Evaluation of canine hepatic masses by use of triphasic computed tomography and B-mode, color flow, power, and pulsed-wave Doppler ultrasonography and correlation with histopathologic classification

Erin R. Griebie Department of Veterinary Clinical Sciences, College of Veterinary Medicine, University of Minnesota, Saint Paul, MN 55108.

Search for other papers by Erin R. Griebie in
Current site
Google Scholar
PubMed
Close
 DVM
,
Frederic H. David Department of Veterinary Clinical Sciences, College of Veterinary Medicine, University of Minnesota, Saint Paul, MN 55108.

Search for other papers by Frederic H. David in
Current site
Google Scholar
PubMed
Close
 DEDV
,
Christopher P. Ober Department of Veterinary Clinical Sciences, College of Veterinary Medicine, University of Minnesota, Saint Paul, MN 55108.

Search for other papers by Christopher P. Ober in
Current site
Google Scholar
PubMed
Close
 DVM, Phd
,
Daniel A. Feeney Department of Veterinary Clinical Sciences, College of Veterinary Medicine, University of Minnesota, Saint Paul, MN 55108.

Search for other papers by Daniel A. Feeney in
Current site
Google Scholar
PubMed
Close
 DVM, MS
,
Kari L. Anderson Department of Veterinary Clinical Sciences, College of Veterinary Medicine, University of Minnesota, Saint Paul, MN 55108.

Search for other papers by Kari L. Anderson in
Current site
Google Scholar
PubMed
Close
 DVM
,
Arno Wuenschmann Department of Veterinary Clinical Sciences, College of Veterinary Medicine, University of Minnesota, Saint Paul, MN 55108.

Search for other papers by Arno Wuenschmann in
Current site
Google Scholar
PubMed
Close
 Dr med vet
, and
Carl R. Jessen Department of Veterinary Clinical Sciences, College of Veterinary Medicine, University of Minnesota, Saint Paul, MN 55108.

Search for other papers by Carl R. Jessen in
Current site
Google Scholar
PubMed
Close
 DVM, Phd

Click on author name to view affiliation information

Abstract

OBJECTIVE To determine clinical relevance for quantitative and qualitative features of canine hepatic masses evaluated by use of triphasic CT and B-mode, color flow, power, and pulsed-wave Doppler ultrasonography and to compare diagnostic accuracy of these modalities for predicting mass type on the basis of histopathologic classification.

ANIMALS 44 client-owned dogs.

PROCEDURES Dogs with histopathologic confirmation (needle core, punch, or excisional biopsy) of a hepatic mass were enrolled. Triphasic CT and B-mode, color flow, power, and pulsed-wave Doppler ultrasonography of each hepatic mass were performed. Seventy quantitative and qualitative variables of each hepatic mass were recorded by 5 separate observers and statistically evaluated with discriminant and stepwise analyses. Significant variables were entered in equation-based predictions for the histopathologic diagnosis.

RESULTS An equation that included the lowest delayed-phase absolute enhancement of the mass and the highest venous-phase mass conspicuity was used to correctly classify 43 of 46 (93.5%) hepatic masses as benign or malignant. An equation that included only the lowest delayed-phase absolute enhancement of the mass could be used to correctly classify 42 of 46 (91.3%) masses (with expectation of malignancy if this value was < 37 Hounsfield units). For ultrasonography, categorization of the masses with cavitations as malignant achieved a diagnostic accuracy of 80.4%.

CONCLUSIONS AND CLINICAL RELEVANCE Triphasic CT had a higher accuracy than ultrasonography for use in predicting hepatic lesion classification. The lowest delayed-phase absolute enhancement of the mass was a simple calculation that required 2 measurements and aided in the differentiation of benign versus malignant hepatic masses.

Abstract

OBJECTIVE To determine clinical relevance for quantitative and qualitative features of canine hepatic masses evaluated by use of triphasic CT and B-mode, color flow, power, and pulsed-wave Doppler ultrasonography and to compare diagnostic accuracy of these modalities for predicting mass type on the basis of histopathologic classification.

ANIMALS 44 client-owned dogs.

PROCEDURES Dogs with histopathologic confirmation (needle core, punch, or excisional biopsy) of a hepatic mass were enrolled. Triphasic CT and B-mode, color flow, power, and pulsed-wave Doppler ultrasonography of each hepatic mass were performed. Seventy quantitative and qualitative variables of each hepatic mass were recorded by 5 separate observers and statistically evaluated with discriminant and stepwise analyses. Significant variables were entered in equation-based predictions for the histopathologic diagnosis.

RESULTS An equation that included the lowest delayed-phase absolute enhancement of the mass and the highest venous-phase mass conspicuity was used to correctly classify 43 of 46 (93.5%) hepatic masses as benign or malignant. An equation that included only the lowest delayed-phase absolute enhancement of the mass could be used to correctly classify 42 of 46 (91.3%) masses (with expectation of malignancy if this value was < 37 Hounsfield units). For ultrasonography, categorization of the masses with cavitations as malignant achieved a diagnostic accuracy of 80.4%.

CONCLUSIONS AND CLINICAL RELEVANCE Triphasic CT had a higher accuracy than ultrasonography for use in predicting hepatic lesion classification. The lowest delayed-phase absolute enhancement of the mass was a simple calculation that required 2 measurements and aided in the differentiation of benign versus malignant hepatic masses.

Technical advancements of CT now allow scanning to be performed more quickly, which often reduces the need for animals to be anesthetized and enables image acquisition for animals that only have been sedated.1 Use of multidetector row CT also allows triphasic CT angiography of the abdomen to become part of routine protocols.2–4 However, CT may not be accessible or available; thus, ultrasonography has been a mainstay for evaluating animals with abdominal disease and is often more available in smaller veterinary clinics and hospitals. The increased number of animals undergoing abdominal CT and ultrasonography has resulted in more frequent discovery of hepatic masses. In 1 study,5 70% of dogs had benign nodular hyperplasia of the liver, which was grossly visible as hepatic nodules and masses during necropsy, with an increasing incidence as the dogs aged. These benign nodules and masses are most often incidental and do not require specific treatment, whereas neoplastic masses may require removal or administration of chemotherapeutics (or both). Determining the nature and importance of hepatic masses remains a diagnostic challenge for veterinary radiologists and clinicians.

In human medicine, the dynamic CT characteristics of hepatocellular carcinomas are so distinct that diagnosis has been made solely on the basis of imaging findings and tumor markers.6 In dogs, triphasic helical CT is a useful tool for differentiating hepatic carcinomas from nodular hyperplasia and metastasis7 as well as for characterizing the enhancement characteristics of hepatocellular carcinoma, hepatocellular adenoma, and nodular hyperplasia.8 However, the diagnostic value of each imaging characteristic has not been tested statistically.7

Despite advances with triphasic CT, the cost of CT and the need for sedation or anesthesia of veterinary patients remain a drawback when compared with more widely available and less expensive ultrasonographic techniques. The B-mode characteristics of hepatic lesions have been widely evaluated but have limited accuracy for use in diagnosis, which does not allow clinicians to make a definitive diagnosis without the collection and evaluation of hepatic samples.9,10 In dogs, use of contrast harmonic ultrasonography improves the diagnostic capability for differentiating benign and malignant hepatic lesions.11 However, the need for costly equipment (including specific transducers and contrast harmonic software), the high cost of contrast medium, and variations in local regulations limit the use of this diagnostic technique.11

Color flow and power Doppler ultrasonography are commonly used in veterinary medicine because they are available on most commercial ultrasound machines.12 In a study13 of human patients, color flow and power Doppler ultrasonography were used to evaluate pulsatile or continuous wave forms of the blood flow signal in 42 hepatocellular carcinoma nodules (< 2 cm in diameter). Power Doppler ultrasonography achieved a higher or almost equal tumor vascular detection rate, compared with the detection rate for dynamic CT or CT hepatic arteriography, respectively.13 Color flow and power Doppler ultrasonography have been used to investigate the difference between malignant splenic masses and benign lesions in 31 dogs.12 The authors of that study12 concluded that the identification of aberrant and tortuous vessels within a mass could suggest malignancy; however, further data are needed to support this finding.

Thus, triphasic CT and color flow and power Doppler ultrasonography have been used in the evaluation of hepatic lesions. However, to the authors’ knowledge, no veterinary study has been conducted to compare these modalities for the evaluation of hepatic lesions.

Therefore, the first objective for the study reported here was to determine the clinical relevance of qualitative and quantitative tomographic variables for their use in classifying hepatic lesions, with a goal of separating benign from malignant lesions, and to characterize the hierarchy of the value for each variable or combination of variables in establishing an accurate diagnosis for hepatic masses seen on CT. The second objective was to determine the clinical relevance of qualitative and quantitative B-mode and Doppler ultrasonographic variables for use in classifying hepatic lesions, with a goal of separating benign from malignant lesions. The final objective was to compare the diagnostic yield of triphasic CT and B-mode and Doppler ultrasonography of the same hepatic masses and to determine the modality that was most useful for classifying hepatic masses as malignant or benign.

Materials and Methods

Animals

Client-owned dogs admitted to the Veterinary Medical Center at the University of Minnesota between May 2014 and March 2016 because of various clinical signs and that underwent a triphasic abdominal CT scan were recruited for inclusion in the study. Dogs with a hepatic mass identified on CT (defined as a focal hepatic lesion with a diameter ≥ 2 cm) were prospectively enrolled. The inclusion of masses with a diameter of at least 2 cm was selected to increase the likelihood of appropriate sample collection by use of needle core biopsy. Client consent was obtained for inclusion of all dogs. The study was approved by the University of Minnesota Institutional Animal Care and Use Committee.

CT procedures

All dogs were sedated or anesthetized for CT; sedation or anesthesia was performed under the direct supervision of members of the anesthesia department at our institution and as directed by each dog's primary clinician. In addition, dogs were anesthetized or sedated prior to enrollment in the study, and the ultrasonographic examination and biopsy were performed during that same anesthetic episode.

All CT images were obtained with a 64-slice scanner.a The protocol for CT consisted of precontrast images and postcontrast images in the arterial, venous, and delayed phases, as was routinely performed for all abdominal CT scans at our institution. Dogs were positioned in dorsal or sternal recumbency. Acquisition parameters included rotation time of 0.5 to 0.75 seconds, helical scan with a pitch factor of 0.828 to 0.844, 120 kVp, and 50 to 400 mAs automatically calculated by a software package,b slice thickness of 1.0 mm, and slice interval of 1.0 mm. Image reconstruction (in the transverse, sagittal, and dorsal planes) used a slice thickness of 2 mm and slice interval of 2 mm. Contrast mediumc (770 mg of I/kg, IV) was administered by use of a power injector with a standard pressure of 2.2 × 106 Pa at a rate determined on the basis of each dog's body weight (range, 1.0 to 3.5 mL/s). Bolus tracking of an aortic ROI (placed over the aorta at the level of the diaphragm) with a threshold at 180 HU was used to trigger the arterial scan; venous and delayed images were acquired at 20 and 90 seconds after the start of the arterial scan, respectively.

Ultrasonography procedures

All ultrasonographic images were obtained with a microconvex (frequency range, 8 to 11 MHz) transducer,d coupled to an ultrasound machine capable of image fusion. It was initially believed that image fusion would be needed to help identify the masses with low ultrasonographic conspicuity, but image fusion was not necessary because all of the masses visible with CT were identified ultrasonographically. Standard ultrasonographic images of the hepatic masses were obtained by a resident in the third year of a veterinary radiology training program (ERG) or a board-certified veterinary radiologist (FHD). The B-mode images of each hepatic mass and surrounding parenchyma were acquired in transverse and sagittal planes. Optimized color flow and power Doppler images of each hepatic mass were also obtained in transverse and sagittal planes. If adequate arterial blood flow was visible within the mass, as evident by visual identification of a high to intermediate resistance flow pattern, Doppler spectral analysis was performed by use of pulsed-wave Doppler ultrasonography. The highest quality pulsatility waveform was chosen for determination of resistive and pulsatility indices, which were calculated by use of ultrasound software. For some masses, respiratory motion prevented acquisition of good-quality pulsatility waveforms, despite adequate intralesional blood flow.

Some dogs did not undergo surgery (punch biopsy or excision of the hepatic mass). In those dogs, ultrasound-guided biopsy specimens were obtained with a core needle biopsy instrument, either automatic (18 gauge × 15 cm × 11 mm or 18 gauge × 16 cm × 22 mm)e or semiautomatic (18 gauge × 15 cm × 15 mm).f The utensil chosen for collection of biopsy specimens was determined on the basis of the size of the dog and location of the hepatic mass. When a biopsy specimen could not be safely obtained because of the location of the mass, that dog was excluded from the study.

Histologic examination

Specimens for histologic examination were obtained by use of ultrasound-guided biopsy immediately after the ultrasonographic examination, by resection of the mass or punch biopsy during exploratory surgery, or during necropsy. For some dogs, a core needle biopsy specimen was obtained during the initial examination and a specimen was also obtained during subsequent surgery; for these dogs, the specimen acquired during surgery was used to make the diagnosis for the purposes of this study. Histologic examinations were performed by a board-certified veterinary pathologist (AW).

Image analysis

Four board-certified veterinary radiologists (FHD, CPO, DAF, and KLA) and a resident in the third year of a veterinary radiology training program (ERG) evaluated all of the images by use of image analysis software.g All of the observers were unaware of results of the histologic examinations. A standardized collaborative tracking algorithm (low-pass filter with beam-hardening correction) with a window width of 500 HU and window level of 70 HU was used for interpretation of all CT images.

For CT images, the number of hepatic masses or nodules within each liver was recorded. Qualitative tomographic evaluation included the degree of heterogeneity (scale of 1 to 5, with 1 indicating homogenous and 5 indicating highly heterogenous), margination (scale of 1 to 5, with 1 indicating poorly defined and 5 indicating well defined), and presence or absence of mineralization, rim enhancement, or vascularity. The presence or absence of blood vessels was recorded for each image during the postcontrast phases and graded (scale of 0 to 5, with 0 indicating absent and 5 indicating marked). Blood vessel distribution (central, peripheral, or diffuse) within each hepatic mass as well as the vascular tortuosity (graded on a scale of 1 to 5, with 1 indicating none [ie, straight] and 5 indicating highly tortuous) were also recorded, when present. Categorical variables were numerically transformed (0 = absent and 1 = present) for statistical analysis.

Quantitative tomographic evaluation included the measurement of mass size as well as multiple attenuation values. Maximal diameter (in millimeters) of hepatic masses was measured with electronic calipers and recorded. Attenuation (in HU) of hepatic masses and surrounding normal hepatic parenchyma was measured by placing a circular ROI within the area of interest; attenuation was measured 3 times for each phase. An attempt was made to maintain a uniform size and location for the ROI across phases and to make the ROI as large as feasible while still keeping it within an area of relative homogeneity. When a hepatic mass was subjectively considered to be at least moderately heterogeneous (> 2 on a scale of 1 to 5), ROIs were obtained from the most hyperattenuating (3 ROIs) and the most hypoattenuating (3 ROIs) regions (subjectively identified visually) within the mass for each phase; these were recorded as the highest and lowest regions of attenuation. Mean attenuation values for each region in each phase were determined and used to calculate the absolute enhancement of the mass and liver for each phase (ie, postcontrast attenuation of the region at a given phase minus precontrast attenuation of the region). Maximum and minimum absolute enhancement of the mass was recorded as well as the phase of maximum and minimum absolute enhancement (arterial, venous or delayed) for each mass. Relative enhancement of the mass was calculated (ie, absolute enhancement of the mass at a given phase minus absolute enhancement of the normal liver at the same phase) for each postcontrast phase (arterial, venous, and delayed). Mass conspicuity was also calculated (ie, mass attenuation minus liver attenuation) for each precontrast and postcontrast phase. When a mass was subjectively considered heterogeneous, calculations were performed for both the highest and lowest regions of attenuation.

Qualitative ultrasonographic evaluation included margination (poorly defined, moderately defined, and well-defined), echogenicity (hypoechoic, hyperechoic, or isoechoic when compared with the surrounding liver parenchyma), and the degree of heterogeneity of the hepatic mass (scale of 1 to 5, with 1 indicating homogenous and 5 indicating highly heterogeneous). Presence or absence of cavitations within a mass was recorded. Presence of blood vessels within a mass detected by use of both color flow and power Doppler ultrasonography was recorded and graded (scale of 0 to 5, with 0 indicating absent and 5 indicating marked). Distribution of blood vessels (central, peripheral, or diffuse) as well as vascular tortuosity (graded on a scale of 1 to 5, with 1 indicating none [ie, straight] and 5 indicating highly tortuous) were recorded.

Quantitative ultrasonographic evaluation included size of each mass and Doppler ultrasonographic measurements, when available. Maximal diameter (in millimeters) of each hepatic mass was measured by use of electronic calipers and recorded. When a hepatic mass was vascular and a good-quality Doppler ultrasonographic spectral tracing was obtained, resistive and pulsatility indices were measured during image acquisition by use of ultrasound software. Measurements for both the resistive (3 measurements) and pulsatility (3 measurements) indices were obtained, when possible, and mean values were calculated. Because resistive and pulsatility indices must be acquired at the time of image acquisition, these measures were obtained by only a single observer.

Statistical analysis

Statistical analysis was performed by use of commercially available software.h Two classifications were used: a broad classification (benign or malignant) and a specific diagnosis classification (hepatocellular carcinoma, other carcinomas, hemangiosarcoma, spindle-cell sarcoma, corticosteroid hepatopathy nodular hyperplasia, hepatitis, vacuolar degeneration, biliary cystadenoma, fibrosis, and normal liver). Each CT and ultrasonographic criterion was first tested for a significant association with the histopathologically confirmed disease by use of univariate analysis. Sensitivity, specificity, and accuracy were calculated for criteria that retained significant associations with the diagnosis after multivariate analysis. Values were considered significant at P ≤ 0.05.

All statistical tests were selected and performed by 1 investigator (CRJ). Observations by each observer were compared (cross tabulated) for each criterion by use of the statistical software package to evaluate similarity of variable utilization. Both χ2 and McNemar analyses were performed. The 2 × 2 χ2 P value reported was from the Fisher exact test. Histopathologic yield values were evaluated with the independent χ2 test.

Stepwise discriminant analysis techniques were used for the criteria, with variable entrance into equations at P ≤ 0.05 and removal at P ≥ 0.10. All triphasic CT and ultrasonographic variables (with values of 0 or 1) were included in the initial step of the stepwise analysis. Fisher prediction equations were developed by use of the Wilk λ method. Fisher equations were used to evaluate the importance of the triphasic CT and ultrasonographic variables with respect to the histopathologic classification groups. Coefficients from these equations allowed comparisons within and among histopathologic classification groups for each observer, which provided insight into the importance of each variable with regard to the histopathologic classification.

Results

Animals

A total of 506 dogs underwent CT angiography of the abdomen at the Veterinary Medical Center during the study period (May 20, 2014, through March 15, 2016). There were 83 dogs that met the inclusion criteria, and 45 dogs were enrolled in the study. The 38 dogs not enrolled in the study were excluded for various reasons, including location of the mass and inaccessibility for biopsy, concurrent patient comorbidities, and owners who declined participation.

Of the 45 dogs, 1 was excluded from the study because of a nondiagnostic core needle biopsy specimen. A total of 46 hepatic masses from the remaining 44 dogs (26 neutered males and 18 spayed females) were included for data analysis. Dogs ranged from 4 to 13 years of age (mean, 10.6 years). Labrador Retrievers (9 [20%]) and Golden Retrievers (6 [14%]) were overrepresented in the study population. The remaining breeds included 2 Border Collies, 2 Fox Terriers, 2 German Shepherd Dogs, 2 German Short-haired Pointers, 2 Shih Tzus, 2 Welsh Corgis, and 1 each of American Eskimo Dog, American Staffordshire Terrier, Basset Hound, Bichon Frise, Briard, Brittany Spaniel, Cocker Spaniel, Collie, Doberman Pinscher, Jack Russell Terrier, Pomeranian, Pug, Rat Terrier, Shetland Sheepdog, and West Highland White Terrier; there also was a Golden Retriever–Poodle cross and a mixed-breed dog.

CT and ultrasonographic examinations and collection of biopsy specimens

Ultrasonography was performed immediately after CT, except for 2 dogs for which ultrasonography was performed immediately before abdominal surgery at 1 and 3 days after CT. Tissue specimens were obtained from 19 hepatic masses by use of ultrasound-guided biopsy; specimens were obtained from the remaining masses by punch biopsy during surgery (n = 3) or by surgical excision of the mass (23). One dog died during surgery, and tissue was collected during necropsy. Exploratory surgery was performed the day of or between 1 and 3 days after CT and ultrasonographic examinations, except for 4 dogs in which surgery was performed 17 to 36 days after the CT and ultrasonographic examinations.

A core needle biopsy specimen of a hepatic mass of 1 dog was obtained by the referring veterinarian. A resected piece of liver tissue was sent to the veterinary pathologist for evaluation and was therefore included in the data analysis.

Histopathologic classification

Histologic examination revealed 30 malignant masses and 16 benign masses. For the malignant masses, there were 23 hepatocellular carcinomas, 2 undifferentiated carcinomas, 2 metastatic hemangiosarcomas, 1 cholangiocarcinoma, 1 biliary carcinoma, and 1 spindle-cell sarcoma. Benign masses included hyperplastic nodules (n = 5; Figure 1), corticosteroid-induced hepatopathy (4), vacuolar degeneration (2), hepatitis (2), biliary cystadenoma (1), fibrosis (1), and normal liver (1). Results for examination of a core biopsy specimen were equivocal in 2 dogs that subsequently underwent surgery; on the basis of the histopathologic classification of liver tissue excised during surgery, a hepatocellular carcinoma and a biliary cystadenoma (Figure 2) were diagnosed. For 2 dogs for which the histopathologic classification was obtained by examination of core biopsy specimens, nodular hyperplasia was considered the most likely diagnosis. Although low-grade hepatocellular carcinoma could not be entirely excluded, both of these dogs were recorded as a classification of nodular hyperplasia for the study.

Figure 1—
Figure 1—

Triphasic CT images of a benign nodular hyperplastic mass in a dog obtained prior to administration of contrast medium (A) and during arterial (B), venous (C), and delayed (D) phases of image acquisition. Notice the mass (asterisk) in the precontrast image and the relatively homogeneous appearance with at least moderate enhancement of all areas of the mass in the delayed phase.

Citation: American Journal of Veterinary Research 78, 11; 10.2460/ajvr.78.11.1273

Figure 2—
Figure 2—

Triphasic CT images of a biliary cystadenoma in a dog obtained prior to administration of contrast medium (A) and during arterial (B), venous (C), and delayed (D) phases of image acquisition. Notice the biliary cystadenoma (asterisk) in the precontrast image. The mass is largely hypoattenuating and nonenhancing with a thin wall, which led to misclassification by use of the lowest delayed-phase absolute enhancement equation.

Citation: American Journal of Veterinary Research 78, 11; 10.2460/ajvr.78.11.1273

Image analysis

A total of 70 variables (57 CT and 13 ultrasonographic variables) were included in the initial univariate analysis. Of the 57 triphasic CT variables analyzed, 41 were continuous variables (interval or ratio data) and 16 were nonparametric (ordinal or categorical). Of the 13 ultrasonographic variables analyzed, 3 were continuous and 10 were nonparametric. Evaluations conducted after the univariate analysis revealed that 21 continuous CT variables, 8 nonparametric CT variables, 1 continuous ultrasonographic variable, and 3 nonparametric ultrasonographic variables were significant with respect to broad classification of disease (benign vs malignant; Table 1). No variables were significant with respect to specific classification of disease.

Table 1—

Mean values of CT and ultrasonographic variables found in univariate analysis to significantly differentiate between benign and malignant hepatic masses in 44 dogs.

VariableBenignMalignantP value*
Continuous CT variables
 Maximal diameter of mass (mm)47.894.9< 0.001
 Attenuation of normal liver on venous phase (HU)135.8119.00.026
 Attenuation of normal liver on delayed phase (HU)142.4123.0< 0.001
 Low attention of mass on precontrast image (HU)49.335.90.03
 Low attenuation of mass on arterial phase (HU)71.940.50.002
 Low attenuation of mass on venous phase (HU)106.947.3< 0.001
 High attenuation of mass on delayed phase (HU)133.2108.40.006
 Low attenuation of mass on delayed phase (HU)114.644.5< 0.001
 Absolute delayed enhancement of the liver (HU)72.857.00.002
 Low absolute arterial enhancement of mass (HU)22.64.60.008
 High absolute venous enhancement of mass (HU)83.458.60.035
 Low absolute venous enhancement of mass (HU)57.611.4< 0.001
 High absolute delayed enhancement of mass (HU)78.551.8< 0.001
 Low absolute delayed enhancement of mass (HU)65.28.6< 0.001
 Maximum absolute enhancement of mass (HU)91.862.80.009
 Minimum absolute enhancement of mass (HU)20.22.50.002
 Low relative arterial enhancement of mass (HU)8.2–8.20.041
 Low relative venous enhancement of mass (HU)–8.7–41.80.007
 Low relative delayed enhancement of mass (HU)–7.5–48.5< 0.001
 Low venous conspicuity of mass (HU)–28.9–71.70.004
 Low delayed conspicuity of mass (HU)–27.8–78.50.001
Nonparametric CT Variables
 Margination of mass2.03.80.003
 Preheterogeneity score1.43.0< 0.001
 Arterial heterogeneity score2.43.5< 0.001
 Venous heterogeneity score2.44.0< 0.001
 Delayed heterogeneity score1.83.8< 0.001
 Blood vessels present in the arterial phase§2.23.40.007
 Vascular tortuosity score1.92.70.017
 Phase of minimal absolute mass enhancementArterialArterial or venous0.009
Continuous ultrasonographic variables
 Maximal diameter of mass (mm)41.581.2< 0.001
Nonparametric ultrasonographic variables
 Margination of mass2.73.30.002
 Presence of cavitations within mass0.00.8< 0.001
 Vascular tortuosity score1.21.60.039

Values were considered significant at P ≤ 0.05.

Scale of 1 to 5, with 1 indicating poorly defined and 5 indicating well-defined.

Scale of 1 to 5, with 1 indicating homogenous and 5 indicating highly heterogenous.

Scale of 0 to 5, with 0 indicating absent and 5 indicating marked.

Scale of 1 to 5, with 1 indicating none (ie, straight) and 5 indicating highly tortuous.

Binary score, with 0 = absent and 1 = present.

Two calculated continuous CT variables were significant with respect to broad classification of disease after the stepwise multivariate analysis. Lowest delayed-phase absolute enhancement of the mass was calculated by measuring the CT value of the lowest region of attenuation within the mass on the delayed-phase images and subtracting the CT value of the lowest region of attenuation within the mass on the precontrast images (Figure 3). Mean ± SD value was 8.6 ± 16.7 HU for malignant masses and 65.2 ± 32.7 HU for benign masses. Highest venous-phase mass conspicuity was calculated by subtracting attenuation of normal liver on venous-phase images from the highest region of attenuation within the mass on the venous-phase images. Mean value was −3.8 ± 35.9 HU for malignant masses and 2.24 ± 34.9 HU for benign masses. These variables were entered into the Fisher equation process, whereby an equation was derived to predict mass type (benign or malignant) on the basis of these 2 variables. Accuracy was 93.5% (Table 2). Because of the inclusion of 2 variables in this model, a specific cutoff value could not be calculated. Use of the equation-based predictions resulted in correct classification of 29 malignant masses as malignant and incorrect classification of 1 malignant mass as benign and correct classification of 14 benign masses as benign and incorrect classification of 2 benign masses as malignant.

Figure 3—
Figure 3—

Triphasic CT images of a hepatocellular carcinoma in a dog that received a heterogeneity score of between 3 and 5 on all imaging phases by each of 5 observers. This mass is notably heterogenous, compared with the relatively homogenous mass in Figure 1. Images were obtained prior to administration of contrast (A) and during arterial (B), venous (C), and delayed (D) phases of image acquisition. Notice the ROI depicting the lowest attenuating region of the mass (white circle) and highest attenuating region of the mass (black circle) in each panel. The area of lowest attenuation on the precontrast image has minimal enhancement on the delayed phase, which is considered a marker of malignancy.

Citation: American Journal of Veterinary Research 78, 11; 10.2460/ajvr.78.11.1273

Table 2—

Prediction methods used to differentiate between malignant and benign hepatic masses in dogs.

Factor2-variable CT diagnosis1-variable CT diagnosisUltrasonography diagnosis
Variable or variables used in methodA = Lowest absolute delayed enhancement of the mass and B = highest venous phase conspicuity of the massA = Lowest absolute delayed enhancement of the massPresence or absence of cavitation
EquationMalignant = −0.835 + (0.027 × A) – (0.013 × B) Benign = −6.301 + (0.174 × A) – (0.064 × B)Malignant = (0.016 × A) − 0.76 Benign = (0.119 × A) – 4.571NA
Classification of prediction methodEquation with highest score predicts the type of mass< 37 HU predicts a malignant mass and > 37 HU predicts a benign massPresence of cavitation predicts a malignant mass and absence of cavitation predicts a benign mass
Sensitivity for predicting malignancy (%)96.793.376.6
Specificity for predicting malignancy (%)87.587.587.5
Accuracy (%)93.591.380.4

NA = Not applicable.

When analyzed separately, only the lowest delayed-phase absolute enhancement of the mass was significant. This variable was separately entered into the Fisher equation process, whereby an equation was derived to predict classification of a mass (Table 2). A threshold of 37 HU was calculated for the lowest delayed-phase absolute enhancement of the mass. Thus, by use of this equation, a lowest delayed-phase absolute enhancement of a mass of < 37 HU predicted malignancy, and a value > 37 HU predicted a benign lesion. This threshold value correctly classified 91.3% of the masses as benign or malignant. By use of this single variable, 28 malignant masses and 14 benign masses were correctly classified, whereas 2 malignant and 2 benign masses were incorrectly classified. Use of the threshold of 37 HU revealed that 23 of 230 observations of individual observers were misclassified. Twenty observations were for the same 4 masses (2 hepatocellular carcinomas, 1 biliary cystadenoma, and 1 corticosteroid-induced hepatopathy), which were misclassified by all 5 observers. The other 3 misclassified observations were for benign lesions, each of which was misclassified by only 1 observer.

Assessment of the presence or absence of cavitations within a mass by use of B-mode ultrasonography was the only ultrasonographic variable significant in the stepwise analysis (Figure 4). By use of this variable, masses with cavitations were classified as malignant, and masses without cavitations were classified as benign. This classification scheme correctly classified 80.4% of the masses as benign or malignant (23 malignant masses were correctly classified as malignant, 7 malignant masses were incorrectly classified as benign, 14 benign masses were correctly classified as benign, and 2 benign masses were incorrectly classified as malignant).

Figure 4—
Figure 4—

The B-mode ultrasonographic image of a hepatocellular carcinoma (long-axis view) with cavitations throughout the mass. Height (electronic calipers [plus signs] A to A, 6.3 cm) and length (electronic calipers B to B, 9.0 cm) are depicted.

Citation: American Journal of Veterinary Research 78, 11; 10.2460/ajvr.78.11.1273

Blood vessels were detected with color flow or power Doppler ultrasonography (or both) in 42 of 46 masses. There was no mass in which detection of blood flow by use of power Doppler ultrasonography was not also identified by use of color flow Doppler ultrasonography. Of the 42 masses with detectable blood flow, only 21 had arterial blood flow with an adequate waveform that allowed for calculation of resistive and pulsatility indices. Respiratory motion was a large limiting factor in achieving adequate waveforms because good-quality waveforms were not repeatedly obtained for 3 consecutive measurements in many masses that had detectable arterial blood flow. The presence or absence of vascularity (or any additional blood flow variable), including the resistive or pulsatility indices, was not significant for differentiating benign from malignant hepatic masses after stepwise analysis.

Discussion

The lowest region of absolute enhancement of the mass, as measured on the delayed phase of the triphasic CT scan, had an accuracy of 91.3% for predicting classification of a hepatic mass (benign or malignant). Sensitivity and specificity for use of this variable for the classification of a mass as malignant were 93.3% and 87.5%, respectively. An equation that combined this variable with the highest venous-phase mass conspicuity yielded an accuracy of 93.5%, sensitivity of 96.7%, and specificity of 87.5% for predicting the classification of a mass as malignant. The only ultrasonographic imaging characteristic that was significant during the stepwise analysis was the presence or absence of cavitations within a mass visible during B-mode ultrasonography. Use of this variable yielded an accuracy of 80.4%, sensitivity of 76.6%, and specificity of 87.5% for classifying masses as malignant. In the study reported here, CT or ultrasonographic variables could not be used to differentiate the specific cause of each liver mass, possibly in part because of the limited number of masses within each specific classification.

In the study reported here, the presence, absence, distribution, and subjective degree of tortuosity of blood vessels within any phase of the postcontrast CT images or Doppler ultrasonographic (power or color flow) images was not significant for differentiating benign from malignant hepatic masses after stepwise analysis. This differs from findings in human medicine, where the diagnosis of hepatocellular carcinoma is suggested by the presence of intratumoral pulsatile blood flow with high velocity (as determined by use of Doppler ultrasonography)14 or the presence of arterial tumor vascularity (as determined by use of dynamic CT).13 This difference in imaging characteristics may be attributable to differences in histologic and vascular characteristics of primary hepatic neoplasia between dogs and humans, because most human patients with hepatocellular carcinoma have cirrhosis,15 but to the authors’ knowledge, a correlation between primary hepatic neoplasia and the presence of cirrhosis in dogs has not been detected.

Measuring the lowest region of attenuation for each phase of contrast enhancement allowed identification of a significant result for the lowest absolute enhancement of the mass on the delayed phase in the multivariate analysis. By use of this variable alone, a cutoff of 37 HU could be used to accurately predict 91.3% of the masses as benign or malignant. In the multivariate analysis, the lowest degree of absolute enhancement of a mass was only significant on the delayed phase of contrast enhancement, which may have been attributable to the ability to differentiate hypoattenuating and hyperattenuating regions because of complete parenchymal enhancement of the mass by the time of acquisition of delayed-phase images. A measurement of < 37 HU was essentially a region with lack of or extremely limited uptake of contrast medium and was theorized to represent a region of necrosis. These findings are consistent with results for a study8 of CT features of hepatic tumors in dogs in which cyst-like necrotic lesions were seen frequently in hepatocellular carcinomas and less commonly identified in benign lesions, including lesions attributable to hepatic adenoma and nodular hyperplasia. The presence of necrosis within a tumor is strongly related to aggressive pathological characteristics and may be the result of tumor-related hypoxia16 as tumors outgrow their blood supply.

Presence or absence of cavitations within a mass (as determined by use of B-mode ultrasonography) had an accuracy of 80.4% for classification of hepatic masses as benign or malignant. This finding was not surprising because this ultrasonographic characteristic likely related to tumor necrosis, similar to the CT variable of the lowest delayed-phase absolute enhancement of the mass. Because this was a subjective criterion, there was likely more observer variation, which resulted in a lower accuracy for lesion classification than for the lowest delayed-phase absolute enhancement of a mass determined by use of CT.

Diagnostic accuracy for classification of masses as malignant or benign was higher for CT than for B -mode, color flow, or power Doppler ultrasonography. Several factors could explain this result. First, triphasic CT provides dynamic information on the evolution of contrast uptake over time. Performing separate CT scans of the liver after contrast medium administration in the arterial, venous, and delayed phases highlighted the unique portal blood system of the liver, thereby enhancing the ability to define lesions on the basis of their blood supply and uptake of contrast medium. Second, the ability to measure attenuation values within a defined region enabled more quantitative measurements and subsequent calculations or comparisons with CT than with ultrasonography, and such quantitative measures may have less interobserver variability, compared with interobserver variability for subjective evaluations. Respiratory motion inhibited acquisition of good-quality spectral Doppler waveforms for quantitative analysis for many of the masses with intratumoral blood flow; had respiratory motion been limited with the use of anesthetic breath hold techniques, the accuracy of Doppler ultrasonography for classification of benign and malignant masses may have improved. It is acknowledged that there was a degree of subjectivity because each observer drew an ROI for the highest and lowest region of attenuation within a given mass; however, interobserver agreement for the significant CT variables was good to excellent.

It should be mentioned that use of the threshold of 37 HU for the lowest delayed-phase absolute enhancement of the mass resulted in misclassification for 23 individual observer observations. Of these, 20 observations were for the same 4 masses (2 hepatocellular carcinomas, 1 biliary cystadenoma, and 1 corticosteroid-induced hepatopathy) that were misclassified by all 5 observers. Three benign lesions were misclassified incorrectly by only 1 observer each. Therefore, the 4 masses misclassified by all observers may have been a result of inherent characteristics of the masses rather than incorrect ROI placement. Retrospective review revealed that the 2 hepatocellular carcinomas had limited heterogeneity and were on the small (3 to 4 cm) side of the size spectrum. It is possible that nonenhancing areas or necrosis (or both) was not yet present. The biliary cystadenoma was likely misclassified as a malignant lesion because of the presence of fluid-filled cavities. This was the only mass with this cause in the study, but it is possible that such masses, which are known to have cystic components, may yield a typical false-positive result for neoplasia when the threshold of 37 HU is used for the lowest delayed-phase absolute enhancement of the mass (Figure 2). The reason for misclassification of the corticosteroid-induced hepatopathy was unclear.

Limitations of the present study included the small sample size. Of 83 dogs eligible for inclusion, only 44 were enrolled. Dogs not enrolled had a liver mass that could not be safely accessed for core biopsy, were not clinically stable for biopsy or surgery in the opinion of the attending clinician, or were not enrolled by the owners for personal reasons. The first 2 reasons for study exclusion emphasized the relevance and pertinence of the ultimate objective for this study (ie, to increase the ability to assess masses in the liver detected during diagnostic imaging for which it would not be feasible to obtain a biopsy specimen for histologic examination). A second limitation of the study was the small tissue sample size obtained by use of ultrasound-guided core needle biopsy. Although histologic examination is more advantageous than cytologic examination for diagnosis of focal liver lesions,17 the sample size obtained for histologic examination can affect the accuracy of diagnosis. Investigators of 1 study18 found that 16% of core needle biopsy specimens of the liver in clinically normal dogs were inadequate after processing. Additionally, depending on the site from which the sample was collected, accurate representation of the pathological process may not have been obtained. We attempted to address this inherent limitation of core biopsy by obtaining 3 core biopsy specimens from each liver mass whenever possible.

In the study reported here, accuracy for the prediction of benign versus malignant hepatic masses was higher for triphasic CT than for B-mode, color flow, and power Doppler ultrasonography. Although it should not be considered as a replacement for histologic examination, the measurement of 2 CT variables during triphasic CT that are then entered into a single-variable equation may increase the accuracy of the classification for benign and malignant hepatic masses and thus better direct the need for additional diagnostic testing and appropriate treatment.

Acknowledgments

Supported by a Small Companion Animal Grant from the College of Veterinary Medicine at the University of Minnesota.

Presented in abstract form at the 2016 American College of Veterinary Radiology Annual Scientific Conference, Orlando, Fla, October 2016.

The authors thank Dr. Tim O'Brien for assistance with the histologic examinations and Dr. Betty Kramek for assistance with recruiting patients for the study.

ABBREVIATIONS

HU

Hounsfield unit

ROI

Region of interest

Footnotes

a.

Toshiba Aquilion 64 CFX CT, Toshiba Medical Systems, Tustin, Calif.

b.

Sure Exposure 3D (SureIQ), Toshiba Medical Systems, Tustin, Calif.

c.

Optiray 350 (Ioversol), Mallinckrodt Inc, Hazelwood, Mo.

d.

Aplio 500, Toshiba, Tochigi, Japan.

e.

Bard Monopty disposable core biopsy instrument, Bard Peripheral Vascular Inc, Tempe, Ariz.

f.

E-Z core single-action biopsy needle, Products Group International, Lyons, Colo.

g.

Vue PACS, version 11.3, Carestream, Rochester, NY.

h.

SPSS Inc, version 18, IBM Corp, Armonk, NY.

References

  • 1. Fields EL, Robertson ID, Osborne JA, et al. Comparison of abdominal computed tomography and abdominal ultrasound in sedated dogs. Vet Radiol Ultrasound 2012; 53: 513517.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 2. Lee S, Jung J, Chang J, et al. Evaluation of triphasic helical computed tomography of the kidneys in clinically normal dogs. Am J Vet Res 2011; 72: 345349.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 3. Cáceres AV, Zwingenberger AL, Hardam E, et al. Helical computed tomographic angiography of the normal canine pancreas. Vet Radiol Ultrasound 2006; 47: 270278.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 4. Winter MD, Kinney LM, Kleine LJ. Three-dimensional helical computed tomographic angiography of the liver in five dogs. Vet Radiol Ultrasound 2005; 46: 494499.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 5. Bergman JR. Nodular hyperplasia in the liver of the dog: an association with changes in the Ito cell population. Vet Pathol 1985; 22: 427438.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 6. Koda M, Matsunage Y, Ueki M, et al. Qualitative assessment of tumor vascularity in hepatocellular carcinoma by contrast-enhanced coded ultrasound: comparison with arterial phase of dynamic CT and conventional color/power Doppler ultrasound. Eur Radiol 2004; 14: 11001108.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 7. Kutara K, Seki M, Ishikawa C, et al. Triple-phase helical computed tomography in dogs with hepatic masses. Vet Radiol Ultrasound 2014; 55: 715.

  • 8. Fukushima K, Kanemoto H, Ohno K, et al. CT characteristics of primary hepatic mass lesions in dogs. Vet Radiol Ultrasound 2012; 53: 252257.

    • Search Google Scholar
    • Export Citation
  • 9. Murakami T, Feeney DA, Bahr KL. Analysis of clinical and ultrasonographic data by use of logistic regression models for prediction of malignant versus benign causes of ultrasonographically detected focal liver lesions in dogs. Am J Vet Res 2012; 73: 821829.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 10. Warren-Smith CMR, Andrew S, Mantis P, et al. Lack of associations between ultrasonographic appearance of parenchymal lesions of the canine liver and histological diagnosis. J Small Anim Pract 2012; 53: 168173.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 11. O'Brien RT, Iani M, Matheson J, et al. Contrast harmonic ultrasound of spontaneous liver nodules in 32 dogs. Vet Radiol Ultrasound 2004; 45: 547553.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 12. Sharpley JL, Marolf AJ, Reichle JK, et al. Color and power Doppler ultrasonography for characterization of splenic masses in dogs. Vet Radiol Ultrasound 2012; 53: 586590.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 13. Furuse J, Maru Y, Yoshino M, et al. Assessment of arterial tumor vascularity in small hepatocellular carcinoma. Comparison between color Doppler and ultrasonography and radiographic imagings with contrast medium: dynamic CT, angiography, and CT hepatic arteriography. Eur J Radiol 2000; 36: 2027.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 14. Shimamoto K, Sakuma S, Ishigaki T, et al. Hepatocellular carcinoma: evaluation with color Doppler US and MR imaging. Radiology 1992; 182: 149153.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 15. Sanyal AJ, Yoon SK, Lencioni R. The etiology of hepatocellular carcinoma and consequences for treatment. Oncologist 2010; 15: 1422.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 16. Richards CH, Mohammed Z, Qayyum T, et al. The prognostic value of histological tumor necrosis in solid organ malignant disease: a systematic review. Future Oncol 2011; 7: 12231235.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 17. Bahr KL, Sharkey LC, Murakami T, et al. Accuracy of US-guided FNA of focal liver lesions in dogs: 140 cases (2005–2008). J Am Anim Hosp Assoc 2013; 49: 190196.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 18. Vasanjee SC, Bubenik LJ, Hosgood G, et al. Evaluation of hemorrhage, sample size, and collateral damage for five hepatic biopsy methods in dogs. Vet Surg 2006; 35: 8693.

    • Crossref
    • Search Google Scholar
    • Export Citation
All Time Past Year Past 30 Days
Abstract Views 311 0 0
Full Text Views 2032 1274 137
PDF Downloads 1512 781 159
Advertisement