Accuracy of B-mode ultrasound and ARFI elastography in predicting malignancy of canine splenic lesions

The objective was to evaluate the accuracy of B-mode ultrasonography and ARFI elastography in detecting malignancy in canine splenic lesions. Thirty-seven spleens with abnormalities (16 benign and 21 malignant) from dogs of different breeds and ages were evaluated. Echogenicity, echotexture, organ length and height were evaluated using B-mode. By ARFI elastography, tissue stiffness was evaluated qualitatively (elastogram) and quantitatively (measuring the shear wave velocity—SWV). Lesions were classified as diffuse, focal or multifocal (cranial, medial or caudal portion) and comparisons of the SWV between the injured and non-injured areas were performed. In the B-mode, no features were associated to malignancy (P > 0.05). In the elastogram, 35 spleens were non-deformable and 2 deformable, having no association with malignancy. The greater SWV was observed in malignant lesions (3.4 ± 0.6 m/s), followed by areas free from alterations (2.1 ± 0.3 m/s) and benign lesions (1.7 ± 0.5 m/s), with difference between groups (P < 0.0001). It was found that a SWV > 2.6 m/s indicates malignancy of canine splenic lesions (sensitivity of 95%, specificity of 100%, PPV of 100%, NPV of 94% and accuracy of 97%), concluding that ARFI elastography is a promising technique for differentiating malignancy in these lesions.

Splenic tumors have clinical relevance in the clinical routine of small animals, presenting a malignancy prevalence of 58% when patients have masses or nodules larger than 1 cm 1 . These animals may show nonspecific clinical signs such as weakness, anorexia, and lethargy 2 . Among the most diagnosed malignant and benign alterations in the spleen are hemangiosarcoma with an occurrence of 44.1% and nodular lymphoid hyperplasia affecting 20.1% of the animals 3 . The prognosis varies from favorable to reserved depending on the diagnosis of the splenic lesion 4 , which can be performed using ultrasound-guided fine-needle aspirate cytology or histopathological analysis of material collected by biopsies or splenectomy [5][6][7] .
In medicine, accurate and rapid diagnosis is essential for patient prognosis and therapy 8 . Ultrasonography is a sensitive technique to detect subtle alterations or abnormalities that affect the splenic parenchyma in small animals 7 , as it allows evaluating the organ in terms of size, contours, echogenicity, echotexture and identifying focal or diffuse lesions 9 . It is difficult to differentiate between benign and malignant processes using B-mode, as malignant and benign splenic tumors may show similar patterns of echogenicity and echotexture while tumors with the same histopathological result may show different echo patterns 10 .
Other ultrasound techniques such as Doppler and contrast-enhanced ultrasound (CEUS) have been studied as adjuvants in the differentiation of splenic lesions. CEUS showed no significant difference between malignant and benign parenchymal lesions in dogs 11 as well as there was similarity in color Doppler and power Doppler between these lesions 12 . Thus, both methodologies, so far, cannot be used as predictors of malignancy in dog spleens.
Elastography provides complementary information to conventional ultrasound examinations, adding stiffness assessment as another measurable property. Among its modalities, the Acoustic Radiation Force Impulse (ARFI) is able to provide quantitative and qualitative information by generating shear waves 13  www.nature.com/scientificreports/ notoriety in veterinary medicine and medicine for demonstrating applicability in differentiating malignancy in thyroid 14 , cervical lymph nodes 15 , pancreas 16 and mammary tumors 17 . The quantitative evaluation of ARFI elastography is performed by measuring the shear wave velocity (SWV) and, through this technique, it has already been possible to differentiate benign from malignant lesions in different types of canine neoplasms (mammary, cutaneous and subcutaneous) and obtain cut-off values for this differentiation 17,18 . Also, in patients with breast cancer, it was possible to verify an increase in SWV in metastatic lymph nodes, differentiating from normal or inflamed tissues.
Believing that there are ultrasound and stiffness differences between malignant and benign splenic lesions in dogs, the present study aimed to evaluate the accuracy of B-mode ultrasound and ARFI elastography in detecting malignancy in splenic lesions in dogs. Our first hypothesis is that malignant tumors are more rigid than benign lesions, corroborating the results already found in other tissues. The second hypothesis is that benign tumors have greater rigidity when compared to normal splenic tissue.

Results
In the present study, 37 spleens were evaluated, of which 10 (27%) had diffuse lesions, 6 (16%) lesions in the cranial region, 9 (24%) in the medial region and 12 (33%) in the caudal region. Of these lesions, 16 (43%) were classified as benign and 21 (57%) as malignant. The histopathological diagnosis of the lesions is shown in Table 1.
The size of the spleen was similar in patients with benign and malignant lesions, and the height and length of the organ were not indicative of malignancy. Findings of echogenicity and echotexture were also similar between malignant and benign splenic alterations ( Table 2).
In the qualitative elastography evaluation, the spleen was classified as non-deformable in 35 (95%) and as deformable in 2 (5%) of the 37 examinations performed, however this classification is not related to the malignancy of the lesions (P = 0.465). Furthermore, there was a homogeneous stiffness pattern in 14 (88%) and heterogeneous in 2 (12%) of the 16 organs with benign lesions; in the 21 malignant lesions, the pattern was homogeneous in 16 (76%) and heterogeneous in 5 (24%). As with deformability, tissue stiffness homogeneity pattern was also not associated with malignancy (P = 0.2596).
The quantitative ARFI assessment of the different regions of the spleen was similar (P = 0.8856) between the cranial, caudal and medial portions and, in this way, the averages of each of these regions were calculated and used for the subsequent analysis, which were considered similar to each other in this general analysis (cranial 2.6 ± 0.8 m/s; medial 2.6 ± 0.9 m/s; caudal 2.5 ± 0.8 m/s; P = 0.9705).  www.nature.com/scientificreports/ Comparing the SWV from the lesion-free regions (2.1 ± 0.3 m/s) and the regions that presented focal or multifocal benign lesions (1.7 ± 0.5 m/s), it was observed that the free regions have a higher SWV (P = 0.0036). Focal or multifocal malignant lesions, on the other hand, had a SWV (3.4 ± 0.6 m/s) significantly greater than benign lesions and lesion-free areas (P < 0.0001).
In the analysis of the ability to predict malignancy through elastography, the anatomical regions in the organs that presented diffuse lesions were compared, and the medial region had the largest area under the curve (AUC = 0.9077) when compared to the cranial region (AUC = 0.8899) and the caudal region (AUC = 0.8661). For this reason, the SWV value of the medial area was used as the region of interest for the diagnosis of diffuse lesions. Table 3 contains the mean values of the comparative descriptive statistical analyzes and the results of the analysis of the ROC curves to differentiate malignant lesions in the studied spleens. When comparing the SWV of the localized lesion regions and the medial region of diffuse lesions, it was possible to observe that in the benign lesions (1.7 ± 0.5 m/s) this variable is significantly lower (P < 0.0001) when compared to malignant lesions (3.4 ± 0.5 m/s), the discriminative power analysis indicated that a SWV > 2.6 m/s in focal lesions or in the medial region in diffuse lesions, indicates malignancy of the splenic lesion with a sensitivity of 95%, specificity of 100%, positive predictive value of 100%, negative predictive value of 94%, and an accuracy of 97%. The Fig. 1 illustrates the ARFI elastographic assessment in a benign (splenic hematoma) and a malignant (lymphoma) lesion. The standard deviation of these SWVs, as well as the depth of assessment, were similar in the different types of lesions and without any predictive power.

Discussion
The present study provided important information regarding the prediction of malignancy in splenic lesions of dogs using ARFI elastography. Malignant lesions presented greater tissue stiffness when compared to benign lesions, with a shear velocity (SWV) greater than 2.6 m/s, with a diagnostic accuracy of 97%. These results are important to determine a more adequate treatment and prognosis, as well as enabling a more agile therapeutic approach, in addition to these data being able to direct studies to other species.
The spleen is a reticuloendothelial organ 20 and, due to its functional and anatomical characteristics, it is prone to develop neoplastic and non-neoplastic lesions 21 . Changes in the splenic parenchyma can be detected by conventional ultrasound 10 , however, this technique has low accuracy for differentiating malignancy from splenic changes 22 . In our study, there was no difference between the B-mode findings of malignant and benign lesions, which makes this method inconclusive, requiring more accurate techniques, such as cytology or histopathology or magnetic resonance 23 .
Ultrasound-guided aspiration puncture is considered invasive and complementary in splenic lesions, but reliable results for the diagnosis of abnormalities are often not obtained 7 . In contrast, histopathological examination is considered the gold standard for diagnosing these conditions as it is capable of providing accurate results 24 and, therefore, it was the method of choice for diagnosing splenic lesions in this study. However, the collection of material for histopathology is more invasive and, in many cases, the delay in obtaining the diagnosis can retard the initiation of adequate therapy and, consequently, worsen the patient's prognosis 24 .
ARFI elastography proved to be a non-invasive technique capable of differentiating malignancy in different types of splenic lesions, with excellent predictive values and higher accuracy than other tests such as magnetic resonance which has 94% accuracy for differentiation between malignant and benign splenic lesions in dogs 23,25 .
Even if no association of qualitative characteristics (deformity and homogeneity) with malignancy of splenic lesions was observed, it is known that these assessments are evaluator-dependent 26 and, consequently, are subject to different interpretations. However, in this study, it was possible to establish a significant difference in the shear velocity values of malignant and benign lesions, being an effective method to complement the study in B-mode, highly predictive of malignancy and that does not depend on the observer or the observations made and this so that it was possible to establish a cutoff value for the determination of malignancy (> 2.6 m/s).
The applicability of ARFI elastography has grown exponentially in recent years in medicine and veterinary medicine, so that this diagnostic modality can be used, in humans, as a predictor of malignancy in lymph node lesions 15,27 , in thyroid 28 and splenomegaly 29,30 . Through this technique, it was already possible to verify that, in cases of breast cancer in bitches, metastatic lymph nodes had a higher shear wave velocity than healthy lymph nodes 31 , which is a promising technique in veterinary oncology.
In a previous study 29 , it was possible to verify differences in stiffness between spleens of healthy human patients, with hepatoportal, myeloproliferative and infectious diseases, with the control group having the lowest Table 3. Mean and standard deviation of shear wave velocity (SWV) for predicting malignancy in splenic lesions of dogs. NA not applicable, CP cutoff point, Se sensibility, Sp specificity, PPV positive predictive value, NPV negative predictive value.  30 , demonstrating that the technique, in addition to being highly accurate for differentiating pathological processes, has good reproducibility of results. Even though studies demonstrating the applicability of ARFI elastography in the study of splenic lesions in veterinary medicine have not been published so far, it has already been possible to differentiate malignancy between different mammary tumors of bitches, where malignant lesions were more rigid by ARFI elastography, with a sensitivity of 94.7% and a specificity of 97.2% 17 . www.nature.com/scientificreports/ The increased stiffness observed in malignant splenic lesions in this study may be correlated with different pathological processes, such as type III collagen deposition, areas of fibrosis or microcalcifications 17 , while benign lesions, such as for example hematomas/clots, are basically formed by red blood cells and fibrin 32 , causing a decrease in this stiffness, making it even less rigid than in healthy tissue and free from injury.
The use of non-invasive tests without risk to patients and that allow for a differential diagnosis between malignant and benign lesions is of vital importance for animals and humans. Our results demonstrated that ARFI is useful in differentiating malignant and benign lesions in the spleen regardless of their location.
In conclusion, in this preliminary study the ARFI elastography made it possible to differentiate malignant and benign lesions in the spleen of dogs, and it was established that lesions with a shear wave velocity greater than 2.6 m/s are more likely to be malignant, with excellent predictive values. Thus, this imaging technique proved to be superior to conventional ultrasonography, which did not show satisfactory results for this differentiation. In B-mode, the spleen was evaluated in transverse and longitudinal sections along its entire length, aiming to perform a complete tissue scan to identify the presence of lesions. Splenic alterations were classified as focal when there was a nodule in the splenic parenchyma, multifocal when there was more than one delimited area of lesion, or diffuse when there was change in the entire splenic parenchyma. Regarding ultrasonographic features, echogenicity (anechoic, hypoechoic, hyperechoic or mixed), echotexture (homogeneous or heterogeneous), contours/borders (regular or irregular) and organ size were evaluated, in addition to the size of focal lesions (length, width and length/width ratio). In diffuse lesions, the types (multiple circumscribed or disseminated alteration), quantity (when quantifiable) and size of the lesions (length, width and length/width ratio) were determined.

Methods
After the B-mode evaluation, ARFI elastography (qualitative and quantitative) was performed using the Virtual Touch Tissue Imaging and Quantification Elastography method, with the same ultrasound device. Qualitative ARFI provided color images (elastogram) that were evaluated through visual deformability (deformable or non-deformable), according to the shades of color observed, with bluish areas indicating soft or deformable tissues (elastic or not very rigid) and reddish shades indicating hard or non-deformable (rigid) tissues. The software itself features image quality control in which homogeneous greenish images indicated high quality of the technique and yellowish and heterogeneous images as low-quality technique. When inappropriate images were obtained, the exam was repeated.
For quantitative ARFI evaluation, tissue elasticity was automatically measured using the shear wave velocity (SWV) of the regions of interest (ROIs) defined by the operator by placing a 2.5 mm 2 calipter on the elastogram images according to the following criteria: In spleens with one or more circumscribed lesions, at least three ROIs (those necessary to include the entire affected area) were selected from the abnormal area to obtain the mean SWV values, taking care to exclude vascular structures and cystic areas as described previously in evaluations of focal liver lesions in humans 32 . In addition, two ROIs were selected in each of the anatomical regions of the spleen (without apparent abnormalities): cranial extremity, body and caudal extremity to calculate the mean SWV of the apparently healthy tissue as already described in dogs 19 .
In spleens with diffuse lesions, a minimum of 12 ROIs were included, four in each of the previously defined areas of the splenic parenchyma (cranial, body and caudal), including the largest area of tissue and superficial, middle and deep subregions; for the calculation of the mean SWV of the spleen with diffuse lesion, taking care to exclude vascular structures, areas with a cystic, necrotic or calcified appearance as described by the consensus of the society of radiologists for the evaluation of diffuse liver lesions in humans 33 . Biopsy and splenic histopathological analysis. After ultrasound evaluation, tissue samples were collected, both from normal and altered areas, from all evaluated spleens. Ultrasound-guided incisional biopsy procedures with a trucut needle or surgical excisional biopsy under general anesthesia were performed, according to the protocol defined by the veterinarian responsible for the patient. The collected fragments were fixed in a 10% formalin solution, buffered with phosphate (pH 7.4) and processed until inclusion in paraffin and the prepared slides stained with hematoxylin/eosin. All samples were evaluated by the same experienced pathologist and were classified as malignant neoplastic lesion, benign neoplastic lesion or non-neoplastic lesion; focal or diffuse for each of the categories, totaling 6 types of classification.

Data availability
Statistical analysis was performed using the R software (R® Foundation for Statistical Computing, Vienna, Austria). Initially, the normal distribution (Shapiro-Wilk test) and the homoscedasticity of the variances (Barlett www.nature.com/scientificreports/