Received: 17 November 2023 Revised: 25 May 2024 Accepted: 8 June 2024

DOI: 10.1111/vsu.14142

CLINICAL RESEARCH

Computed tomography scan accuracy for the prediction

of lobe and division of liver tumors by four
board-certified radiologists

Brian J. Thomsen DVM 1 | Michael Ward BVSc, MSc, MPVM, PhD, FACVSc, DVS 2 |
Jin Y. Heo DVM, MS, DACVR 3 | Elizabeth Huynh DVM, MS, DACVR 3 |
Marc A. Ledesma DVM, MS, DACVR 4 | Jason A. Fuerst DVM, MS, DACVR 5 |
Arathi Vinayak DVM, DECVS (Small Animal), ACVS Fellow, Surgical Oncology 6

1
Department of Surgery, BluePearl Pet
Hospital, Levittown, Pennsylvania, USA Abstract
2
Sydney School of Veterinary Science, The Objective: (1) Evaluate the accuracy of computed tomography (CT) scans for
University of Sydney, Camden, New localization of liver masses. (2) Assess the agreement between radiologists on
South Wales, Australia
3
localization. (3) Determine if location influences the accuracy of localization
Department of Diagnostic Imaging, VCA
West Coast Specialty and Emergency
and histopathologic diagnosis. (4) Determine what lobar vasculature radiolo-
Animal Hospital, Fountain Valley, gists found most useful for localization.
California, USA Study design: Retrospective.
4
Department of Diagnostic Imaging, VCA
Animals: A total of 67 client-owned dogs with a total of 75 hepatic masses.
California Veterinary Specialists Ontario,
Ontario, California, USA Methods: Records were reviewed for relevant data. Localization for each
5
Department of Diagnostic Imaging, VCA hepatic mass was performed by four radiologists (JH, EH, ML, JF)
California Veterinary Specialists Murrieta, independently.
Murrieta, California, USA
6
Results: Overall accuracy of mass localization was 217/292 (74.3%) by lobe
Department of Surgical Oncology, VCA
West Coast Specialty and Emergency and 264/300 (88%) by division. Accuracy for the quadrate lobe (11/27, 40.7%)
Animal Hospital, Fountain Valley, was lower (p < .05) than for the caudate process of the caudate lobe (19/24,
California, USA
79.2%), left medial lobe (47/64, 73.4%) and left lateral lobe (95/101, 89.6%).
Correspondence Accuracy for the right lateral lobe (17/35, 48.6%) was lower (p < .05) lower
Brian J. Thomsen, Department of Surgery, than for the left lateral lobe (95/101, 89.6%). Accuracy of localization was
BluePearl Pet Hospital, Levittown,
173/192 (90.1%) for masses located in the left division, 37/48 (77.1%) in the
PA 19056, USA.
Email: [email protected] central division, and 53/60 (88.3%) for the right division. The agreement
(kappa) between radiologists was good (0.61–0.8) to excellent (0.81–1) for divi-
sion and moderate (0.41–0.6) to good for lobe localization.
Conclusion: CT localization was more accurate for division than lobe localiza-
tion of canine hepatic masses. Similarly, radiologists had a better agreement
for division than lobe localization.
Clinical significance: This study supports CT as a useful modality for liver
mass localization based on division. CT localization to specific lobes should be
interpreted with some caution.

Veterinary Surgery. 2024;53:1313–1325. wileyonlinelibrary.com/journal/vsu © 2024 American College of Veterinary Surgeons. 1313
 1532950x, 2024, 7, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vsu.14142 by Mississippi State University, Wiley Online Library on [03/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1314 THOMSEN ET AL.

1 | INTRODUCTION Although other modalities have some utility for

detecting liver masses, computed tomography (CT) scan
Primary hepatobiliary tumors in dogs are uncommon is generally considered the gold standard in veterinary
and are found in 0.9% of all canine necropsies.1 They rep- medicine.3,10 The accuracy of ultrasound for detecting
resent between 0.6% and 1.3% of all canine neoplasia.2 liver masses is reported to be 99%.3 The accuracy of
The liver is also a common site of metastatic spread. Pri- ultrasound for mass localization by division has been
mary liver tumors are typically broken into three main reported between 51.8% and 74%.3,11,12 However, the
morphologic categories.3 Large solitary masses are lobe was only able to be correctly identified 19% of the
defined as massive liver tumors and are typically con- time.3 In human medicine CT and surgical descriptions
fined to a single lobe. Multifocal masses involving several were found to agree on the location of 92% of hepatic
lobes are defined as nodular tumors. Diffuse tumors lesions.13 To the author's knowledge, only three other
occur in multiple lobes and can be the result of coalesc- studies exist that have evaluated the accuracy of CT for
ing nodular disease or significant effacement of the liver liver mass localization in dogs.8,11,14 Using the location
parenchyma. of the hepatic lobar vein or portal vein nearest to a mass
Approximately 70% of dogs diagnosed with liver mass as landmarks on CT had 93% agreement with the surgi-
have clinical signs.3 These signs include lethargy, vomit- cal assessment of division and 56% agreement with
ing, inappetence, anorexia, weight loss, diarrhea, poly- regards to lobation in one study.14 This agreement
uria/polydipsia, melena, seizure, and ataxia. Most dogs information has limited utility given that the evaluation
(95%) will have at least one liver value elevated.3 Given was performed by a single radiologist. An agreement of
the nonspecific nature of clinical signs and laboratory 90.5% and 84% between divisions based on CT and
abnormalities, imaging is of critical importance for diag- operative findings has been reported in other stud-
nosing liver tumors and for surgical planning. ies.8,11 However, only liver division was assessed, not
The liver in dogs is deeply fissured and can be ana- lobation. Additionally, only a single radiologist and a
tomically divided at multiple levels of specificity.4 The consensus between two radiologists were compared to
canine liver can be classified as left, central, and right surgical findings.
divisions.5,6 The left division is subdivided into the left A multiradiologist study of the accuracy of CT for
lateral and left medial lobes.5,6 The central division is liver mass localization has yet to be performed in veteri-
subdivided into the quadrate lobe and the right medial nary medicine. The goal of this study was to: (1) evaluate
lobe.5,6 The right division consists of the right lateral lobe the accuracy of CT scans for localization of liver masses;
and the caudate lobe.5,6 The caudate lobe is further subdi- (2) assess the agreement between radiologists on localiza-
vided into the caudate process and the papillary tion; (3) determine if location influences the accuracy of
process.5,6 localization and histopathologic diagnosis and (4) deter-
Surgery is the hallmark of treatment for well-defined mine what lobar vasculature radiologists found most use-
liver masses.4 Accurate preoperative localization of ful for localization.
masses is paramount in determining surgical resectability
and expected outcomes.4,7–9 The lobes of the left division
are anatomically well separated and mobile, making re- 2 | MATERIALS AND METHODS
section less challenging. The attachments and associated
vascular anatomy make the resection of central and right 2.1 | Study subjects
divisional tumors more difficult. A significantly higher
risk of complications was found with right divisional This was a retrospective study. Medical records
masses and two dogs with masses in the right division between September 2016 and August 2022 were inter-
exsanguinated after attempted resection.7 An association rogated for inclusion criteria. Dogs were eligible for
between right medial lobe excision and a higher risk of inclusion if they underwent an abdominal CT scan
perioperative death (odds ratio [OR] = 9.2) has also been (with accessible images), had a liver mass greater than
identified.8 Disease within the quadrate lobe was also 2 cm, had a surgically determined localization, and
identified as a negative prognostic indicator.8 More had complete medical records. Dogs were excluded if
recently, complications during surgery were found to be the mass was less than 2 cm in maximal diameter, if
less common with left divisional masses than right divi- the CT images were incomplete, and if diffuse liver
sional masses but not central.9 The anatomical divisions disease prevented the identification of any singular
of the liver lobes and, therefore mass location, are typi- mass. Dogs with two liver masses were included if the
cally clear at the time of surgery but may not always be masses were located in two separate lobes, to prevent
as apparent on imaging. ambiguity.
 1532950x, 2024, 7, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vsu.14142 by Mississippi State University, Wiley Online Library on [03/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
THOMSEN ET AL. 1315

2.2 | Procedures (portal, hepatic, both) were made. Data was collected
from medical records including breed, age, sex, weight,
From the start of the study until April 17, 2022 CT scans date of CT, histopathologic diagnosis, size of mass on his-
of the abdomen were captured using a multidetector topathology, location (lobe and division) at the time of
(64 slice) CT scanner (Aquilion Prime, Toshiba Medical surgery, and if resection was complete on histopathology.
Systems, Tokyo, Japan) with helical acquisition. After
April 17, 2022 CT scans of the abdomen were captured
using a multidetector (128 slice) CT scanner (Aquilion 2.3 | Measures of outcome
Prime, Toshiba Medical Systems, Tokyo, Japan) with
helical acquisition. Both scanners were determined to The primary measure of outcome for this study was the
produce images adequate for evaluation. All scans were accuracy of radiologists to localize liver masses by divi-
performed in sternal recumbency with 2–5 mm slice sion and by lobe. Division, lobation, and histopathologic
thickness depending on body size. Transverse precontrast diagnosis were also evaluated for their influence on
images were reconstructed in soft tissue window accuracy. The secondary measure of outcome was the
(400 WW, 40 WL). Pre- and post-contrast images agreement between radiologists in their localizations by
obtained in the arterial, portal, and delayed phases were division and lobe. Tertiary measures were radiologist
acquired and viewed at the radiologist's discretion. Non- agreement on CT measured size; CT measured size com-
ionic, iodinated positive contrast medium (2.2 mL/kg; pared to histologic size; the correlation between histo-
Omnipaque [iohexol] 350 mgI/ml, GE Health Care, pathologic diagnosis and lobation determined by
Princeton, New Jersey) was administered to all dogs surgery; the correlation between histopathologic diagno-
intravenously into the cephalic vein as a bolus using a sis and division determined by surgery; and what vascu-
power injector (MEDRAD Stellant [Bayer, New Jersey, lature on CT radiologists found to be most useful in
USA]) at injection dosages of 1 mL/kg for dogs with a localization.
bodyweight of less than 10 kg and 2 mL/kg for dogs with
a bodyweight of 10 kg or greater. A 30-s and 5-min delay
was used for the portal and delayed phases, respectively. 2.4 | Statistical analysis
The CT images were reviewed by four ACVR board-
certified radiologists (JH, EH, ML, JF) trained and Statistical analysis was performed by a statistician (MW).
employed at three private practice hospitals with The Shapiro–Wilk test was used to test the normality of
4–10 years post-residency experience. The radiologists data distribution. The Kruskal-Wallis test was used to
were blinded to all data except the name and patient evaluate the association between weight and histologic
number to retrieve the CT scans. It is possible that two of diagnosis, and between age and histologic diagnosis, with
the radiologists could have analyzed the scans previously; pairwise comparisons adjusted to p-value <.05. Pearson
however, well over 1000 CT scans were performed in the χ 2 tests were used to evaluate associations between the
study period and radiologists are not routinely informed histologic diagnosis and whether complete margins were
on surgical outcomes so there was very little concern for obtained; the liver lobe and whether complete margins
bias. Multiplanar images reformatted into soft tissue algo- were obtained; liver division and whether complete mar-
rithm (window width = 400, window level = 40) were gins were obtained; histopathologic diagnosis at surgery
viewed and analyzed on Antech Imaging Services Work- and location; histopathologic diagnosis at surgery and
station (Antech Imaging Services Workstation, Antech division; histopathologic diagnosis and accuracy of locali-
Diagnostics Inc, Fountain Valley, California). The liver zation by lobe; and histopathologic diagnosis and accu-
masses were assessed in transverse, dorsal, and sagittal racy of localization by division. A p-value <.05 was
reconstructions; all CTs had arterial, portal, and delayed considered statistically significant. Accuracy was calcu-
venous phases that were assessed. Size was defined as the lated based on what the radiologists reported compared
widest recorded diameter in centimeters. The accepted to findings at surgery, and 95% confidence intervals were
vascular anatomy of the liver was used by radiologists to estimated. Inter-rater reliability was calculated to assess
determine mass location.5,14,15 The location of the mass radiologist agreement for lobation and division using
was recorded as left lateral (LL), left medial (LM), quad- Cohen's kappa coefficient. The criteria for agreement
rate (QU), right middle (RM), caudate processes of the were based on kappa ≤0 = poor, 0.01–0.2 = very weak,
caudate lobe (CC), and papillary process of the caudate 0.21–0.4 = weak, 0.41–0.6 = moderate, 0.61–0.8 = good,
lobe (CP). From this, a division was assigned of left (LD), 0.81–1 = excellent.16 Spearman rank correlation coeffi-
central (CD), or right (RD). Assessments of which cient was used to evaluate agreement between radiolo-
lobar vasculature was most helpful in localization gists in terms of size. Spearman rank correlation
 1532950x, 2024, 7, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vsu.14142 by Mississippi State University, Wiley Online Library on [03/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1316 THOMSEN ET AL.

interpretations were based on the following: 0–0.19 (20%) were in the right division. Two masses spanned
“very weak,” 0.20–0.39 “weak,” 0.40–0.59 “moderate,” more than one lobe but were isolated to a single division.
0.60–0.79 “strong,” 0.80–0.99 “very strong.” The statistical These were double-counted in the statistics below
software used for all tests was IBM SPSS Statistics (77 total) but were excluded from other analyses involv-
version 28. ing lobation (73 total). By lobation, 31/77 (40.26%) of
masses were located in the left lateral lobe, 19/77
(24.68%) were located in the left medial lobe, 3/77 (3.9%)
3 | R E SUL T S were located in the quadrate lobe, 9/77 (11.69%) were
located in the right medial lobe, 5/77 (6.49%)
3.1 | Description of sampled animals were located in the right lateral lobe, 3/77 (3.9%) were
located in the papillary process of the caudate lobe, and
A total of 68 CT scans were eligible for inclusion. One 7/77 (9.09%) were located in the caudate process of the
was excluded because the CT scan was performed at a caudate lobe. All but one case had histopathology results
different facility. One dog was presented with eligible available. This dog died before leaving the hospital. One
masses at two separate points in time (2 years apart) and dog had two neoplasms in a single mass (i.e., collision
was therefore counted separately. These 67 dogs had a tumor) and was counted twice for the reporting of histo-
total of 75 masses eligible for evaluation. Two dogs were pathologic data. For all other analyses using histologic
excluded from lobe analysis because the mass spanned data, the primary diagnosis was used and the mass was
multiple lobes within a single division making for a total not counted twice. Histopathology reports revealed 11/75
of 73 masses included in lobation analysis later in this (14.67%) masses were nodular hyperplasia, 38/75
paper. Of dogs that were included in the study 35/67 (50.67%) were well-differentiated hepatocellular carci-
(52.2%) were spayed females and 32/67 (47.8%) were neu- noma, 10/75 (13.33%) were adenoma, 10/75 (13.33%)
tered males. Of the 75 masses that were included in our were other neoplasms, and 6/75 (8%) were other
study 40/75 (53.3%) were in spayed females and 35/75 non-neoplasms. Other neoplasms consisted of histiocytic
(46.7%) were in neutered males. The weight of the 67 dogs sarcoma, fibrosarcoma, exocrine pancreatic adenocarci-
was not normally distributed (Shapiro–Wilk statistic noma, hemangiosarcoma, tubular adenocarcinoma, spin-
p < .001) with a median of 13.8 kg and an interquartile dle cell sarcoma, poorly differentiated hepatocellular
range (IQR) of 8.22–27.9. The weight of dogs across the carcinoma, biliary cystadenoma, liposarcoma, and
75 masses was not normally distributed (Shapiro–Wilk moderately differentiated hepatocellular carcinoma.
statistic p = .019) with the same median and IQR. Of the Non-neoplastic masses consisted of abscess, suppurative
67 dogs, 42/67 (62.7%) were >10 kg (large breed) and hepatitis, sinus telangiectasia, hepatic cyst, and diffuse
25/67 (37.3%) were < 10 kg (small breed). Of the hepatocellular vacuolar change.
75 masses, 48/75 (64%) were in dogs >10 kg (large breed)
and 27/75 (36%) were in dogs <10 kg (small breed). The
age of the 67 dogs was not normally distributed (Shapiro– 3.3 | Results of statistical analyses
Wilk statistic p = .018) with a median of 11.1 years and
IQR 9.5–13.0. The age of dogs across the 75 masses was The radiologists made 292 (73 eligible masses  4 radiol-
not normally distributed (Shapiro–Wilk statistic ogists) localizations by lobe. Of these, 217/292 (74.3%,
p = .040) with a median of 11.1 years and IQR 9.3–13. Of 95% CI: 69.6–79.6) were correct. A summary of individual
the 67 dogs, 22/67 (32.8%) were mixed breed dogs, 6/67 radiologist accuracy in lobation evaluation is in Table 1.
(8.9%) were Golden Retrievers, and 4/67 (6%) were Chi- A summary of accuracy in each lobe is in Table 2.
huahuas. Of the 75 masses, 25/75 (33.3%) were in mixed Accuracy for quadrate was lower than for the caudate
breed dogs, 7/75 (9.33%) were in Golden Retrievers and process of the caudate lobe, left medial lobe, and left
4/75 (5.3%) were in Chihuahuas.

3.2 | Description of the masses T A B L E 1 The number of correct and incorrect mass
localizations by lobe by individual radiologists.

The mean measured size of the masses on histopathology Radiologist 1 2 3 4

was available for 34/75 masses and was 9.11 cm. The Correct (n) 40 62 57 58
mean measured diameter on CT was 8.95 cm. At the time
Incorrect (n) 33 11 16 15
of surgery, 48/75 (64%) of the masses were in the left divi-
Total (n) 73 73 73 73
sion, 12/75 (16%) were in the central division, and 15/75
 1532950x, 2024, 7, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vsu.14142 by Mississippi State University, Wiley Online Library on [03/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
THOMSEN ET AL. 1317

TABLE 2 The number of correct and incorrect mass localizations by liver lobe.

Left Left Right Right Caudate Caudate

Lobe lateral medial Quadrate medial lateral (papillary) (caudate) Overall
Correct (n) 95 47 11 20 17 8 19 217
Incorrect (n) 11 17 16 6 18 2 5 75
Total (n) 106 64 27 26 35 10 24 292

TABLE 3 The number of correct and incorrect mass localizations by liver division by individual radiologists.

Radiologist 1 2 3 4

Division Left Cent Right Left Cent Right Left Cent Right Left Cent Right
Correct (n) 42 9 11 46 9 14 43 9 14 43 10 14
Incorrect (n) 6 3 4 2 3 1 5 3 1 5 2 1
Total (n) 48 12 15 48 12 15 48 12 15 48 12 15

T A B L E 4 The number of correct and incorrect mass Analysis of histopathologic diagnosis and division as
localizations by liver division. well as histopathologic diagnosis and lobe was
Division Left Central Right Overall attempted but could not be satisfactorily completed due
to small case numbers for some cross-tabulation cells.
Correct (n) 173 37 53 264
The single case with a collision tumor is represented
Incorrect (n) 4 16 17 36 twice. The distribution of histopathologic diagnosis by
Total (n) 177 53 70 300 division and lobe is in Tables 7 and 8, respectively. In
the analysis attempted with the limited data, no signifi-
cant association between localization for lobe or divi-
lateral lobe (p < .05). Accuracy for the right lateral lobe sion and histologic diagnosis was found. Similarly, the
was lower than for the left lateral lobe ( p < .05). Overall, correlation between histopathologic diagnosis and accu-
the radiologists made 300 (75 eligible masses  4 radiolo- racy of localization by division and by lobe were ana-
gists) localizations by division. Of these localizations, lyzed but could not be satisfactorily completed due to
264/300 (88%, 95% CI: 84–91.4) were correct. A summary small case numbers. No significant association between
of individual radiologist accuracy in each division is in accuracy and diagnosis for any radiologist was found,
Table 3. A summary of the overall accuracy for each divi- however, the small case numbers mean this result
sion is in Table 4. No difference ( p > .05) in accuracy should be interpreted with caution. A summary of this
overall by division or with any radiologist individually data is in Table 9. The radiologists made 295 assessments
was found. of what circulation (hepatic vasculature) was most use-
All four radiologists agreed on the lobation in 41/73 ful in localizing the mass. For 103/295 (34.9%) the portal
(56%, 95% CI: 44.8–67.2) masses. Examples of CT images venous phase was reported to be most useful, for 79/295
from each lobe can be seen in Figures 1–7. All pairs of (26.8%) the hepatic venous phase was reported to be
radiologist comparisons had at least moderate agreement, most useful, and for 113/295 (38.3%) both were reported
with good agreement for radiologist 2 versus radiologist to be equally useful.
3, radiologist 2 versus radiologist 4, and radiologist 3 ver- No association between weight and histologic diagno-
sus radiologist 4. Individual radiologist agreement data sis (p = .545) or between age (<10 vs. >10 years old) and
on lobation is in Table 5. All four radiologists agreed on histologic diagnosis (p = .337) was found using the
the division in 61/75 (81%, 95% CI: 72.1–89.9) masses. All Kruskal-Wallis test of median weights for each category.
pairs of radiologists had at least good agreement, with an Histologic margins were available for 55 masses. No asso-
excellent agreement for radiologist 2 versus radiologist ciation between liver division and whether complete
3, radiologist 2 versus radiologist 4, and radiologist 3 ver- margins were obtained ( p = .376), between the liver lobe
sus radiologist 4. Individual radiologist agreement data is and whether complete margins were obtained (p = .055),
in Table 6. All radiologist pairs had strong correlation or between histologic diagnosis and whether complete
(p > .838) in size measurements and correlations were margins were obtained ( p = .547) was found, using the
highly significant (p < .001). Pearson χ 2 test.
 1532950x, 2024, 7, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vsu.14142 by Mississippi State University, Wiley Online Library on [03/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1318 THOMSEN ET AL.

F I G U R E 1 Computed tomography
(CT) images of a well-defined, mildly
heterogeneously contrast enhancing left lateral
liver lobe mass. (A) Transverse plane, arterial
phase post-contrast, soft tissue window CT
image of the mass in the left medial liver lobe.
(B) Dorsal plane, portal phase post-contrast, soft
tissue window CT image of the mass in the left
medial liver lobe. (C) Sagittal plane, portal phase
post-contrast, soft tissue window CT image the
mass in the left medial liver lobe.

F I G U R E 2 Computed tomography
(CT) images of a well-defined, mildly
heterogeneously hypoenhancing mass in
the left medial liver lobe. (A) Transverse
plane, arterial phase post-contrast, soft
tissue window CT image of the mass in
the left medial liver lobe. (B) Dorsal
plane, portal phase post-contrast, soft
tissue window CT image of the mass in
the left medial liver lobe. (C) Sagittal
plane, portal phase post-contrast, soft
tissue window CT image of the mass in
the left medial liver lobe.

4 | DISCUSSION 74.3%). Similarly, radiologists tended to agree more for

divisional localization (good-excellent) than lobe local-
This study supports CT as a useful modality for canine ization (moderate to good). All radiologist comparisons
liver mass localization. CT localization was more accu- had a strong correlation for size measurements. No sig-
rate for division (264/300, 88%) than lobe (217/292, nificant association between weight or age and
 1532950x, 2024, 7, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vsu.14142 by Mississippi State University, Wiley Online Library on [03/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
THOMSEN ET AL. 1319

F I G U R E 3 Computed tomography
(CT) images of a well-defined, rim
enhancing mass in the quadrate liver
lobe. (A) Transverse plane, portal phase
post-contrast, soft tissue window CT
image of the mass in the quadrate liver
lobe. (B) Dorsal plane, portal phase post-
contrast, soft tissue window CT image of
the mass in the quadrate liver lobe that
is positioned caudal and slightly to the
left of the gallbladder in the caudal
aspect of the liver. The mass has
excellent delineation between liver lobes
due to fat invagination between lobes.
(C) Sagittal plane, portal phase post-
contrast, soft tissue window CT image of
the mass in the quadrate liver lobe.

F I G U R E 4 Computed tomography
(CT) images of a well-defined,
heterogeneously contrast enhancing
mass in the right medial liver lobe.
(A) Transverse plane, arterial phase
post-contrast, soft tissue window CT
image of the mass in the right medial
liver lobe. (B) Dorsal plane, arterial
phase post-contrast, soft tissue window
CT image of mass in the right medial
liver lobe that is positioned to the right
of the gallbladder and the hepatic hilus
in the cranial aspect of the liver.
(C) Sagittal plane, arterial phase post-
contrast, soft tissue window CT image of
the mass in the right medial liver lobe.

histologic diagnosis was found. No significant associa- was found. For 103/295 (34.9%) responses, the portal
tion between histologic diagnosis, liver division or venous phase was reported to be the most useful, for
lobe, and whether complete margins were obtained 79/295 (26.8%) responses, the hepatic venous phase
 1532950x, 2024, 7, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vsu.14142 by Mississippi State University, Wiley Online Library on [03/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1320 THOMSEN ET AL.

F I G U R E 5 Computed tomography
(CT) images of a large, heterogeneous,
and heterogeneously contrast enhancing
mass in the right lateral liver lobe.
(A) Transverse plane, portal phase post-
contrast, soft tissue window CT image of
the mass in the right lateral liver lobe.
(B) Dorsal plane, portal phase post-
contrast, soft tissue window CT image of
the mass in the right lateral liver lobe.
The common right divisional portal
branch is visible medial to the mass.
(C) Sagittal plane, portal phase post-
contrast, soft tissue window CT image of
the mass in the right lateral liver lobe.
The common right divisional portal
branch is visible dorsal.

F I G U R E 6 Computed tomography
(CT) images of a poorly delineated,
heterogeneously contrast enhancing
mass in the papillary process of the
caudate lobe of the liver. (A) Transverse
plane, portal phase post-contrast, soft
tissue window CT image of the mass in
the papillary process of the caudate lobe
of the liver. (B) Dorsal plane, portal
phase post-contrast, soft tissue window
CT image of the mass in the papillary
process of the caudate lobe of the liver.
(C) Sagittal plane, portal phase post-
contrast, soft tissue window CT image of
the mass in the papillary process of the
caudate lobe of the liver.

was reported to be the most useful, and for 113/295 The nearly equal distribution of gender in this study
(38.3%) responses, both were reported to be equally is similar to the 57% male and 43% female previously
useful. reported in dogs with liver masses.3 Our subset of dogs
 1532950x, 2024, 7, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vsu.14142 by Mississippi State University, Wiley Online Library on [03/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
THOMSEN ET AL. 1321

F I G U R E 7 Computed tomography
(CT) images of a cavitated, noncontrast
enhancing mass in the caudate process
of the caudate lobe of the liver.
(A) Transverse plane, portal phase post-
contrast, soft tissue window CT image of
the mass in the caudate process of the
caudate lobe of the liver. Note the
caudate branch from the right hepatic
vein. (B) Dorsal plane, portal phase post-
contrast, soft tissue window CT image of
the mass in the caudate process of the
caudate lobe of the liver. Note the
caudate branch from the right hepatic
vein. (C) Sagittal plane, portal phase
post-contrast, soft tissue window CT
image of the mass in the caudate process
of the caudate lobe of the liver. Note the
caudate branch from the right
hepatic vein.

T A B L E 5 Radiologist agreement
Radiologist Asymptotic Approximate
for lobe localization by each comparison Kappa value standard errora Approximate Tb significance ( p)
radiologist pair.
1 versus 2 0.537 0.070 9.934 <.001
1 versus 3 0.578 0.066 11.341 <.001
1 versus 4 0.554 0.070 10.117 <.001
2 versus 3 0.785 0.055 14.029 <.001
2 versus 4 0.732 0.059 12.799 <.001
3 versus 4 0.751 0.058 13.572 <.001
a
Not assuming the null hypothesis.
b
Using the asymptotic standard error assuming the null hypothesis.

T A B L E 6 Radiologist agreement
Radiologist Kappa Asymptotic Approximate
for division localization by each comparison value standard errora Approximate Tb significance (p)
radiologist pair.
1 versus 2 0.739 0.069 8.844 <.001
1 versus 3 0.771 0.065 9.217 <.001
1 versus 4 0.748 0.068 8.888 <.001
2 versus 3 0.856 0.055 9.913 <.001
2 versus 4 0.857 0.055 10.039 <.001
3 versus 4 0.885 0.049 10.384 <.001
a
Not assuming the null hypothesis.
b
Using the asymptotic standard error assuming the null hypothesis.

was smaller with a mean weight of 13.8 kg compared to a mean age of 10.9 years and the tumor-specific age previ-
prior study which reported a mean weight of 20.6 kg.3 ously reported for dogs with liver masses.1,3 There were
The median age of the dogs in this study is similar to the numerous breeds represented in the study and these
 1532950x, 2024, 7, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vsu.14142 by Mississippi State University, Wiley Online Library on [03/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1322 THOMSEN ET AL.

TABLE 7 Histologic diagnosis broken down by division.

Nodular Other WD hepatocellular

Adenoma hyperplasia Noncancer cancer carcinoma No diagnosis Total
a
Division Left (n) 4 7 2 7 27 1 48
Central (n) 2 2 1 2 5 0 12
Right (n) 4 2 3 1 6 0 16
a
Total (n) 10 11 6 10 38 1 76
a
This dog died immediately postoperatively.

TABLE 8 Histologic diagnosis broken down by lobe.

Nodular Other WD hepatocellular

Adenoma hyperplasia Noncancer cancer carcinoma No diagnosis Total
a
Lobe Left lateral (n) 2 6 2 1 17 1 29
Left medial (n) 2 1 0 5 9 0 17
Quadrate (n) 1 0 0 0 2 0 3
Right medial (n) 1 2 1 2 3 0 9
Right lateral (n) 3 0 0 0 2 0 5
Caudate process (n) 0 2 1 1 4 0 8
Papillary process (n) 1 0 2 0 0 0 3
a
Total (n) 10 11 6 9 37 1 74
a
This dog died immediately postoperatively.

T A B L E 9 Correlation of histologic diagnosis and accuracy of reported.9,14 The lobation distribution is also similar to
localization by lobe by individual radiologists. the 33%–40% left lateral, 21%–25% left medial, 11%–16%
quadrate, 15%–19% right medial, 7% right lateral, and
Two-sided
6%–8% caudate (both processes) lobe previously
Pearson Degrees of asymptotic
Radiologist χ 2 value freedom significance ( p) reported.3,9 Hepatocellular carcinoma was the most com-
mon tumor type in our study, which is similar to previous
Lobe 1 3.386 5 .641
reports.3,9 Overall, the tumor population in this study
2 3.882 5 .567
was similar to that reported in previous studies and is
3 5.473 5 .361 likely a good representation of the general populous of
4 20.760 5 <.001 liver masses.
Division 1 2.734 2 .255 Previous reports of CT accuracy are limited in
2 5.707 2 .058 scope.8,11,14 The overall accuracy of liver division localiza-
tion in this study was 264/300 (88%). This is similar to
3 2.438 2 .295
the 84%–93% previously reported.8,11,14 Although no sig-
4 0.708 2 .702
nificant difference in accuracy overall by division was
demonstrated, the trend in accuracy of localization
was highest for the left division (173/192, 90.1%) com-
findings are consistent with a lack of breed predilection pared to masses located in the central division (37/48,
previously reported.17 The data seems to suggest a popu- 77.1%) and right division (53/60, 88.3%). These results are
lation of dogs similar to that of previous studies and is similar for the right and left divisions but not central
likely a good representation of the general populous. compared to Cheyney et al who reported a positive pre-
The mean measured size of the masses in this study is dictive value of 88% for the left division, 43% for the cen-
similar to the previously reported average of 11.3 cm.3 tral division, and 75% for the right division.11 These
The distribution of masses by division in the current divisional differences might be true on ultrasound as
study is similar to the 53%–58% left division, 12%–27% well.11,12 It would seem logical for the left division to
central division, and 14%–35% right division previously have the highest accuracy as it has the largest volume
 1532950x, 2024, 7, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vsu.14142 by Mississippi State University, Wiley Online Library on [03/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
THOMSEN ET AL. 1323

and the lobes of the left division are anatomically well accuracies previously published mainly depict the localiza-
separated and mobile.4 Lamb et al found that all masses tion accuracy of that one radiologist. In human medicine,
left of midline on CT were left divisional masses making manual segmentation of the liver by radiologists is time-
the landmarks extremely clear. In contrast, only 60% of consuming and prone to error.21 The use of artificial intel-
masses to the right of midline were right divisional.14 ligence in the form of convolutional neural networks
Based on the information in the present study it appears (CNNs) has demonstrated significant promise for auto-
that CT offers at least good reliability for localization to mated segmentation of the human liver.21–24 These same
all divisions, which is essential for surgical planning and principles of machine learning have already been applied
client preparation of potential surgical complications in the veterinary space to evaluate CT imaging characteris-
given the higher risk of complications with right divi- tics of hepatic masses as predictors of malignancy.25,26 It
sional masses.7,9 stands to reason that this technology could be applied to
Although liver division appears important to accu- canine liver mass localization. Future research should be
rately predict surgical risks, the effect of specific lobation considered for localization as this field continues to
has yet to be examined. Masses in a different lobe within advance.
the same division can require different surgical equip- It has been previously reported that CT features
ment and approaches or be considerably more challeng- including the postcontrast enhancement pattern of the
ing to remove. In this way, there is a logical advantage to lesion and the lesion's maximal transverse diameter have
localization by lobe over division. The present study has a high accuracy level for predicting malignancy.27 To the
a lower accuracy for identification of mass location by authors' knowledge no studies exist evaluating locations
lobe compared to division ( p < .05). In addition to the influence on malignancy. No apparent association
lower overall accuracy, the accuracy for the quadrate lobe between histopathologic diagnosis and location by divi-
was significantly lower than for the caudate process of sion or lobe was determined in the current study but the
the caudate lobe, the left medial lobe, and the left lateral case numbers were too small for a reliable analysis. In
lobe. Accuracy for the right lateral lobe was significantly addition, no association between histopathologic diagno-
lower than for the left lateral lobe. The particular lobar sis and accuracy of localization by division or lobe was
anatomy might contribute to these differences. Although found. Again, low case numbers prevented full analysis.
the present study found a much higher accuracy Future studies with larger case numbers are warranted to
(217/292, 74.3%) than previously reported (56%), there is evaluate these associations.
still a relatively high likelihood (25.7%) that a tumor The current study represents the most comprehensive
could be incorrectly localized.14 The lower accuracy by liver mass localization study to date but does have several
lobe versus division might be related to the smaller size limitations. The retrospective nature creates several possi-
of individual lobes, less distinct anatomic landmarks for ble errors. It relies on the information provided in medical
lobes, individual vascular variance, or distortion of anat- records and cannot be fully verified. Additionally, masses
omy by the masses.5,15,18 CT localization by lobe should that did not have surgery were excluded which may have
be interpreted with some level of caution. There is some introduced some selection bias towards smaller masses
evidence that techniques such as positive contrast CT that were evaluated as resectable. In this study, the radiol-
peritoneography could aid in more accurate localization, ogists only evaluated the liver and were aware of the goal
but this has only been demonstrated in concept.19 The of the study. This could have encouraged heightened
importance of lobation as it relates to surgical risks war- attention and artificially raised accuracy. A blinded pro-
rants further research. spective study could have eliminated some of these poten-
When comparing the individual accuracy of radiolo- tial biases but would have been considerably more
gists by lobe there are significant differences and individ- challenging. Several analyses in this study were under-
ual accuracy ranges from 40/73 (54.8%) to 62/73 (83.1%). powered due to small case numbers. A larger study would
The agreement between radiologists was good to excellent be required to have large enough numbers in all lobes and
for division but only moderate to good for lobation. This of all mass types. Regardless, this study demonstrates that
supports that lobe localization cannot be reliably evaluated CT is an effective and reliable tool for the localization of
on CT, whereas division can. Training and radiologist liver masses by division. Localization by lobe should be
experience might have a role to play and studies in human interpreted cautiously as the accuracy and inter-radiologist
medicine support the relationship between radiologist agreement are considerably less than for division.
experience and diagnostic accuracy with CT scan.20 The
most experienced radiologist in this study was the most AUTHOR CONTRIBUTIONS
accurate in lobation and division analysis. Prior veterinary Thomsen BJ, DVM: Contributed to the design of the
studies lack inter-radiologist comparisons thus, the study, performed a record search and identified suitable
 1532950x, 2024, 7, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vsu.14142 by Mississippi State University, Wiley Online Library on [03/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1324 THOMSEN ET AL.

medical records, performed data collection from medical 7. Liptak JM, Dernell WS, Monnet E, et al. Massive hepatocellular
records, compiled all data into spreadsheets, interpreted carcinoma in dogs: 48 cases (1992-2002). J Am Vet Med Assoc.
statistical results for relevance, and primarily drafted 2004;225(8):1225-1230.
8. Moyer J, Lopez DJ, Balkman CE, Sumner JP. Factors associ-
and revised the manuscript. Ward M, BVSc, MSc,
ated with survival in dogs with a histopathological diagnosis of
MPVM, PhD, FACVSc, DVS: Contributed to the design hepatocellular carcinoma: 94 cases (2007-2018). Open Vet J.
of the study, performed all statistical analysis, primarily 2021;11(1):144-153.
authored the materials and methods section related to 9. Moore VW, White J, Marchevsky AM. Association between
statistical analysis, critically reviewed the manuscript, divisional location and short-term outcome of liver mass
and endorsed the final version. Heo JY, DVM, MS, resection in 124 dogs. Vet Surg. 2023;52(4):513-520.
DACVR: Contributed to the design of the study and was 10. Armbrust LJ, Biller DS, Bamford A, Chun R, Garrett LD,
Sanderson MW. Comparison of three-view thoracic radiogra-
primarily responsible for determining all CT criteria to
phy and computed tomography for detection of pulmonary
be evaluated, evaluated CT scans, critically reviewed the
nodules in dogs with neoplasia. J Am Vet Med Assoc. 2012;
manuscript, and endorsed the final version. Huynh E, 240(9):1088-1094.
DVM, MS, DACVR: Contributed to the design of the 11. Cheney DM, Coleman MC, Voges AK, Thieman Mankin KM,
study, evaluated CT, primarily authored the materials Griffin JF. Ultrasonographic and CT accuracy in localising
and methods section related to imaging, critically surgical-or necropsy-confirmed solitary hepatic masses in dogs.
reviewed the manuscript, and endorsed the final ver- J Small Anim Pract. 2019;60(5):274-279.
sion. Ledesma MA, DVM, MS, DACVR: Contributed to 12. Wormser C, Reetz JA, Giuffrida MA. Diagnostic accuracy of
ultrasound to predict the location of solitary hepatic masses in
the design of the study, evaluated CT scans, critically
dogs. Vet Surg. 2016;45(2):208-213.
reviewed the manuscript, and endorsed the final ver- 13. Nelson RC, Chezmar JL, Sugarbaker PH, Murray DR,
sion. Fuerst JA, DVM, MS, DACVR: Contributed to the Bernardino ME. Preoperative localization of focal liver lesions
design of the study, evaluated CT scans, critically to specific liver segments: utility of CT during arterial portogra-
reviewed the manuscript, and endorsed the final ver- phy. Radiology. 1990;176(1):89-94.
sion. Vinayak A, DVM, DECVS (Small Animal), ACVS 14. Lamb CR, Steel R, Lipscomb VJ. Determining the anatomical
Fellow, Surgical Oncology: Assisted in medical record origin of canine hepatic masses by CT. J Small Anim Pract.
search, performed data collection from medical records, 2018;59(12):752-757.
15. Carlisle CH, Wu J-X, Heath TJ. Anatomy of the portal and
interpreted statistical results for relevance, and was the
hepatic veins of the dog: a basis for systematic evaluation of
senior collaborator who provided in-line scientific edit- the liver by ultrasonography. Vet Radiol Ultrasound. 1995;
ing. All authors are aware of their respective contribu- 36(3):227-233.
tions and have confidence in the integrity of all 16. Toma B. Dictionary of Veterinary Epidemiology. 1st English
contributions. language ed. Iowa State University Press; 1999 xvi, 284.
17. Hammer AS, Sikkema DA. Hepatic neoplasia in the dog
CONFLICT OF INTEREST STATEMENT and cat. Vet Clin North Am Small Anim Pract. 1995;25(2):
419-435.
The authors declare no conflict of interest.
18. Oishi Y, Tani K, Nakazawa H, Itamoto K, Haraguchi T,
Taura Y. Anatomical evaluation of hepatic vascular system in
R EF E RE N C E S healthy beagles using X-ray contrast computed tomography.
1. Patnaik AK, Hurvitz AI, Lieberman PH. Canine hepatic neo- J Vet Med Sci. 2015;77(8):925-929.
plasms: a clinicopathologic study. Vet Pathol. 1980;17(5): 19. Rivas LR, Seiler GS, Robertson ID, Ryu H, Mathews KG. Abil-
553-564. ity of positive and negative contrast computed tomographic
2. Vilkovyskiy IF, Vatnikov YA, Kulikov EV, et al. Influence of peritoneography to delineate canine liver lobe fissures. Vet
hepatic neoplasia on life expectancy in dogs. Vet World. 2020; Surg. 2019;48(8):1444-1449.
13(3):413-418. 20. Tsurusaki M, Numoto I, Oda T, et al. Assessment of liver
3. Kinsey JR, Gilson SD, Hauptman J, Mehler SJ, May LR. Factors metastases using CT and MRI scans in patients with pancreatic
associated with long-term survival in dogs undergoing liver ductal adenocarcinoma: effects of observer experience on diag-
lobectomy as treatment for liver tumors. Can Vet J. 2015;56(6): nostic accuracy. Cancers (Basel). 2020;12(6):1455.
598-604. 21. Ibragimov B, Toesca D, Chang D, Koong A, Xing L. Combining
4. Johnston SA, Tobias KM, Peck JN, Kent M. Veterinary surgery: deep learning with anatomical analysis for segmentation of the
small animal. Vol 2. 2nd ed. Elsevier; 2018:86 Illustrations portal vein for liver SBRT planning. Phys Med Biol. 2017;
(chiefly colour). 62(23):8943-8958.
5. Mari L, Acocella F. Vascular anatomy of canine hepatic venous 22. Zhou LQ, Wang JY, Yu SY, et al. Artificial intelligence in medi-
system: a basis for liver surgery. Anat Histol Embryol. 2015; cal imaging of the liver. World J Gastroenterol. 2019;25(6):
44(3):212-224. 672-682.
6. Covey JL, Degner DA, Jackson AH, Hofeling AD, 23. Lu F, Wu F, Hu P, Peng Z, Kong D. Automatic 3D liver loca-
Walshaw R. Hilar liver resection in dogs. Vet Surg. 2009; tion and segmentation via convolutional neural network and
38(1):104-111.
 1532950x, 2024, 7, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/vsu.14142 by Mississippi State University, Wiley Online Library on [03/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
THOMSEN ET AL. 1325

graph cut. Int J Comput Assist Radiol Surg. 2017;12(2): 27. Leela-Arporn R, Ohta H, Shimbo G, et al. Computed tomo-
171-182. graphic features for differentiating benign from malignant liver
24. Sun C, Guo S, Zhang H, et al. Automatic segmentation of liver lesions in dogs. J Vet Med Sci. 2019;81(12):1697-1704.
tumors from multiphase contrast-enhanced CT images based
on FCNs. Artif Intell Med. 2017;83:58-66.
25. Shaker R, Wilke C, Ober C, Lawrence J. Machine learning How to cite this article: Thomsen BJ, Ward M,
model development for quantitative analysis of CT heterogene- Heo JY, et al. Computed tomography scan
ity in canine hepatic masses may predict histologic malignancy.
accuracy for the prediction of lobe and division of
Vet Radiol Ultrasound. 2021;62(6):711-719.
liver tumors by four board-certified radiologists.
26. Burti S, Zotti A, Bonsembiante F, Contiero B, Banzato T. Diag-
nostic accuracy of delayed phase post contrast computed tomo- Veterinary Surgery. 2024;53(7):1313‐1325. doi:10.
graphic images in the diagnosis of focal liver lesions in dogs: 1111/vsu.14142
69 cases. Front Vet Sci. 2021;8:611556.