For this retrospective study, the medical records of all canine patients who underwent a diagnostic workup for various symptoms at the Clinic for Small Animals of the University of Veterinary Medicine Hannover Foundation between November 2021 and March 2023 were reviewed. This included a SDCT examination of the abdomen (Philips IQon Spectral CT, Philips Health Care Germany). The inclusion criteria were splenic lesions visible in SDCT and a histopathological examination of these lesions. Samples were collected through splenectomy or ultrasound-guided Tru-Cut biopsies.

From all dogs that met the inclusion criteria, data on breed, sex, age at the time of imaging, weight, and reason for initial presentation were collected using simple calculation software (Microsoft® Excel, Version 16.0). Since the data used was part of routine clinical work, no approval from an ethics committee was required.

All histopathological samples were acquired through ultrasound-guided Tru-Cut biopsy or total splenectomy during laparotomy at the Clinic for Small Animals of the University of Veterinary Medicine Hannover Foundation and were subsequently fixed in 10% buffered formalin. Soft tissue semi-automatic biopsy needles with adjustable penetration depth (BIO CORE 2, 16 G (1.67 mm x 10 cm), HVM Medical Products GmbH, Fulda, Germany) were used for the biopsies. The location for the Tru-Cut biopsies was selected based on ultrasound imaging and depended on the decision of the ultrasonographer. The number of biopsies taken depended on the appearance of the lesions and the quality of the samples obtained (2–3 per patient). Due to the retrospective nature of the study, the location and number of Tru-Cut biopsies varied and were not standardized.

All tissue samples underwent examination at the Institute for Pathology of the University of Veterinary Medicine Hannover Foundation by Diplomates of the European College of Veterinary Pathologists (ECVP). The samples were dehydrated in a graded series of ethanol and then embedded in paraffin wax. From these samples, 2–3 μm thick slices were prepared and stained with hematoxylin–eosin. Subsequently, these histopathological preparations were examined under a light microscope at up to 100x magnification according to ECVP standards and were classified into the following groups based on their predominant findings: nodular hyperplasia, extramedullary hematopoiesis, hemorrhage, thrombosis, hemosiderosis, neoplasia, and inflammation. If more than one pathological finding was present, for instance, hemangiosarcoma and hemosiderosis, the patients were classified according to the predominant and clinically more relevant diagnosis, which in this case was categorized as hemangiosarcoma. All recorded findings were entered into the patient’s records.

Appropriate software (Microsoft Excel 2019; GraphPad Prism 10.2.2) was utilized for the statistical evaluation of the collected data. Initially, the differences in HU-values of corresponding VNC and TUE images of each ROI for reference organs, normal spleen, and splenic lesions were determined through subtraction (19) and verified by establishing cut-off values of ≤ 5, ≤ 10, and ≤ 15 HUs based on veterinary and human medical studies (13–15, 17, 19). Values with a difference in HUs of ≤ 10 were deemed negligible, while differences between 10 and 15 HUs were considered acceptable (13, 14). The cut-off value of ≤ 5 HUs was also included to assess and illustrate the potential accuracy of VNC images. A two one-sided t-test (TOST) was performed to test the equivalence of TUE and VNC series, utilizing an opposing null hypothesis that mean differences between TUE and VNC were greater than 5, 10, or 15 HUs, respectively. To emphasize the analysis of various lesions and pathologies, all measured ROIs for each respective category were grouped and analyzed independently of the number of dogs.

Subsequently, multiple comparisons of the mean differences were conducted on various imaging characteristics to assess their influence on the accuracy and comparability of VNC with TUE images, using a one-way analysis of variance (ANOVA) test, including Tukey’s multiple comparisons test for p-value correction. An unpaired t-test was performed to evaluate the influence of the location of regions of interest situated at the center or the periphery of focal lesions. For testing normal distribution, the D’Agostino and Pearson test was performed. Simple column bar graphs and box-and-whisker plots were created. Summarised data were reported as mean ± standard deviation (SD). A p-value of less than 0.05 was deemed significant. All tests were two-sided.

In total, 30 dogs met the inclusion criteria. The affected breeds included the following: Rhodesian Ridgeback (4), Golden Retriever (3), Labrador Retriever (3), American Staffordshire Terrier (1), Appenzeller Mountain Dog (1), Beagle (1), Fox Terrier (1), French Bulldog (1), Galgo Español (1), Standard Poodle (1), Magyar Vizsla (1), Nova Scotia Duck Tolling Retriever (1), Whippet (1), and 10 crossbreeds. Fourteen dogs were female (of which eleven were spayed), while sixteen were male (also eleven were spayed). The median age of the dogs at presentation was 11 years, ranging from 4.1 years to 13.2 years. The median weight was 26.3 kg, ranging from 4.8 kg to 50.4 kg.

Eighteen of the 30 patients were initially presented for general staging due to other findings, such as prostatopathy or neoplasia, in which a structurally altered spleen was a secondary finding. Five patients were referred to the clinic with acute abdomen, and seven patients exhibited lethargy, where a splenic lesion was the primary finding.

Twenty-eight histopathological samples were obtained following a splenectomy, and two samples were collected via a tru-cut biopsy under ultrasonographic guidance. The histopathological examination revealed ten malignant pathologies: five hemangiosarcomas, four lymphomas, and one metastatic anal sac carcinoma, along with 18 benign lesions, which included nine nodular hyperplasias, four cases of extramedullary hematopoiesis, three hematoma, one thrombus in the splenic body, and one instance of hemosiderosis. There were no cases of splenitis or splenic abscessation. Two splenic samples were unremarkable, one being a Tru-Cut biopsy and the other a splenectomy.

Hemosiderosis was a common finding in 14 cases of both benign and malignant splenic pathologies, while extramedullary hematopoiesis was noted as an additional finding in four benign cases and one malignant case.

Among the 30 spleens examined, 21 displayed multifocal changes, seven had focal changes, and two exhibited diffuse changes. Twelve patients had free peritoneal fluid (mild: 3/30, moderate: 6/30, severe: 3/30). Of the 21 cases of multifocal splenopathies, 13 patients had more than 10 lesions, four had between 6 and 10 lesions, and four had between 2 and 5 lesions. The lesions were categorized based on their location: splenic head: 2/30, splenic body: 2/30, splenic tail: 4/30, head and body: 1/30, body and tail: 1/30, all three regions: 20/30.

Four patients exhibited partial mineralization in at least one of their lesions (mild: 2/30, moderate: 1/30, severe: 1/30). Capsule formation was only observed in one lesion. Eleven patients presented lesions with cavitation (8/30 in one lesion, 3/30 in multiple lesions). The attenuation of lesions pre-contrast compared to the surrounding tissue was hypoattenuating in 15, isoattenuating in six, and hyperattenuating in two patients. Five patients displayed different types of attenuation within the same lesion or across several lesions. In the post-contrast series, eight of the 28 patients with focal or multifocal lesions showed hypoattenuation of lesions compared to surrounding tissue, eleven patients had hyperattenuating lesions, and nine patients exhibited different types of attenuation within the same lesion or in multiple lesions. The enhancement pattern of splenic lesions, or the entire spleen in diffuse cases, was homogeneous in nine patients and heterogeneous in 13 patients. Seven patients demonstrated various enhancement patterns across different lesions, with three of these patients having one target lesion each. One spleen showed no enhancement at all. The maximum extent of the splenic lesions ranged from 0.7 cm to 18.2 cm, measured across all three planes.

A total of 305 ROIs were drawn, of which 215 were located in the splenic parenchyma. This resulted in 610 Hounsfield unit values (in VNC and TUE series) that were statistically evaluated.

The differences in HUs between TUE and VNC for all measured ROIs were ≤ 15 HUs in 98.0%, ≤ 10 HUs in 93.1%, and at least 66.2% were below the limit of 5 HUs.

The values for the reference organs exhibited a strong correlation between VNC and TUE images, as well as conventional post-contrast images with MonoE images at an energy level of 70 keV (Figure 3).

For the reference organs and normal splenic tissue, 100% of the calculated differences in HUs between TUE and VNC were below the threshold of 15 HUs. In diffusely altered spleens and those with focal lesions, the differences between TUE and VNC were below the threshold of 10 HUs in 100% of cases. In spleens with multifocal lesions, the differences were ≤ 15 HUs in 94.7% and ≤ 10 HUs in 91.2% of 114 drawn ROIs (Figure 4). Regarding their histopathological diagnoses, patients with hemangiosarcoma (5), lymphoma (4), metastatic carcinoma (1), nodular hyperplasia (9), hematoma (3), and thrombus (1) showed that 100% of the evaluated differences between TUE and VNC were ≤ 10 HUs. Among patients with extramedullary hematopoiesis (4), 81.5% had differences ≤ 15 HUs, and 66.7% had differences ≤ 10 HUs. In one patient with hemosiderosis as the primary histopathological finding, 75% of measured differences between TUE and VNC were ≤ 10 HUs. Categorized into the cut-off values of ≤ 5, ≤ 10, and ≤ 15 HUs, the differences in HUs between TUE and VNC images for the reference organs, various groups regarding their imaging characteristics, and the splenic pathologies are listed in Table 4. The collected data showed predominantly a Gaussian distribution.

The two one-sided t-tests conducted to evaluate the equivalence of TUE and VNC regarding the limits of ≤ 15, ≤ 10, and ≤ 5 HU differences, aimed at supporting the previously mentioned results, indicated a strong agreement between TUE and VNC images across all ROIs (p < 0.0001 for differences ≤ 10 HUs). However, the group with splenic lesions due to extramedullary hematopoiesis showed no significant equivalence of HUs in the TUE and VNC series, with a p-value of 0.0693 for differences ≤ 15 HUs. All results of the TOST are illustrated in Table 5.

Only two patients had differences between the TUE and VNC values of more than 15 HUs. One patient, who had four small soft tissue attenuating lesions, exhibited a difference of 15.2 HUs in one of these lesions. The remaining three lesions in this spleen showed differences below the threshold of 10 HUs. This patient had hemosiderosis on histopathology.

The second patient, who demonstrated larger discrepancies between VNC and TUE values, had several elongated mineralizations at the splenic hilus (Figure 5) and multiple central parenchymal lesions, both with and without mineralization. The hilar mineralizations were misinterpreted as areas of contrast enhancement and were incorrectly removed by the VNC algorithm (Figure 5D). The central parenchymal lesions became evident following contrast enhancement (e.g., orange arrows in Figures 6B,C). All ROIs that included mineralizations in this patient showed significant differences between VNC and TUE images. The discrepancies measured 53.2 HUs in the polygonal ROI of the largest mineralized parenchymal lesion, with 27 and 18.2 HUs in its periphery and 54.2 and 61.2 HUs in its center. The largest parenchymal lesion of this patient was not mineralized, and the polygonal ROI of this lesion exhibited a difference of less than 10 HUs. One of these non-mineralized parenchymal lesions is indicated by the orange arrows in Figure 6. The VNC algorithm accurately removed the contrast medium in these lesions (Figure 6D). This patient had extramedullary hematopoiesis and hemosiderosis on histopathology and was unique in having some lesions that displayed a moderate degree of mineralization.

The patients were categorized according to specific imaging characteristics of the splenic lesions to evaluate whether these influenced the performance of the VNC algorithm.

The ordinary one-way ANOVA conducted to assess the influence of mineralization in lesions on the accuracy of the VNC technique produced a significant result, with p < 0.0001. The differences in HUs between VNC and TUE images for lesions with moderate mineralization (1) were considerably higher than those for lesions with no (26), mild (2), or severe (1) mineralization, with an adjusted p-value < 0.0001 in each comparison (Figure 7).

The comparison of differences in HU values between VNC and TUE images in patients with diffusely, focally, and multifocally altered spleens showed no significant difference, despite the patient with mineralized lesions being part of the multifocal group (p_ANOVA = 0.0768). The adjusted p-value for the comparison of diffusely altered to focally altered spleens was 0.9134; for diffusely altered to multifocally altered spleens, it was 0.6087; and for focally altered spleens compared to the multifocal altered group, it was 0.0750.

The comparison of patients regarding their type of attenuation before the administration of CM showed significant results (p_ANOVA < 0.0001). The largest difference in HUs between VNC and TUE images was noted in splenic lesions that were hyperattenuating compared to the surrounding tissue in TUE images, as observed in two patients. With an adjusted p-value < 0.0001 in each comparison, these results were significant when compared to lesions that were iso-attenuating (6), hypoattenuating (15), or lesions with various types of attenuation (5) in TUE images (Figure 8). Nevertheless, three of these four mean differences remained below the threshold of 10 HUs, which is considered negligible, while the mean difference for the group with hyperattenuating lesions was 17.37 HUs and included the patient with the highest differences due to incorrect measurement of its mineralized lesions. The two patients with diffusely altered spleens were excluded from this comparison due to the absence of normal-appearing parenchyma for comparison.

The comparison of various types of pathology showed a significantly higher mean difference in the HU values between TUE and VNC in patients with extramedullary hematopoiesis (4) compared to those with hemangiosarcoma (5), lymphoma (4), metastasis of anal-sac carcinoma (1), nodular hyperplasia (9), and hematoma (3) (p_ANOVA < 0.0001; adjusted p-values ranging from < 0.0001 to 0.0353). All other individual comparisons, including hemosiderosis (1), thrombus (1), and histopathologically unremarkable spleens (2), showed no significance. A comparison of malignant and benign lesions revealed no statistical significance.

All results of the multiple comparisons, including the presence of cavitation, the type of attenuation of the splenic lesions post-CM compared to the surrounding tissue, the type of enhancement pattern, and the degree of enhancement, are presented in Supplementary Table S1.

An unpaired t-test showed no significant differences in HU values between TUE and VNC images within regions of interest located at the center compared to the periphery of the focal lesions (p = 0.1810), irrespective of the presence of mineralization (p = 0.2215).

The quality of the VNC images was slightly superior to that of the TUE images, receiving an average score of 4.07 (SD ± 0.25) on a five-point Likert scale. In two patients the image quality of VNC images was assessed as much better (score 5) than the image quality of TUE images. All other patients (28/30) received a score of 4.

The iodine subtraction in the pancreas, gallbladder, spinal musculature, liver, spleen, and main lesion was nearly complete, with an average score of 4.37 points (SD ± 0.30). The spinal musculature achieved the highest score, at 4.97 points (SD ± 0.18). Overall, the splenic tissue scored 4.13 points (SD ± 0.72), while the main lesions of the spleen (excluding the two diffuse pathologies) received a minimum score of 3.4 points (SD ± 1.2).