Ultrasonographic evaluation is an integral part of the assessment of liver disease in dogs and cats.1-11 Diffuse (infiltrative but not nodular) liver disease is a common problem in dogs and cats.12-28 Despite this volume of information, there is no information on the accuracy of ultrasonographic interpretation for differentiation among the types of diffuse liver infiltrates (inflammatory, vacuolar hepatopathy, round-cell neoplasia, prenodular [early] metastatic disease, lipidosis, or other nonnodular liver disease). Furthermore, there is no information on the accuracy of ultrasonographic interpretation for differentiating diffuse liver disease from normal liver.
The group of diseases that can manifest as an infiltrative (but not architecturally disruptive) ultrasonographic pattern (eg, not masses, not nodules, without intrahepatic dilatation of bile duct, or without vascular disruption-distortion) include cholangiohepatitis (all forms), diffuse prenodular (early) metastatic carcinoma or sarcoma, round-cell neoplasia (including lymphomas, mast cell disease, and histiocytic neoplasms), patchy or diffuse fatty infiltrate, vacuolar hepatopathy, storage diseases (eg, amyloidosis or copper), toxic hepatopathy, early degenerative changes attributable to micronodular hyperplasia with various degrees of fibrosis (precirrhosis conditions), and vascular abnormalities.1-6,11-13,15-17,19,21,23-26 Currently, the role of grayscale ultrasonography in identifying hepatic parenchymal disease is limited to defining evidence of nodules or masses, guiding percutaneous collection of samples for use in cytologic or histologic examinations, and searching for related nonhepatic morphologic abnormalities that could create the biochemical or ultrasonographic abnormalities. These would include identifying portosystemic shunts, abdominal lymphadenopathy, or other masses and characterizing morphologic characteristics of the adrenal glands and pancreas.
The combination of liver ultrasonography and sonographically guided collection of samples for cytologic or histologic examination is part of an in-depth evaluation of liver disease.29,30 Unfortunately, there has been no large-scale, organized attempt to objectively define clinically applicable ultrasonographic criteria for naturally developing diffuse liver disease and to statistically correlate them with microscopic or biochemical findings. Furthermore, there has been no objective study completed to determine the number of animals with abnormal serum liver enzyme concentrations that actually have an abnormal ultrasonographic appearance to a clinician who is unaware of the biochemical analysis. Although there have been numerous studies31-49 in humans to compare the findings of ultrasonography with those of cytologic and histologic examinations, a specific, criterion-based comparison could not be found in the veterinary literature. In addition to brightnessmode ultrasonography, Doppler and ultrasonographic contrast techniques have been used,50-56 but they appear to address only specific circumstances. In 1 report57 in the veterinary literature, ultrasonographic criteria were compared to induced steroid hepatopathy.
Interpretation of any image, including liver ultrasonograms, is a mental amalgamation of background data (historical, physical, hematologic, and biochemical), information embedded in the images, risk assessment (age-related likelihood of neoplasia or specific breed-associated diseases), and knowledge of the pathologic and ultrasonographic findings among the diseases under consideration.32,35-37,39,40,42,46,48,49 To appropriately weight ultrasonographic appearance in this process, it must be determined whether there is any clinically exploitable diagnostic relationship between the ultrasonographic appearance (definable criteria) and any subset of the aforementioned diffuse diseases. This knowledge would appropriately place the ultrasonographic appearance of the liver parenchyma into the category of only an aid in the disease search or the category of a clinically applicable tool for use in differentiating among diffuse liver diseases in dogs and cats.
Because of the need to define the diagnostic yield from and the appropriate patient management role of liver-related ultrasonographic procedures for infiltrative disease, a retrospective study was devised. The objective was to determine whether commonly used ultrasonographic criteria for liver interpretation (eg, relative echogenicity in relation to that of other abdominal organs, characterization of liver surface and edges, and clarity of portal vein walls) without knowledge of patient status (normal or abnormal) or any other patient data would provide statistically relevant value for use in differentiating among microscopically defined groups of naturally developing diffuse liver diseases in dogs and cats. The hypotheses tested were that a significant relationship exists among specific or groups of ultrasonographic criteria and the various categories of diffuse liver disease and that regardless of the relationship defined by ultrasonographic criteria alone, it can be improved by statistical integration of routine biochemical and hematologic laboratory data.
Materials and Methods
Sample population—A retrospective evaluation was conducted of grayscale liver ultrasonograms obtained from dogs and cats examined because of possible liver disease at the University of Minnesota Veterinary Medical Center between February 2001 and November 2003. Ultrasonograms with previous interpretations of mass or nodular disease were excluded from the study. Ultrasonographic images of the remaining patients were screened by 2 investigators (DAF, BMD) to determine whether there was a representative number of liver images available and the images were of adequate quality for evaluation.
Ultrasonographic evaluation—Static hard-copy ultrasonographic images from 403 animals (299 dogs and 104 cats) were analyzed by 3 board-certified veterinary radiologists (DAF, KLA, LEZ) by use of a defined series of criteria (Appendices 1 and 2). Investigators did not have knowledge of the animal with regard to suspected disease, results of hematologic or serum biochemical analysis, or patient status. Randomly inter spersed among these images were images obtained from 60 animals (53 dogs and 7 cats) that did not have relevant pathologic changes in the liver (ie, normal liver), as determined on the basis of antemortem histologic or cytologic examinations or during necropsy. An option was included whereby a radiologist could refuse to interpret images for a specific animal on the basis that, in the opinion of the radiologist, the images were inadequate.
Interpretations of cytologic, histologic, or necropsy findings were screened by an investigator who was board-certified in veterinary internal medicine (RMH). These microscopic findings were then classified into 1 of 7 categories, which included no important findings (normal liver); inflammation (mononuclear or polymorphonuclear); round-cell neoplasia (lymphoma, mast cell, or histiocytosis); non–round-cell (infiltrative prenodular metastatic) neoplasia; lipidosis; vacuolar hepatopathy; and other. Equivocal or insufficiently specific data resulted in the elimination of that patient from the study. Only official cytologic, biochemical, or histologic findings from the University of Minnesota Veterinary Medical Center Laboratories or 2 regional outlets of national commercial veterinary clinical laboratoriesa,b were used in this study. To be included in the analysis, serum biochemical data had to be obtained within 14 days of the date on which the ultrasonographic examination was performed. Variables analyzed were activities of alanine transaminase, alkaline phosphatase, aspartate transaminase, and γ-glutamyltransferase; serum concentrations of albumin, BUN, creatinine, and total bilirubin; a CBC; WBC differential counts; and, when applicable, adrenal gland–related data (results for ACTH stimulation, plasma cortisol:urine cortisol ratio, and results for dexamethasone suppression tests).
Data analysis—All data were entered into a commercially available spreadsheetc and analyzed by use of a statistical analysis package.d Significance for all analyses was defined as values of P < 0.05. Because all patients did not have biochemical or hematologic data (ie, laboratory data) that met inclusion criteria, 2 groups of patients were defined (those with laboratory data and those without laboratory data). The groups were pooled for comparisons to categories of liver disease and ultrasonographic criteria. However, additional discriminant analyses were performed on the group with applicable laboratory data that included use of both the laboratory and ultrasonographic data.
Applied ultrasonographic criteria (Appendices 1 and 2) were compared to the 7 categories of liver status by use of χ2 analysis to identify significant relationships for each radiologist as well as to determine whether there was a pattern among radiologists that could be exploited for clinical identification of the 7 categories of liver status. Ultrasonographic criteria used by > 1 radiologist (on the basis of results of the χ2 analysis) were subjected to discriminant analysis by use of a forward stepwise regression procedure for variable inclusion (P < 0.05 for inclusion and P > 0.1 for exclusion) for each radiologist to evaluate the predictive capacity of these selected criteria for the 7 categories of liver status among all patients. A similar additional analysis was performed separately on the subset of patients with laboratory data to determine whether prediction of the category of liver status (percentage predicted correctly) could be improved by the addition of laboratory data. Another discriminant analysis was performed for each radiologist by use of their applied ultrasonographic criteria to evaluate the predictive capacity of these criteria for the 7 categories of liver status among all patients. Similar to analysis for the shared criteria, an additional analysis for specific applied ultrasonographic criteria was performed separately on the subset of patients with laboratory data.
Subsequently, ultrasonographic variables were created by use of ordinal metric values assigned to the ultrasonographic criteria (Appendices 1 and 2). A similar χ2 analysis and similar 4 rounds of discriminant analyses were performed by use of the metric ultrasonographic variables that were performed for the aforementioned nonmetric ultrasonographic criteria. A Spearman correlation matrix was used to compare each microscopic diagnosis (including that of normal liver) to the interpreted ultrasonographic criteria to determine whether significant relationships existed among the ultrasonographic, histologic, cytologic, biochemical, or hematologic variables. The null hypothesis was that no relationship existed between any of the ultrasonographic criteria, the microscopic diagnoses, or the laboratory values.
Further discriminant analyses were performed separately on each species to determine whether differences existed in the relationship among the ultrasonographic, histologic, cytologic, biochemical, or hematologic variables for the 7 categories of liver status. An additional discriminant analysis was performed by use of only the ordinal metric values for biochemical or hematologic data to determine whether significant relationships existed among the histologic, cytologic, biochemical, or hematologic variables for the 7 categories of liver status.
Results
The species and number of animals for each category of liver disease were defined (Table 1). Sex of animals did not have a significant effect, but there were only 12 sexually intact females and 15 sexually intact males in our data set. Two subgroups (403 animals [all animals] and 307 animals with laboratory data) were analyzed. Data were organized with regard to these 2 groups but subcategorized on the basis of species.
Categories of diffuse liver disease for 403 dogs and cats.
From the comparison among radiologists, 30 common ultrasonographic criteria (from the 51 variables in Appendices 1 and 2) were identified as having a relationship with 1 or more of the 7 categories of liver status in dogs, on the basis that ≥ 2 radiologists used a given criteria. These 30 criteria were identified by use of the χ2 test of homogenicity. An ultrasonographic criterion was selected for further analysis when the χ2 value of that comparison was ≥ 2. For example, a value of 0 in a cell within a χ2 table indicated no relationship, and a summed χ2 value of 3.84 for all cells in a table indicated significance with 1 df.
One subgroup included comparative echogenicity among parenchymal organs (liver, spleen, or kidneys), diffuse or patchy hyperechoic or hypoechoic character of the liver, uniform or coarse liver echotexture, variations in portal venous clarity, geometry of liver lobe edges and contours, variability in diameter of the caudal vena cava, identifiable peritoneal fluid, and character of peritoneal surfaces. However, when analyzed by use of a correlation matrix, none of these criteria had a Spearman correlation coefficient > 0.24, when compared to the 7 categories of liver status in dogs for any radiologist despite a significant outcome for the χ2 test. Similarly, none of these criteria had a Spearman correlation coefficient > 0.30 in a correlation matrix that compared use of ultrasonographic criteria among the 3 radiologists for dogs, despite a significant outcome for the χ2 test. These were all the available morphologic descriptors for the liver parenchyma of dogs, except parenchymal attenuation.
An additional subgroup of ultrasonographic criteria used by all radiologists had higher Spearman correlation coefficients for a similar analysis (Table 2). These included increased sound attenuation in liver parenchyma with increasing depth, thickness of the gallbladder wall, diameter of the bile duct, no or limited precipitate in the gallbladder, nondependent shadowing in the gallbladder, diameter of the hepatic vein relative to the diameter of the caudal vena cava, identifiable peritoneal fluid, spleen echotexture (normal vs abnormal), specific patterns of abnormal spleen echotexture, comparison of the echogenicity of liver parenchyma in relation to echogenicity of kidney parenchyma, and kidney echotexture.
Spearman correlation coefficients for pairs of radiologists between ultrasonographic ordinal values and any category of liver disease* in dogs.
From a similar comparison among radiologists, 30 common ultrasonographic criteria (from the 51 variables in Appendices 1 and 2) were identified as having a relationship with 1 or more of the 7 categories of liver status in cats. One subgroup included diffuse or patchy hyperechoic or hypoechoic character of the liver, uniform or coarse liver echotexture, patchy or nodular liver echotexture, variations in portal venous clarity, geometry of liver lobe edges and contours, thickness of the gallbladder wall, variability in diameter of the caudal vena cava, diameter of the hepatic vein relative to the diameter of the caudal vena cava, identifiable peritoneal fluid, and character of peritoneal surfaces. When criteria related directly to the liver parenchyma were analyzed by use of a correlation matrix, only diffuse liver hyperechogenicity, liver echogenicity less than that of the kidney cortex, lack of sound attenuation with increasing depth, and clarity of the portal vein wall were found to have correlation coefficients > 0.3 and < 0.5 when compared among the liver categories in cats for any radiologist, despite significant differences for the χ2 test. When extrahepatic criteria were analyzed by use of a correlation matrix, only evidence of complex peritoneal fluid, splenic nodules, splenic mass, or splenic infiltrative pattern were found to have correlation coefficients > 0.3 when compared among the categories of liver status in cats for any radiologist. In only 1 instance for 1 radiologist did the correlation coefficient for any criterion in cats exceed 0.5 (0.56 for complex peritoneal fluid). Similarly, none of these criteria had a Spearman correlation coefficient > 0.30 in a correlation matrix that compared the use of ultrasonographic criteria among the 3 radiologists for cats, despite significant differences for the χ2 test.
Similar to the results for dogs, these included most of the available morphologic descriptors for liver parenchyma, except parenchymal attenuation and liver echogenicity compared with kidney and spleen echogenicity. There was an additional subgroup of ultrasonographic criteria used in common among the radiologists that had higher Spearman correlation coefficients with the 7 categories of liver status (Table 3). These included liver echogenicity compared with kidney and spleen echogenicity, increased sound attenuation of the liver parenchyma with increasing depth, diameter of the bile duct, no or limited precipitate in the gallbladder, nondependent shadowing in the gallbladder, identifiable peritoneal fluid, and specific patterns of abnormal kidney echotexture.
Spearman correlation coefficients for pairs of radiologists between ultrasonographic ordinal values and any category of liver disease* in cats.
Use of the ultrasonographic criteria yielded significant identifiable relationships within each radiologist. Thus, the use of 1 variable may predict the use of another variable. An example of this would be the use of coarse liver echotexture at the same time as patchy hyperechoic or hypoechoic echogenicity. Despite similarities such as these in the descriptors, there were only 6 pairs (ie, 1 variable predicted the use of the other) of variables for both species, and these involved only 2 of the 3 radiologists.
Accuracy of the various biochemical, hematologic, and ultrasonographic variables for the correct classification of a patient in 1 of the 7 categories of liver status by use of discriminant analysis was limited. When only biochemical data were used in dogs, the accuracy for predicting the appropriate category of liver status (normal liver, vacuolar hepatopathy, lipidosis, inflammation, round-cell neoplasia, or non–round-cell neoplasia) was 25.8%, whereas the random probability of classifying a dog in the appropriate category of liver status was 16.7%. The only significantly relevant biochemical variables for dogs were total bilirubin and serum albumin concentrations. Similarly, when only hematologic data were used, accuracy for predicting the appropriate category of liver status in dogs was 23.6%. The only significantly relevant hematologic variables were total lymphocyte and unsegmented neutrophil counts. When only biochemical and hematologic data were used, accuracy for predicting the appropriate category of liver status in dogs was 31.4%. Accuracy when only defined ultrasonographic criteria chosen by the 3 radiologists were used to classify dogs in the appropriate category of liver status ranged from 32% to 41% among the radiologists, with 16.7% being the probability of a correct random classification. When only the biochemical and ultrasonographic variables were included in the discriminant analysis of dogs, accuracy ranged from 35% to 42% among the radiologists. When the hematologic, biochemical, and ultrasonographic variables were included in the discriminant analysis of dogs, maximum accuracy ranged from 33.7% to 37.4% among the radiologists.
When only biochemical data was used in cats, accuracy for predicting the appropriate category of liver status was 19.7%, whereas the random probability of correctly classifying a cat in the appropriate category of liver status was 16.7%. The only significantly relevant biochemical variables in cats were serum activities of alanine transaminase and γ-glutamyltransferase. Similarly, when only hematologic data were used, accuracy for predicting the appropriate category of liver status in cats was 16.9%. The only significantly relevant hematologic variable in cats was total RBC count. When both biochemical and hematologic data were used, accuracy for predicting the appropriate category of liver status in cats was 18.3%. Accuracy when only defined ultrasonographic criteria chosen by the 3 radiologists were used to classify cats in the appropriate category of liver status ranged from 43% to 63% among the radiologists (random probability of appropriate classification was 16.7%). When only the biochemical and ultrasonographic variables were included in the discriminant analysis of cats, accuracy ranged from 51% to 64% among the radiologists. When the hematologic, biochemical, and ultrasonographic variables were included in the discriminant analysis of cats, the maximum accuracy ranged from 31.7% to 54.9% among the radiologists.
For both species and all 3 radiologists, the applied group of ultrasonographic variables was available in all analyses. From the 30 ordinal metric values of variables in Appendices 1 and 2, 7 diffuse liver-related architecture categories were defined: portal vein clarity (clear or unclear), liver border contour (sharp or round), thickness of the gallbladder wall, splenic architecture (nodule or mass), echogenicity of liver parenchyma (hyperechoic or hypoechoic), sound attenuation of liver parenchyma (evident at any or all probe frequencies), and character of any identifiable peritoneal fluid (simple or complex). Ultrasonographic variables in dogs that were significantly related to the liver-status categories included gallbladder status and sound attenuation of the liver parenchyma. These were used twice as often as any of the other categories of variables; the other categories were all used equally. Ordinal metric values for ultrasonographic variables in cats that were significantly related to the liver-status categories included echotexture of the liver parenchyma, followed by spleen pattern and character of the peritoneal fluid, and then by portal vein clarity and sound attenuation of liver parenchyma, with the least used being liver border contour.
On the basis of discriminant analyses, overall accuracy of the ultrasonographic criteria chosen by the radiologists to predict liver-status category varied by approximately 10%, with a couple of notable but inconsistent results that were dependent on whether serum biochemical variables were included in the discriminant analysis along with the nonparametric ultrasonographic criteria (Tables 4 and 5). However, there was high variation in criteria-based accuracy among liver categories (including normal liver) for each radiologist, and there was no specific pattern for any specific radiologist. Use of biochemical variables improved the accuracy of diagnosis of vacuolar hepatopathy and lipidosis in dogs and vacuolar hepatopathy in cats. However, this was generally no better than an accuracy among radiologists of approximately 50% for dogs and between 50% and 65% for cats. Discriminant analyses by use of ordinal metric values assigned to ultrasonographic criteria were somewhat less accurate than analyses that used only the nonparametric ultrasonographic criteria.
Calculated accuracy of categorization of diffuse liver disease in dogs for each of 3 radiologists by use of individually applied ultrasonographic criteria and with or without inclusion of available serum biochemical variables in discriminant analysis.
Calculated accuracy of categorization of diffuse liver disease in cats for each of 3 radiologists by use of individually applied ultrasonographic criteria and with or without inclusion of available serum biochemical variables in discriminant analysis.
Discussion
Mean accuracy for 3 board-certified veterinary radiologists when their choices of ultrasonographic criteria were statistically analyzed against the 7 categories of diffuse liver disease was 36.5% for dogs. When biochemical variables were combined with the ultrasonographic criteria, the mean accuracy for dogs improved slightly to 39.1%. For comparison, mean accuracy for the same radiologists when their choices of ultrasonographic criteria were statistically analyzed against the 7 categories of diffuse liver disease was 54.6% for cats. When combined with biochemical variables, accuracy for cats improved slightly to 57.5%. Similar analyses relating biochemical or hematologic variables without the ultrasonographic criteria to the 7 categories of liver status did not exceed an accuracy of 32% in either species, whereas the random probability of a correct response was 16.7%. Even the accuracy of identification of normal liver when only ultrasonographic criteria were used was highly variable (0% to 80%) for each species and among the radiologists.
In short, overall accuracy of prediction of the category of liver status when only ultrasonographic criteria were used was < 40% for dogs and < 60% for cats, regardless of whether biochemical or hematologic variables were added to the statistical model. The exception may be fatty liver in cats because criterion-based assessment yielded a correct classification in at least 50% of cats by all radiologists and exceeded 70% in 2 of 3 radiologists with or without inclusion of biochemical variables. For comparison, accuracy of prediction of the category of liver status when biochemical or hematologic (or both) variables were used was only nominally better than random assignment. The addition of hematologic or biochemical variables to the statistical model along with the ultrasonographic criteria did not significantly improve the accuracy of correlations with the 7 liver-status categories in either species.
Of interest in both dogs and cats, almost none of the parenchymal descriptors (comparative echogenicity among parenchymal organs [liver, spleen, and kidneys], diffuse or patchy hyperechoic or hypoechoic liver character, uniform or coarse liver echotexture, variations in portal venous clarity, or geometry of liver lobe edges and contours) had a clinically applicable correlation between pairs of radiologists, despite significant relationships for each radiologist. The problem was that the low correlation coefficients (≤ 0.3) explained < 10% of the variation in radiologists' use of these criteria. However, sound attenuation of liver parenchyma as well as other less parenchymal-specific criteria were more frequently used by > 1 radiologist. Similarly, there were some ultrasonographic criteria that were significantly related within a radiologist in that the use of 1 criterion would be highly predictive of the use of another. Because there were only about 6 such pairs of related ultrasonographic criteria for 2 or 3 radiologists, the application of descriptive ultrasonographic criteria was apparently highly specific among the radiologists. It seems that individuals have a set of criteria that they typically group together and use more than other criteria, even when a larger array of choices is available. Furthermore, when the ultrasonographic criteria were compared to the categories of liver status in dogs for each radiologist, neither the intrahepatic nor extrahepatic criteria had clinically applicable correlation coefficients (> 0.3). By comparison, when the ultrasonographic criteria were compared to the categories of liver status in cats for each radiologist, only 6 liver parenchymal and 4 extrahepatic criteria had correlation coefficients with limited clinical applicability (0.3 < r < 0.6). These criteria appeared primarily related to liver lipidosis, which comprised approximately 50% of the conditions in cats.
It appeared that the nonparenchymal findings (thickness of the gallbladder wall, diameter of the bile duct, no or limited precipitate in the gallbladder, nondependent shadowing in the gallbladder, diameter of the hepatic vein relative to the diameter of the caudal vena cava, identifiable peritoneal fluid, spleen echotexture [normal vs abnormal], and specific patterns of abnormal spleen echotexture) and sound attenuation of liver parenchyma with increasing depth played an important role in the categorization process, but the specific use of these findings outside of a statistical model would be limited, depending on the disease category in question. Of interest, sound attenuation of liver parenchyma with increasing depth appeared to facilitate categorization of liver status to a greater degree in dogs than in cats, despite the greater accuracy of status categorization in cats and the preponderance of lipidosis among cats in the sample population.
We acknowledge that there are shortcomings in the use of results of cytologic examinations, compared with results of histologic examinations, as a basis for the diagnosis of certain types of liver disease.29,30,58,59 The use of only cytologic assessments coupled with a review by an experienced specialist in veterinary internal medicine seemed an acceptable, although not ideal, approach to accruing sufficient numbers of certain diseases (eg, hepatic lipidosis in cats) to permit assessment of the clinical relevance of ultrasonographic criteria. From our database, because of prevailing clinical philosophy, some diseases were not assessed by examination of core tissue biopsy specimens because of the clinical circumstances and risk of hemorrhage. Therefore, rather than exclude these data, we accepted the cytologic data when histologic data were not available. We also acknowledge that retrospective application of criteria, although used in an objective and unbiased manner, of only available static images may limit the full appreciation of subtleties evident on high-resolution screens in real-time evaluations. However, it was the only practical means available to enable us to separate a radiologist who selected the ultrasonographic criteria from the patient and its associated information.
It must be remembered that the analyses reported here were based on the use of individual ultrasonographic criteria combined or not combined statistically with individual hematologic or biochemical variables in an attempt to appropriately classify diffuse liver disease of dogs and cats into 1 of 7 categories by use of discriminant analyses. This is not a comparison of diagnoses rendered by the involved radiologists. Specific diagnoses were not solicited during the criteria-based assessment, and each radiologist had no knowledge about the clinical status or results of laboratory analyses for the patients evaluated. The objective was to determine whether a criterion-based approach to categorization of diffuse liver disease would be acceptably accurate to be useful clinically, regardless of whether used alone or in combination with other objective data such as biochemical or hematologic variables. The use of K analyses among radiologists was not possible for most analyses because of variation in the use of the available criteria by each of the radiologists. In addition, it was hypothesized that specific ultrasonographic criteria (or combinations of criteria) would be found with strong correlations to specific disease categories to facilitate diagnoses among radiologists. However, it appears that a criterion-based approach is not a clinically applicable solution to the dilemma of diffuse liver disease in dogs and cats, regardless of how rigorous the application or whether it is paired with hematologic or biochemical data. Reports31-49 in the human literature on diffuse liver disease revealed that the conclusions were varied and depended on the population studied, the inclusion criteria applied, the classification and data assessment schemes used, and the vintage of equipment used. These findings provided no additional applicable insights to our dilemma.
Application of defined ultrasonographic criteria in an objective unbiased (unaware of the patient, the clinical suspicion, and the biochemical values) manner yielded insufficient accuracy to be clinically useful in the categorization of naturally developing diffuse liver disease in dogs and cats. Therefore, it appears that the ultrasonographic characterization of diffuse liver disease in dogs and cats has limited value in discriminating among the categories of diffuse liver diseases that do not disrupt the ultrasonographic architecture of the liver. Even for lipidosis in cats, the calculated accuracy was, at best, only approximately 70% by use of this approach. It remains debatable whether a clinician with knowledge of the clinical situation and any other patient data would be better able to predict or even rank the categories of diffuse liver disease on a case-by-case basis. Therefore, we conclude that care must be exercised in the diagnosis of (or speculation on) the various types of diffuse liver disease on the basis of the ultrasonographic appearance alone or when used in combination with routine clinical biochemical or hematologic data. In addition, it appears that discrimination of normal from abnormal is not sufficiently accurate or predictable to be applicable in a clinical situation. We presume that there are insufficient differences among the acoustic impedances of the various diffuse liver diseases and the normal liver to permit clinically applicable characterization of liver tissue and discrimination among infiltrative liver conditions or abnormal from normal tissues with current ultrasonographic technology. Therefore, a cytologic or histologic analysis is of paramount importance in the assessment of diffuse liver disease in dogs and cats.
Antech (Midwest), Alsip, Ill.
Veterinary Diagnostic Services, Marshfield Clinical Laboratories, Marshfield, Wis.
MS Excel, Microsoft Corp, Redmond, Wash.
SPSS, version 10.05, SPSS Inc, Chicago, Ill.
References
- 1.
Biller DS, Kantrowitz B, Miyabayashi T. Ultrasonography of diffuse liver disease. J Vet Intern Med 1992;6:71–76.
- 2.
Cruz-Arambulo R, Wrigley R, Powers B. Sonographic features of histiocytic neoplasms in the canine abdomen. Vet Radiol Ultrasound 2004;45:554–558.
- 3.
d'Anjou MA, Penninck D, Cornejo L, et al. Ultrasonographic diagnosis of portosystemic shunting in dogs and cats. Vet Radiol Ultrasound 2004;45:424–437.
- 4.
Drost WT, Henry GA, Meinkoth JH, et al. Quantification of hepatic and renal cortical echogenicity in clinically normal cats. Am J Vet Res 2000;61:1016–1020.
- 5.
Lamb CR, Hartzband LE, Tidwell AS, et al. Ultrasonographic findings in hepatic and splenic lymphosarcoma in dogs and cats. Vet Radiol 1991;32:117–120.
- 6.
Nicoll RG, Jackson MW, Knipp BS, et al. Quantitative ultrasonography of the liver in cats during obesity induction and dietary restriction. Res Vet Sci 1998;64:1–6.
- 7.
O'Brien RT, Iani M, Matheson J, et al. Contrast harmonic ultrasound of spontaneous liver nodules in 32 dogs. Vet Radiol Ultrasound 2004;45:547–553.
- 8.
Schwarz LA, Penninck DG, Leveille-Webster C. Hepatic abscesses in 13 dogs: a review of the ultrasonographic findings, clinical data and therapeutic options. Vet Radiol Ultrasound 1998;39:357–365.
- 9.
Szatmari V, Rothuizen J, van den Ingh TS, et al. Ultrasonographic findings in dogs with hyperammonemia: 90 cases (2000–2002). J Am Vet Med Assoc 2004;224:717–727.
- 10.
Voros K, Vrabely T, Papp L, et al. Correlation of ultrasonographic and pathomorphological findings in canine hepatic diseases. J Small Anim Pract 1991;32:627–634.
- 11.
Whiteley MB, Feeney DA, Whiteley LO, et al. Ultrasonographic appearance of primary and metastatic canine hepatic tumors: 48 cases (1981–1987). J Ultrasound Med 1989;8:621–630.
- 12.
Andersson M, Sevelius E. Breed, sex and age distribution in dogs with chronic liver disease—a demographic study. J Small Anim Pract 1991;32:1–5.
- 13.
Blunden AS, Smith KC. Generalized amyloidosis and acute liver hemorrhage in 4 cats. J Small Anim Pract 1992;33:566–570.
- 14.
Brown PJ, Henderson JP, Galloway P, et al. Peliosis hepatis and telangiectasis in 18 cats. J Small Anim Pract 1994;35:73–77.
- 15.
Chapman BL, Hendrick MJ, Washabua RJ. Granulomatous hepatitis in dogs: nine cases (1987–1990). J Am Vet Med Assoc 1993;203:680–684.
- 16.
Carmichael KP, Miller M, Rawlings CA, et al. Clinical, hematologic, and biochemical features of a syndrome in Bernese Mountain Dogs characterized by hepatocellular degeneration. J Am Vet Med Assoc 1996;208:1277–1279.
- 17.
Crawford MA, Schall WD, Jensen RK, et al. Chronic active hepatitis in 26 Doberman Pinschers. J Am Vet Med Assoc 1985;187:1343–1350.
- 18.
Danderslebel JV, Janichen T, Fiebiger I, et al. Lipid storage disease in the Fox Terrier similar to Wolman's disease of man. Tierarztl Prax 1986;14:253–263.
- 19.
Dill-Macky E. Chronic hepatitis in dogs. Vet Clin North Am Small Anim Pract 1995;25:387–398.
- 20.
Doige CE, Furneaux RW. Liver disease and intrahepatic portal hypertension in the dog. Can Vet J 1975;16:209–214.
- 21.
Gagne JM, Armstrong PJ, Weiss DJ, et al. Clinical features of inflammatory liver disease in cats: 41 cases (1983–1993). J Am Vet Med Assoc 1999;214:513–516.
- 22.
Haywood S, Rutgers HC, Christian MK. Hepatitis and copper accumulation in Skye Terriers. Vet Pathol 1988;25:408–414.
- 23.
Hultgren BD, Stevens JB, Hardy RM. Inherited, chronic progressive hepatic degeneration in Bedlington terriers with increased liver copper concentrations: clinical and pathologic observations and comparison with other copper-associated liver diseases. Am J Vet Res 1986;47:365–377.
- 24.
O'Keefe DA, Guillermo-Couto C, Burke-Schwartz C, et al. Systemic mastocytosis in 16 dogs. J Vet Intern Med 1987;1:75–80.
- 25.
Rutgers HC, Haywood S, Kelly DF. Idiopathic hepatic fibrosis in 15 dogs. Vet Rec 1993;133:115–118.
- 26.
Schwartz SG, Mitchell SL, Keating JH, et al. Liver lobe torsion in dogs: 13 cases (1995–2004). J Am Vet Med Assoc 2006;228:242–247.
- 27.
Shiga A, Shirota K, Shida T, et al. Morphologic study on a case of canine hepatic nodular fibrosis. J Vet Med Sci 1996;58:281–284.
- 28.
Waters CB, Morrison WB, DeNicola DB, et al. Giant cell variant of malignant fibrous histiocytoma in dogs: 10 cases (1986–1993). J Am Vet Med Assoc 1994;205:1420–1424.
- 29.
Wang KY, Panciera DL, Al-Rukibat RK, et al. Accuracy of ultrasound-guided fine-needle aspiration of the liver and cytologic findings in dogs and cats: 97 cases (1990–2000). J Am Vet Med Assoc 2004;224:75–78.
- 30.
Willard MD, Weeks BR, Johnson M. Fine-needle aspirate cytology suggesting hepatic lipidosis in four cats with infiltrative hepatic disease. J Feline Med Surg 1999;1:215–220.
- 31.
Afdhal NH, Nunes D. Evaluation of liver fibrosis: a concise review. Am J Gastroenterol 2004;99:1160–1174.
- 32.
Angtuaco TL, Lal SK, Banaad-Omiotek GD, et al. Current liver biopsy practices for suspected parenchymal liver diseases in the United States: the evolving role of radiologists. Am J Gastroenterol 2002;97:1468–1471.
- 33.
Cao GT, Shi PF, Hu B. Liver fibrosis identification based on ultrasound images captured under varied imaging protocols. J Zhejiang Univ Sci B 2005;6:1107–1114.
- 34.
Ding W, He XJ. Fine needle aspiration cytology in the diagnosis of liver lesions. Hepatobiliary Pancreat Dis Int 2004;3:90–92.
- 35.
Gaitini D, Baruch Y, Ghersin E, et al. Feasibility study of ultrasonic fatty liver biopsy: texture vs. attenuation and backscatter. Ultrasound Med Biol 2004;30:1321–1327.
- 36.
Gangeh MJ, Hanmandlu M, Bister M. A fuzzy-based texture analysis for tissue characterization of diffused liver diseases on B-scan images. Biomed Sci Instrum 2002;38:369–374.
- 37.
Graif M, Yanuka M, Baraz M, et al. Quantitative estimation of attenuation in ultrasound video images: correlation with histology in diffuse liver disease. Invest Radiol 2000;35:319–324.
- 38.
Hamer OW, Aguirre DA, Casola G, et al. Imaging features of perivascular fatty infiltration of the liver: initial observations. Radiology 2005;237:159–169.
- 39.
Joseph AE, Savarymuttu SH, al-Sam S, et al. Comparison of liver histology with ultrasonography in assessing diffuse liver disease. Clin Radiol 1991;43:26–31.
- 40.
Joy D, Thava VR, Scott BB. Diagnosis of fatty liver disease: is biopsy necessary? Eur J Gastroenterol Hepatol 2003;15:539–543.
- 41.
Mathiesen UL, Franzen LE, Aselius H, et al. Increased liver echogenicity at ultrasound examination reflects degree of steatosis but not of fibrosis in asymptomatic patients with mild/moderate abnormalities of liver transaminases. Dig Liver Dis 2002;34:516–522.
- 42.
Meziri M, Pereira WC, Abdelwahab A, et al. In vitro chronic hepatic disease characterization with a multiparametric ultrasonic approach. Ultrasonics 2005;43:305–313.
- 43.
Nicolau C, Bianchi L, Vilana R. Gray-scale ultrasound in hepatic cirrhosis and chronic hepatitis: diagnosis, screening, and intervention. Semin Ultrasound CT MR 2002;23:3–18.
- 44.
Taylor KJW, Reily CA, Hammers L, et al. Quantitative ultrasound attenuation in normal liver and in patients with diffuse liver disease. Radiology 1986;160:65–71.
- 45.
Tchelepi H, Ralls PW, Radin R, et al. Sonography of diffuse liver disease. J Ultrasound Med 2002;21:1023–1032.
- 46.
Vehmas T, Kaukiainen A, Luoma K, et al. Liver echogenicity: measurement or visual grading? Comput Med Imaging Graph 2004;28:289–293.
- 47.
Vilgrain V. Ultrasound of diffuse liver disease and portal hypertension. Eur Radiol 2001;11:1563–1577.
- 48.
Yamada H, Ebara M, Yamaguchi T, et al. A pilot approach for quantitative assessment of liver fibrosis using ultrasound: preliminary results in 79 cases. J Hepatol 2006;44:68–75.
- 49.
Yeh WC, Huang SW, Li PC. Liver fibrosis grade classification with B-mode ultrasound. Ultrasound Med Biol 2003;29:1229–1235.
- 50.
Aube C, Winkfield B, Oberti F, et al. New doppler ultrasound signs improve the non-invasive diagnosis of cirrhosis or severe liver fibrosis. Eur J Gastroenterol Hepatol 2004;16:743–751.
- 51.
Bernatik T, Strobel D, Hahn EG, et al. Doppler measurements: a surrogate marker of liver fibrosis? Eur J Gastroenterol Hepatol 2002;14:383–387.
- 52.
Kaneko T, Teshigawara O, Sugimoto H, et al. Signal intensity of the liver parenchyma in microbubble contrast agent in the late liver phase reflects advanced fibrosis of the liver. Liver Int 2005;25:288–293.
- 53.
Martinez-Noguera A, Montserrat E, Torrubia S, et al. Doppler in hepatic cirrhosis and chronic hepatitis. Semin Ultrasound CT MR 2002;23:19–36.
- 54.
Mihmanli I, Kantarci F, Yilmaz MH, et al. Effect of diffuse fatty infiltration of the liver on hepatic artery resistance index. J Clin Ultrasound 2005;33:95–99.
- 55.
O'Donohue J, Ng C, Catnach S, et al. Diagnostic value of Doppler assessment of the hepatic and portal vessels and ultrasound of the spleen in liver disease. Eur J Gastroenterol Hepatol 2004;16:147–155.
- 56.
Oguzkurt L, Yildirim T, Torun D, et al. Hepatic vein doppler waveform in patients with diffuse fatty infiltration of the liver. Eur J Radiol 2005;54:253–257.
- 57.↑
Syakalima M, Takiguchi M, Yasuda J, et al. Comparison of attenuation and liver-kidney contrast of liver ultrasonographs with histology and biochemistry in dogs with experimentally induced steroid hepatopathy. Vet Q 1998;20:18–22.
- 58.
Stockhaus C, Teske E, van den Ingh TS, et al. The influence of age on the cytology of the liver in healthy dogs. Vet Pathol 2002;39:154–158.
- 59.
Weiss DJ, Moritz A. Liver cytology. Vet Clin North Am Small Anim Pract 2002;32:1267–1291.
Appendix 1
Intrahepatic criteria* for objective ultrasonographic assessment of liver disease in dogs and cats.
Appendix 2
Extrahepatic criteria* for objective ultrasonographic assessment of liver disease in dogs and cats.
