUAL RESECTION
CT images were analyzed with the non-commercial
software assistant HepaVision (MeVis, Germany). This
software allows for the calculation of 1) total liver vol-
ume together with graft- and remnant- liver volumes as
well as 2) volume of the intrahepatic vascular tree.
Liver parenchyma imaging was derived from CT data
in a semi-automatic way. Segmentation of liver parenchy-
ma was performed on axial slices with a modified live-
wire algorithm. With this approach, contours between
user-defined boundary points were determined auto-
matically based on CT values and gradients during user-
interaction. Parameters of the algorithm were adapted
to each CT phase, and manual correction of the auto-
matic delineation contours as well as manual drawing of
contour parts was undertaken to ensure accurate liver
segmentation. The live-wire contours were interactively
determined on 3 mm axial 2-D CT slices during venous
(V), arterial (HA) and native = unenhanced (N) phases.
The contours of intermediate slices were automatically
interpolated and optimised initially by the software and
ultimately by the operator, summarizing all segmented
areas and yielding volumetric calculations in milliliters
(ml). All surrounding structures, including major extra-
hepatic vessels (portal vein, hepatic artery, inferior vena
cava) and gallbladder fossa, were excluded.
EUROPEAN JOURNAL OF MEDICAL RESEARCH320July 28, 2008
Fig. 1a-c. Assessment of total liver volume (TLV). Liver con-
tours were traced with a modified live-wire, semi-automated,
contour-finding algorithm approach. The live-wire contours
were independently obtained from venous (a), arterial (b) and
native = unenhanced (c) phase 3-mm, axial, 2-D CT images.
a
b
c

Page 3

During the segmentation of hepatic vessels, arterial,
portal and hepatic venous systems were extracted from
the image data by using a filter for noise reduction and
background compensation and a region-growing algo-
rithm [16]. Intrahepatic vessels were automatically
analysed and transferred into a hierarchical graph de-
picting dependencies between branches and direction
of blood flow. Relevant branches of subtrees were la-
belled during exploration of the 3-D venous graph [17].
Virtual resections were performed in the resulting
individual 3-D liver model that allowed optional dis-
play of vascular trees and territories. The volume aris-
ing from the manually (surgeon line) defined grafts
and remnants was automatically calculated.
5. LIVER PARTITION
The plane of transection in the liver partition (Fig. 3a-
c) follows the course of the MHV (“carving tech-
nique”) [14]. In nearly most cases, the MHV remained
with the graft during the procuring resection. In each
instance the MHV was initially identified by intra-op-
erative ultrasound examination and subsequently
“carved” out of the surrounding remnant liver par -
enchyma. The transection plane lied exactly over the
MHV, leaving its left-or right-sided border exposed on
the transection surface of the graft. The level of divi-
sion of the MHV trunc depended on its anatomical
relationship with tributaries from segments IVa and
VIII and the size of their drainage territories.
6. INTRAOPERATIVE FINDINGS AND STATISTICAL
ANALYSIS
All 62 recipients of the live donor grafts underwent
venous outflow tract reconstruction by means of
our “blanket” technique [18]. Each liver graft was
weighed immediately after retrieval. Comparisons
between preoperatively calculated volumes and
intraoperatively measured weights were performed
by considering a specific weight of healthy liver
parenchyma of 1 gm/ml [19]. Calculations of graft-
LVBWR/LWBWR followed previously described for-
mulas [11, 13, 20].
EUROPEAN JOURNAL OF MEDICAL RESEARCHJuly 28, 2008 321
Fig. 2a-b. Donor virtual hepatectomy. Malagó partition (“carving technique”). The plane of transection runs along the course
of the MHV – 2D view (2a). The MHV is “carved” out of the surrounding hepatic parenchyma – 3D cranial view (2b). RHV
(blue), MHV (yellow), LHV (red), right graft (green), left liver remnant (brown).
Fig. 3a-c. Malagó partition (“carving technique”), intraoperative view. The transaction line on the liver surface follows the
course of MHV (a). Left remnant liver with the “MHV groove” on the transaction surface (b). Right liver graft including the
MHV (c).
a
b
a
b
c

Page 4

Results were expressed as mean volume percentage
(%) ± standard deviation (SD) values. Continuous
variables were analyzed by one-way analysis of vari-
ance and t-test when normal distribution was given.
Univariate and multivariate designs analyzed by factor-
ial Anova examined relations between gender and/or
age (as categorized or continuous variable) and graft
volume error ratio or graft to GVBWR error ratio. P-
values of 0.05 or less were considered significant.
The error ratio (%) was calculated as [E-A] /A x 100,
where E was the estimated volume (ml) and A was the
actual weight (gm) as described by Hiroshige et al. [21].
Preoperative volume- / LVBWR- values with devia-
tions of less than ≤ 1% from intraoperative values
were denoted as “identical”.
Major postoperative morbidity was defined as all
life threatening events requiring invasive procedures,
re-operations, hospital stays longer than 30 days, and
re-hospitalizations.
RESULTS
1. TOTAL LIVER VOLUME (TLV) CALCULATION:
COMPARISON AMONG DIFFERENT CT PHASES
Among the 62 live liver donors prospectively evaluat-
ed, the largest mean TLV (1596 ± 212ml) was ob-
tained with the venous (V) CT phase measurements.
The smallest mean TLV (1456 ± 196ml) was obtained
in the native = unenhanced (N) CT phase in 61(98%)
cases and in the arterial (HA) CT phase in n = 1(2%)
of livers, respectively. The mean difference between
largest (phase V) and smallest (phases: N or HA) TLV
was 142 ± 78ml (p<0.001).
The calculated mean intrahepatic vascular volume
for all 62 live liver donors, who underwent graft hepa-
tectomy was 88 ± 17ml.
2. VIRTUAL LIVER PARTITION: GRAFT- VS.
REMNANT- VOLUME PERCENTAGES
There were 62 donors who underwent resection for live
donation. Based on the 3-D virtual liver partition de-
rived from the CTphaseV,themeanrightandlefthemi -
liver volumes were 63 ± 8% and 37 ± 8% of the TLV.
3. INACCURACY INCIDENCES FOR OVER- VS.
UNDER-ESTIMATED GRAFT-VOLUMES AND -LVBWRS:
3-D CT VOLUMETRY VERSUS INTRAOPERATIVE DATA
Table 1 outlines in detail the incidence of the over- vs.
under-estimated graft-volumes and -LVBWRs when
comparing data derived from our -modus 3-D vol-
umetry and the intraoperative findings in the trans-
planted subgroup (n = 62).
The graft-volume calculations based on 3-D CT
volumetry yielded an overestimation in 61% of in-
stances. Graft-LVBWR was overestimated in 42% of
cases. In 24% (n = 15) of cases, an identical graft-vol-
ume, when compared to the intraoperative graft-
weight, was predicted. A graft-LVBWR identical to the
intraoperative graft-LWBWR value was calculated in
over half (52%) of all transplants. There were 9 un-
derestimations (15%) in the graft-volume, and n = 4
(6%) in the graft-LVBWR, respectively (Table 1).
When addressing the subgroup of 7 graft livers with
“real” intraoperative LWBWR between 0.9 and 0.8,
and an additional 6 recipients with a “real” graft-LWB-
WR < 0.8, our 3-D CT volumetry overestimated graft-
LVBWR values in 6 (46%) of them. There were no
underestimations of graft-LVBWR seen in these n =
13 live liver donors.
4. OVERESTIMATION ERRORS OF GRAFT-VOLUME AND
GRAFT-LVBWR: 3-D CT CALCULATION VERSUS
INTRAOPERATIVE DATA
Our- modus 3-D CT volumetry provided mean graft
volumes of 847±187ml for the 62 cases who under-
went donor hepatectomy. This did not represent sig-
nificant overestimations (p = 0.229) with respect to
the intraoperative mean actual graft weights of 808 ±
169gm as shown in Table 2. There was a mean overes-
timation error of 7% with our 3-D CT volumerty for
graft-volumes, although in 5% (n = 3) of live liver
donors the overestimation error was higher than 25%.
We did not observe any statistically significant differ-
ences (p = 0.287) among virtually calculated graft-
LVBWR and intraoperatively obtained graft-LWBWR
values as delineated in Table 2. There was only a 6%
mean overcalculation error with our 3-D volumetry
EUROPEAN JOURNAL OF MEDICAL RESEARCH322July 28, 2008
Table 1. Comparison of cases with over-estimated, under-estimated, and identical results from our 3-D CT volumetry with re-
spect to intra-operative values for graft-volume and graft-LVBWR.
n = 62 Our modus 3-D CT volumetry
Graft-volume Graft-LVBWR
Cases of overestimation 38 26
(61%) (42%)
Cases of underestimation 9 4
(15%) (6%)
Cases with identical values 15 32
(24%) (52%)
Our-modus: native = unenhanced-phase CT volume (intrahepatic vessel volume subtracted); LVBWR: liver volume body
weight ratio; identical: ≤1% deviation from intraoperative findings.

Page 5

for graft-LVBWR, although an overcalculation higher
than 15% was seen in 5% (n = 3) of our live liver
donors.
5. OVERESTIMATION ERROR OF VIRTUAL GRAFT-VOL-
UME AND -LVBWR CALCULATIONS: ANALYSIS IN
RELATION TO THE DONOR AGE AND GENDER
Seventy four percent (n = 46) of the live liver donors
(f:m = 22:24), who underwent graft hepatectomy had
a less than 10% overestimation error for graft-volume
and –LVBWR when compared to the “actual” intraop-
erative graft-weight and –LWBWR values. Five percent
(n = 3) of donors had a greater than 20% overestima-
tion error for graft-volume and –LVBWR when com-
pared to the intraoperative calculations, including 2 fe-
males and one male. In these 3 cases, the “real” intra-
operative graft-LWBWRs were of: 0.79, 0.83, and 1.04,
respectively. Despite low “real” graft-LWBWR values,
there was no evidence of postoperative small for size
(SFS) syndrome in these 3 graft recipients.
5a. Age correlation test
Neither univariate nor multivariate designs analyzed by
factorial Anova, disclosed any statistically significant
differences considering relations between age (as cate-
gorized or continuous variable) and graft volume error
ratio or graft to GVBWR error ratio (p = 0.4075).
5b. Donor subgroup: females versus males
Twenty nine (33.2%) live liver donor candidates who
underwent graft hepatectomy were females. Among
them, there was a mean overestimation error of 8.17%
(range: 0.1-29%) and 6.75% (range: 0-28%) for graft-
volume and –LVBWR, respectively, when compared to
intraoperative values. Thirty three (46.8%) live liver
donors who underwent resection were males. In these
cases, the mean overestimation errors for graft-volume
and –LVBWR were 6.12 % (range: 0-27.3%) and
5.51% (range: 0-19.1%), respectively. Neither univari-
ate nor multivariate designs analyzed by factorial Ano-
va, disclosed any statistically significant differences
considering relations between age (as categorized or
continuous variable) and graft volume error ratio or
graft to GVBWR error ratio (p = 0.4679).
6. 3-D CT VOLUMETRY: ANALYSIS OF
UNDERESTIMATION-RATE AND -ERROR FOR
GRAFT-VOLUME AND -LVBWR
There were underestimations for graft-volume in nine
(15%) recipients, and for graft- LVBWR in four (6%)
cases. However, we did not observe any significant dif-
ferences (p = 0.526) between their virtually calculated
mean graft-volume of 771 ± 158ml and the mean “ac-
tual” intraoperative graft-weight of 822 ± 156ml. The
mean underestimation error for graft-volume was
7.8%. The difference between the preoperatively cal-
culated mean graft-LVBWR of 1.2 and the mean
“real” intraoperative graft-LWBWR value of 1.29 was
not significant (p = 0.172). The mean underestimation
error for graft-LVBWR was 7.65%.
7. DONOR AND RECIPIENT OUTCOME AFTER 62
LDLTS DURING A MEAN FOLLOW UP PERIOD OF 21 ±
12 MONTHS
7a. Donor subgroup
There were no donor deaths. In 16% (n = 10) of
donors the retrospectively calculated “intraoperative”
LWBWR for the remnant liver was >0.8, while in 37%
(n = 23) of cases a LWBWR value between 0.8 and
0.7 was calculated, respectively. The remaining 29
(47%) donors had a LWBWR value of less than 0.7.
We did not experience postoperative liver insufficiency
associated with SFS remnants in any of the 33 operat-
ed donors who had “intraoperative” LWBWR values
≥0.7. One of these 33 donors developed a subphrenic
abscess due to a bile leak early after a right graft hepa-
tectomy including MHV. He was successfully reoperat-
ed without further morbidity. One donor, who had
“intraoperative” LWBWR of 0.5, developed a tran-
sient SFS syndrome manifested by cholestasis (peak
total serum bilirubin of 21mg/dl) and coagulopathy
(drop in thrombine time to 23%). He recovered spon-
taneously and was discharged from the hospital on
postoperative day 27. There were no late vascular or
biliary complications within 3-6 months after LDLT in
any of the 62 donors. Recovery time ranged from 9-12
weeks, after which all donors resumed their preopera-
tive occupational activities.
EUROPEAN JOURNAL OF MEDICAL RESEARCHJuly 28, 2008 323
Table 2. Comparison of CT-derived and intra-operatively obtained graft-volume and graft-LVBWR values.
Our modus 3-D CTvolumetry versus intra-OP values
n = 62 overestimation error (%) Our-modus OP
Graft
volume / weight mean 7% mean 847±187 mean 808±169
p = 0.229
Graft
LVBWR / LWBWR mean 6% mean 1.16±0.3 mean 1.12±0.03
p = 0.287
Our-modus: smallest CT volume (intrahepatic vessel volume subtracted); OP: weight obtained intraoperatively ; LVBWR: liver
volume body weight ratio; LWBWR: liver weight body weight ratio.