Characterization of Cutaneous and Subcutaneous Neoplasms in Canines and Malignancy Prediction Using B-Mode Ultrasonography, Doppler, and ARFI Elastography

Cutaneous and subcutaneous neoplasms are highly prevalent in dogs, ranging from benign to highly aggressive and metastatic lesions. The diagnosis is obtained through histopathology, however it is an invasive technique that may take a long time to obtain the result, delaying the beginning of the adequate treatment. Thus, there is a need for non-invasive tests that can help in the early diagnosis of this type of cancer. The aim of this study was to verify the accuracy of ultrasonography methods to predict malignancy in cutaneous and subcutaneous canine neoplasms. In addition, we aim to propose an ultrasonography evaluation protocol and perform the neoplasms characterization using these three proposed techniques. When and were proposed and characterization was done for all with at least

Based on the possibility of skin tumors malignancy prediction in canines using ultrasound techniques, this study aimed to evaluate cutaneous and subcutaneous neoplasms using B-mode, Doppler, and ARFI elastography, to determine the accuracy of ultrasonography methods, suggest an evaluation protocol for these neoplasms, and perform the ultrasonographic characterization of the speci c tumor types included.

Histopathologic results
A total of 130 cutaneous neoplasms (98 malignant and 32 benign) were evaluated, resulting in 21 histopathologic classi cations ( Table 1). The most prevalent malignant neoplasms in this study were the high-grade cutaneous mast cell tumors (18.46%). In comparison, the most prevalent benign neoplasms were lipomas (13.07%).

B-mode ultrasonography
In B-mode, measurements of length (3.02±2.85 cm), width (2.58±2.18 cm), and height (1.79±1.71 cm) were not associated with tumor malignancy, as well as echogenicity, capsule, and echotexture pattern (smooth or rough) ( Table 2). It was found that echotexture (P=0.007), invasiveness in adjacent tissues (P=0.002), hyperechogenic spots (P=0.031), and cavitary areas (P=0.001) were shown to be predictive characteristics of malignancy. This way, heterogeneous neoplasms with signs of invasiveness, presence of hyperechogenic spots, and cavitary areas were more likely to be malignant ( Figure 1). The predictive values of sensitivity, speci city, accuracy, PPV and NPV are shown in Table 2, and the ultrasonographic characterization of neoplasms is shown in Table 3. Se -sensitivity; Sp -speci city; Ac -accuracy; PPV -positive predictive value; NPV -negative predictive value. Although an association between vascularization intensity and malignancy was not observed (Table 4), only one benign neoplasm (in ltrative lipoma) presented intense vascularization. There were also no associations between tumor malignancy, location, and vascularization pattern.  Se -sensitivity; Sp -speci city; AUC -area under the curve.
Identi cation of arterial ow using pulsed Doppler was only possible in 51 neoplasms. Of these, 42 were malignant (82.35%) and only nine benign (17.65%). It was found that the peak values of systolic velocity, diastolic velocity, and resistivity index were not predictive of malignancy using the pulsed Doppler. However, the pulsatility index proved to be signi cant in differentiating between malignant and benign neoplasms (P=0.015), with a cut-off value >0.93 as indicative of malignancy, with 90.5% sensitivity, 55.6% speci city, and 75.7% accuracy ( Figure 2). The Doppler ultrasonographic characterization of cutaneous neoplasms is shown in Table 5. Table 5 Ultrasonographic characterization by Doppler (intensity, location and vascularization pattern, systolic peak, diastolic velocity, resistivity index, and pulsatility index) of cutaneous and subcutaneous canine neoplasms for tumor types that presented two or more cases. *Only one neoplasm with the arterial ow; SP -systolic peak; DV -diastolic velocity; RI -resistivity index; PI -pulsatility index; SD -Standard deviation; NA -not applicable.

ARFI Elastography
Both qualitative and quantitative assessments were shown to be signi cant in predicting malignancy (Table 6). Regarding deformability, it was observed that 11 benign and nine malignant neoplasms were classi ed as deformable, while 21 benign and 89 malignant were nondeformable. Deformability was shown to be predictive of tumor malignancy with 90.2% sensitivity, 35.48% speci city, 87.09% accuracy, 81.3% PPV, and 55% NPV.  It was observed that 85 neoplasms had at least four malignancy predictive characteristics (Table 8). Seventy-two neoplasms were indeed malignant, and only 13 were benign. Five or more malignancy predictive characteristics were found in sixty neoplasms, where 53 were malignant and seven were benign. Forty-ve neoplasms had at least six characteristics, where 41 were malignant, and four were benign. When considering all seven malignancy predictive characteristics, 16 neoplasms were computed, where 14 were malignant, and only two were benign.

Discussion
This study provides important information regarding the diagnosis and classi cation of cutaneous and subcutaneous canine neoplasms, as it was possible to establish malignancy predictive characteristics by all techniques used (B-mode, Doppler, and ARFI elastography). In addition, it was possible to determine an ultrasound examination protocol that could contribute to lesions diagnosis and prognosis and provide individual ultrasound characteristics for each studied tumor type. Because it is a complementary method, its characteristics are highly sensitive and have a positive predictive value. These results were obtained in all three ultrasound techniques that were performed.
Given the high number of cutaneous and subcutaneous neoplasm types, it should be considered that they have different structural components and biological behaviors. They can range from benign to highly aggressive and metastatic lesions [18], which justi es the moderate results observed. The authors would like to emphasize the importance of studies regarding speci c cancer types, as the present study results differed from previous canine mammary tumors ndings. In another study, with breast tumors, different characteristics and predictive values of malignancy were found [14].
There were no associations between malignancy and tumor measurements in this study, which can be explained by the fact that neoplasms were diagnosed at different stages. There were no associations with echogenicity, which may be related to the different pathological processes involved, such as active in ammation or tissue necrosis in different tumor types [19]. A preliminary study involving 42 cutaneous neoplasms showed an association between malignancy and hypoechogenicity [15]. A greater number of neoplasms and speci c types of skin cancer that were included in the present study may explain the discrepancy between the two studies.
The heterogeneous echotexture indicative of malignancy seen in cutaneous and subcutaneous tumors is explained by the different structural components, such as the presence of cavitary areas, points of brosis, or microcalci cations. The association between heterogeneous echotexture and malignancy was already demonstrated in previous studies with different types of neoplasms (cutaneous and mammary) in both humans and animals [13,[15][16][17]20].
It was possible to identify the signs of invasiveness in adjacent tissues because of their reactivity or the di cult tumors delimitation and then associate it with malignancy. This association is justi ed because malignant neoplasms tend to be more aggressive and invasive than benign ones, even requiring a greater safety margin when surgically removed [21].
On Doppler, it was not veri ed any qualitative characteristic with malignancy. It is known that tumor growth, both in malignant and benign lesions, is dependent on the blood supply [22]. Therefore, it is reasonable the fact that no signi cant results were obtained in neoplasm differentiation through these characteristics even though other researchers showed associations with malignancy in other tumor types, such as breast cancer in women and canine mast cell tumors [23,24].
Even though no vascularization points were observed in some tumors by color Doppler, the lack of vascularization should not be ruled out. It is known that the color Doppler technique has limitations at microvascular level and tissue perfusion, requiring other methods for diagnostic complementation, such as contrasted ultrasound [14]. Nevertheless, this technique was not available and could not be tested in the present study. This Doppler technique limitation contributed to the impossibility of evaluating all neoplasms by pulsed Doppler, with the Doppler velocimetry indices being calculated for only a portion of those who presented vascularization in color Doppler.
The lack of association between RI, systolic peak, and diastolic velocity with malignancy could be because it was only possible to identify the arterial ow in 9 benign neoplasms, predominantly in malignant lesions (82.35% of cases). However, a PI increase in malignant neoplasms was veri ed. The increase in this index has already been associated with malignancy in other types of lesions, such as ovarian and thyroid tumors in humans and metastases in canine lymph nodes. These may be related to the compressive effect tumor, the angiogenesis process, and the presence of arteriovenous shunts, which promote turbulent ows with high perfusion rates [25][26][27].
In the same way, as B-mode observed heterogeneity, the increased rigidity observed in malignant neoplasms can also be explained by tissue components they may present. In a previous study, greater stiffness was found in malignant mammary tumors in female dogs compared to benign ones, and this increase in stiffness was justi ed by the presence of areas of brosis, microcalci cations, and even collagen deposition [14].
The study of the rigidity of skin neoplasms in dogs has already been carried out qualitatively and semiquantitatively (through scores) through elastography, with greater rigidity being observed in malignant tissues, however no real quantitative values of the shear wave velocity were obtained. only subjective analysis [17]. On the other hand, this study provides more detailed information regarding neoplasms stiffness since it was possible to verify that an SWV greater than 3.52 m/s was indicative of malignancy. In addition, the elastography method used (ARFI method) allows more reliable results that are easy to perform, with greater reproducibility and less interobserver variability than sonoelastography [28].
Some benign neoplasms, such as adenomas, showed high tissue stiffness, justi ed by the accumulation of keratin and predominantly lymphoplasmacytic in ammatory in ltrate [29], that cause rigidity alterations in the keratinocytes and extracellular matrix [30,31].
Because ultrasonography is a complementary exam and should not be used alone to diagnose neoplasms, in this study, we demonstrate the importance of the association between the ndings of the different techniques performed. These have been already described for evaluating breast tumors in women, where an increase in accuracy was found when elastography and Doppler ndings were associated [24]. In our study, as we increased the number of malignancy predictive characteristics, there was a decrease in the number of false positives, increase in protocol speci city, and positive predictive value.
Among the study's limitations, it should be considered that some tumor types had a low experimental number, and as noted in this discussion and we had some values discrepancies (e.g., adenomas), which may be responsible for the low speci city and accuracy values that were observed.

Conclusions
Findings from this study indicate that ultrasonography has good applicability in the malignancy prediction of cutaneous and subcutaneous canine neoplasms through different techniques. It provides quick and noninvasive results and can be used as a complementary method for this diagnosis. Furthermore, we found that the assessment protocol by associating the ndings of different ultrasound techniques allows for greater reliability in diagnosing malignancy in this type of cancer.

Experimental design
This study was carried out according to the ARRIVE guidelines 2.0 (2020). Prospective data collection was conducted between September 2019 and June 2021. Sixty-six dogs of different breeds and ages (9.45±2.58 years) from the hospital routine presented cutaneous or subcutaneous neoplasms were enrolled in the study. The Veterinary Oncology sector previously evaluated all patients.

Ultrasound evaluation
Trichotomy of the tumor region was done with up to two centimeters of the peritumoral region. In order to maintain the patient's comfort during the examination and without sedation or anesthesia, patients were positioned in decubitus according to the anatomical location of the neoplasms. ACUSSON S2000™ equipment (Siemens®, Munich, Germany) was used for all the techniques performed, with a linear transducer and frequency ranging from 8 to 10Mhz. In addition, an ultrasonographic conductive gel was used throughout the examination.

B-mode ultrasound
The transducer was positioned in the central super cial region of the neoplasms, adjusting the focus, gain, and depth as needed. After adjusting the device, the nodules and masses were measured in longitudinal (length and height) and transversal (width) sections. The characteristics of echogenicity (hypoechogenic or hyperechogenic), echotexture (homogeneous or heterogeneous), echotexture pattern (coarse or smooth), invasiveness in adjacent tissues (presence or absence), capsule (presence or absence), cavitary areas (presence or absence), and hyperechogenic points (presence or absence) were evaluated.

Doppler
The color Doppler function was activated to identify neovascularization, and the pulse repetition frequency (PRF) was adjusted to 977 Hz.
When necessary, changes were made to the pre-established PRF. Tumor neovascularization was characterized according to its intensity (absent, mild, moderate, or intense), location (central, peripheral, or diffuse), and pattern (perinodular, mosaic, or network).
The pulsed wave Doppler was activated and used only for those neoplasms that presented vascularization at color Doppler examination. At this stage, the PRF used in the qualitative assessment was maintained, and the caliper was adjusted to cover 2/3 of the vessel's caliber, and using an angulation towards the vessel when necessary, respecting the limit of 60º degrees. At least three spectral traces were obtained [14] to get the peak values of systolic velocity (m/s), diastolic velocity (m/s), resistivity index (RI), and pulsatility index (PI).

ARFI Elastography
The elastographic evaluation was performed using the VTIQ method (virtual touch tissue imaging quanti cation, 2D-SWE technique). Color elastograms were performed in the qualitative study. Where blue colors represented more elastic areas, green and yellow represented intermediate stiffness, and red corresponded to more rigid areas. Based on the color pattern, neoplasms were classi ed according to their deformability (deformable or non-deformable). The same elastograms were used for quantitative analysis, and at least three areas of interest were selected. Those areas were randomly chosen to obtain the mean shear wave velocity (SWV -m/s), quanti ed by the VTIQ software, and using total stiffness as a representative value [14].

Histopathological evaluation
After ultrasound examinations, clinical care was continued, and biopsies (incisional or excisional) were performed to obtain the de nitive diagnosis. Patients were individually anesthetized, and surgical protocols were de ned under the recommendation of the responsible veterinarian. These tumor samples were xed in 10% formalin and sent to the veterinary pathology laboratory within the university, where histological cuts were performed to make slides stained with hematoxylin and eosin and, in cases of mast cell tumors, with toluidine blue.
After histopathological diagnosis, neoplasms were classi ed as benign or malignant, as established by the World Health Organization (WHO).

Statistical analysis
All data were analyzed using the SPSS Statistics 20 package (IBM®, New York, United States), and a signi cance level of 95% was used for all tests (P < 0.05). Echogenicity, echotexture, texture pattern, invasiveness, capsule, hyperechogenic spots, cavitary areas, and deformability were associated with malignancy using the Chi-square test, and sensitivity, speci city, accuracy, and positive (PPV) and negative (NPV) predictive values were calculated. Logistic regression was performed to differentiate malignancy according to the intensity, location, and pattern of vascularization. The other characteristics were submitted to the Kolmogorov-Smirnov normality test. The Mann-Whitney test was performed to analyze length, width, height, systolic peak, diastolic velocity, and pulsatility index. While for the resistivity index and SWV, a ttest was performed for independent samples. A ROC curve was used to obtain the cut-off point, sensitivity, speci city, and area under the curve for signi cant results.
Afterward, the variables with signi cant results were selected, and a descriptive analysis of the association between the different ultrasound techniques was performed. Furthermore, they were grouped into four groups: 1) presence of at least four predictive malignancy characteristics; 2) at least ve characteristics; 3) at least six characteristics; 4) seven characteristics. Thus, the chi-square test veri ed an association with malignancy, and the values of sensitivity, speci city, accuracy, PPV, and NPV were calculated. Additionally, descriptive analysis was performed and expressed in percentages of the qualitative ultrasonographic characteristics and the mean and standard deviation of the quantitative characteristics for each tumor type included in this study, except for single cases.

Declarations
Ethics approval and consent to participate This study was approved by the Animal Care and Use Committee of Universidade Estadual Paulista "Júlio de Mesquita Filho", Jaboticabal, Brazil (Protocol 010047/19) and the owners formally agreed, through signing a term of responsibility, to enroll their animals in this study. All

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. data.xlsx