The angioarchitecture and hemodynamics of lymph nodes are altered by pathologic processes. Blood vessel morphology in nodes affected by malignant neoplastic processes is usually deranged as internal nodal architecture is destroyed by neoplastic infiltration or because neovascularization occurs as a result of malignant angiogenesis.1,2 Inflammation causes dilation of intranodal vessels mediated by the release of humoral agents.1
The diagnostic value of nodal Doppler spectral waveform analysis in human medicine has been investigated, but its reliability in differentiating neoplastic from inflammatory changes in lymph nodes is still a matter of debate.1–9 In general, investigators have determined that pulsatility and RI are higher in nodes with neoplastic versus inflammatory changes1; however, because of the lack of a standardized technique, the optimal pulsatility and RI cutoff values for this distinction have not been widely accepted. It has been shown that analysis of the same data by use of different methods can lead to conflicting results.9
The results of the only study10 of nodal Doppler spectral waveform analysis in dogs of which we are aware confirm the findings in humans: compared with reactive nodes, pulsatility and RI are higher in nodes with malignant changes. From their data, those investigators proposed an RI of 0.68 and pulsatility of 1.49 as optimal cutoff values (giving equal weight to specificity and sensitivity) for the differentiation of malignant neoplastic and reactive lymphadenopathy in dogs.10 However, the technique used to select the analyzed spectral waveforms was not described in that report.10
The present study was based on the hypotheses that spectral Doppler analysis of intranodal blood flow can be used diagnostically to differentiate neoplastic from inflammatory peripheral lymphadenopathy and that derived Doppler indices differ depending on the manner in which the data are analyzed. The purpose of the study reported here was to evaluate pulsed-wave Doppler spectral parameters as a method for distinguishing between neoplastic and inflammatory peripheral lymphadenopathy in dogs and to propose calculation methods and index cutoff values to be used in clinical settings.
Materials and Methods
Animals—Client-owned dogs of any breed, age, and sex with 1 or more abnormally large superficial lymph nodes were considered suitable for inclusion in this prospective clinical study. The ultrasonographic and invasive procedures performed were part of the clinical management of these patients at the Department of Veterinary Clinical Sciences, University of Pisa. All owners were informed of the study and consented to participation of their dogs. Because procedures were considered routine, no approval was sought from an oversight committee.
Each dog was placed in lateral recumbency; the skin over the lymph nodes to be examined ultrasonographically was clipped, and coupling gel was applied. Examinations were performed by use of an ultrasound systema with a 6to 10-MHz linear-array transducer. All examinations were performed by the same ultrasonographer (DDS) who was unaware of the final diagnosis. Lymph nodes were included in the study if, besides being abnormally large, intranodal arterial blood flow was detectable and at least 3 distinct pulsed wave spectral tracings containing at least 3 consecutive cardiac cycles could be recorded during the ultrasonographic examination.
Representative pulsed-wave Doppler spectral waveform pattern recorded from a neoplasia-affected (metastasis of a synovial cell sarcoma) popliteal lymph node in a dog. The sites of measurements of PSV, DNV, and EDV are indicated. In this node, there is a high resistance pattern with a high S:D ratio (4.86) and RI (0.79) and a low N:S ratio (0.13); the D:N ratio is 1.52. The scale to the left of the image is in increments of 10 cm/s.
Citation: American Journal of Veterinary Research 69, 4; 10.2460/ajvr.69.4.478
Ultrasonographic evaluation—Large peripheral lymph nodes were examined by use of gray-scale sonography, color-flow Doppler, power Doppler, and spectral Doppler ultrasonography. Pulsed-wave Doppler signals were sampled under the guidance of color-flow Doppler or power Doppler mapping. The largest longitudinal and transverse diameters of the studied lymph node on the same image were recorded. For each lymph node, 3 areas were examined and 3 spectral tracings (1 from each area) were recorded. Each spectral tracing had to include ≥ 3 consecutive cardiac cycles. A sample volume of 2 mm was centered in the vessel, and the angle of insonation was maintained at < 60°. During the examination, the probe was gently positioned on the skin surface to avoid compression of the lymph node and artificial increase in vascular resistance. Spectral Doppler images were electronically saved in a JPEG (Joint Photographic Experts Group) format on a workstation and measurements were performed manually by the operator with an image processing programb at the end of the examination.
Doppler ultrasonographic parameters—For each cardiac cycle, PSV, EDV, and DNV (defined as the maximal velocity at the deepest point in the early diastolic notch of the arterial waveform11) were recorded (Figures 1 and 2). On the basis of these measurements, Doppler ultrasonographic indices were calculated as follows:
Representative pulsed-wave Doppler spectral waveform pattern recorded from a superficial inguinal lymph node with inflammatory changes in a dog. In this node, EDV is relatively high and there is a low resistance pattern (S:D ratio, 2.54; RI, 0.61; N:S ratio, 0.47; and D:N ratio, 0.83). The scale to the left of the image is in increments of 10 cm/s.
Citation: American Journal of Veterinary Research 69, 4; 10.2460/ajvr.69.4.478
Doppler indices relative to each spectral tracing were obtained by averaging the indices (calculated by use of the aforementioned formulas) of 3 consecutive cardiac cycles.
After the ultrasonographic examination, fine needle aspirates or excisional biopsy specimens of the examined lymph nodes were obtained; lymphadenopathy was classified as either inflammatory or neoplastic (lymphomatous or metastatic) via cytologic or histologic examination.
Values of the Doppler indices for statistical analysis were then calculated (Appendix). The final dataset for each lymph node was composed of 4 sets (each corresponding to a Doppler ultrasonographic index) of 10 values each (corresponding to the different calculation methods).
For statistical purposes, only 1 node in each dog was selected for the receiver operating characteristic analysis (a so-called 1 patient–1 node approach). In dogs in which ≥ 1 lymph node was examined, a popliteal lymph node was selected as a representative node for that dog; if both popliteal lymph nodes were examined, the largest was chosen. Receiver operating characteristic analysis was used to determine sensitivity, specificity, PPV, NPV, and diagnostic accuracy of the Doppler indices, compared with the cytologic or histopathologic classification of the lymphadenopathy as neoplastic or inflammatory. Optimal cutoff values and calculation methods were selected to obtain the overall highest diagnostic accuracy (minimal false-negative and false-positive results) and the maximal specificity at 100% sensitivity. Values were approximated to the second decimal place. Sensitivity, specificity, PPV, and NPV of each obtained cutoff value were determined in the whole population (n = 40) employing a ratio estimator method.12,13 Diagnostical accuracy was also calculated. Descriptive statistical analysis of the evaluated indices that provided the highest diagnostic accuracies was performed. Statistical analysis was performed with commercially available software.c
Table 1—Spectral pulsed-wave Doppler indices (derived from the first recorded spectral tracing [calculation method 1]) determined for a sample population of 40 abnormally large lymph nodes categorized on the basis of neoplastic versus inflammatory changes in 33 dogs.
| Index | Classification of lymphadenopathy | Median (range) | Mean ± SD |
|---|---|---|---|
| Rl | Neoplastic | 0.75(0.62–0.91) | 0.76 ± 0.08 |
| Inflammatory | 0.58 (0.44–0.69) | 0.58 ± 0.09 | |
| S:D ratio | Neoplastic | 4.07(2.67–11.17) | 5.09 ± 2.45 |
| Inflammatory | 2.39(1.80–3.22) | 2.50 ± 0.52 | |
| N:S ratio | Neoplastic | 0.24(0.12–0.45) | 0.26 ± 0.10 |
| Inflammatory | 0.48 (0.31–0.66) | 0.47 ± 0.11 | |
| D:N ratio | Neoplastic | 0.89(0.48–1.43) | 0.95 ± 0.26 |
| Inflammatory | 0.89(0.65–1.00) | 0.87 ± 0.12 |
Results
Thirty-nine dogs with at least 1 abnormally large superficial lymph node were examined via ultrasonography; 6 dogs were excluded from the study because 3 Doppler tracings could not be recorded. Forty lymph nodes from 33 dogs (age range, 18 to 204 months; mean age, 91.5 months) were included in the study. In all the dogs of the present study, the 3 tracings were recorded in < 10 minutes.
Among the lymph nodes examined, there were 1 right mandibular, 1 left mandibular, 2 right superficial cervical, 3 left superficial cervical, 1 right inguinal, 3 left inguinal, 18 right popliteal, and 11 left popliteal nodes. Lymphadenopathy was classified as inflammatory in 17 of 40 nodes and neoplastic in 23 of 40 nodes. Among the 23 lymph nodes with evidence of neoplasia, 5 were considered to represent sites of metastasis (2 mast cell tumors, 2 melanomas, and 1 synovial cell sarcoma); 18 nodes were diagnosed as lymphoma. In 6 dogs (all affected with lymphoma), multiple lymph nodes were examined (2 lymph nodes in 5 dogs and 3 lymph nodes in 1 dog). The spectral Doppler indices (derived via calculation method 1) of all 40 abnormally large lymph nodes were assessed on the basis of neoplastic versus inflammatory categorization (Table 1).
Table 2—Diagnostic yield of spectral Doppler waveform analysis of intranodal blood flow in 33 abnormally large superficial lymph nodes in 33 dogs determined by use of a 1 patient–1 node approach. The endpoint was neoplastic disease; the cutoff value with the overall highest diagnostic accuracy and corresponding diagnostic yield is indicated by an asterisk.
| Cutoff value (sensitivity [%]; specificity [%]; PPV [%]; NPV [%]) | ||||
|---|---|---|---|---|
| Calculation method | RI | S:D ratio | N:S ratio | D:N ratio |
| 1 | 0.69 | 3.22* | 0.45 | 0.87 |
| (88; 100; 100; 89) | (94; 100; 100; 94) | (100; 71; 80; 100) | (56; 64; 64; 56) | |
| 2 | 0.66 | 3.00 | 0.33 | 1.00 |
| (88; 94; 94; 88) | (88; 94; 94; 88) | (75; 100; 100; 78) | (25; 93; 80; 52) | |
| 3 | 0.69 | 3.22 | 0.29 | 0.87 |
| (88; 94; 94; 88) | (94; 94; 94; 94) | (75; 93; 92; 76) | (56; 71; 70; 59) | |
| 4 | 0.65 | 2.88 | 0.39 | 1.04 |
| (82; 94; 93; 83) | (82; 94; 93; 83) | (94; 79; 83; 92) | (44; 94; 87; 59) | |
| 5 | 0.69 | 3.22 | 0.31 | 1.04 |
| (70; 94; 92; 75) | (76; 94; 93; 79) | (69; 100; 100; 74) | (37; 93; 86; 56) | |
| 6 | 0.69 | 3.27 | 0.34 | 1.03 |
| (88; 94; 94; 88) | (87; 94; 93; 88) | (74; 100; 100; 78) | (25; 100; 100; 54) | |
| 7 | 0.68 | 3.16 | 0.31 | 1.00 |
| (88; 94; 94; 88) | (88; 94; 94; 88) | (69; 100; 100; 74) | (19; 100; 100; 52) | |
| 8 | 0.64 | 3.04 | 0.43 | 0.8 |
| (82; 94; 93; 83) | (88; 87; 88; 87) | (100; 64; 76; 100) | (87; 29; 58; 67) | |
| 9 | 0.66 | 2.98 | 0.33 | 0.99 |
| (88; 94; 94; 88) | (88; 94; 94; 88) | (69; 100; 100; 74) | (37; 79; 67; 52) | |
| 10 | 0.67 | 3.08 | 0.32 | 1.00 |
| (82; 94; 93; 83) | (87; 94; 93; 88) | (75; 100; 100; 80) | (25; 100; 100; 54) |
Table 3—Diagnostic yield of spectral Doppler waveform analysis of intranodal blood flow in 33 abnormally large superficial lymph nodes in 33 dogs determined after maximization of sensitivity by use of a 1 patient–1 node approach. The endpoint was neoplastic disease; the cutoff value with the overall highest diagnostic accuracy and corresponding diagnostic yield is indicated by an asterisk.
| Cutoff value (sensitivity [%]; specificity [%]; PPV [%]; NPV [%]) | ||||
|---|---|---|---|---|
| Calculation method | RI | S:D ratio | N:S ratio | D:N ratio |
| 1 | 0.62 | 2.62 | 0.45* | 1.38 |
| (100; 62; 74; 100) | (100; 62; 74; 100) | (100; 71; 80; 100) | (100;0; 53; ND) | |
| 2 | 0.60 | 2.48 | 0.42 | 0.47 |
| (100; 62; 74; 100) | (100; 62; 74; 100) | (100; 64; 76; 100) | (100;0;53;ND) | |
| 3 | 0.62 | 2.62 | 0.41 | 1.34 |
| (100; 56; 71; 100) | (100; 56; 70; 100) | (100; 64; 76; 100) | (100;0; 53; ND) | |
| 4 | 0.54 | 2.18 | 0.46 | 0.55 |
| (100; 56; 71; 100) | (100; 56; 71; 100) | (100; 71; 80; 100) | (100;0; 53; ND) | |
| 5 | 0.58 | 2.39 | 0.45 | 0.39 |
| (100; 37; 63; 100) | (100; 37; 63; 100) | (100; 57; 73; 100) | (100; 7; 55; 100) | |
| 6 | 0.60 | 2.52 | 0.39 | 0.39 |
| (100; 43; 65; 100) | (100; 53; 64; 100) | (100; 64; 76; 100) | (100; 7; 55; 100) | |
| 7 | 0.60 | 2.53 | 0.42 | 0.44 |
| (100; 69; 77; 100) | (100; 62; 74; 100) | (100; 64; 76; 100) | (100;0; 53; ND) | |
| 8 | 0.60 | 2.49 | 0.43 | 0.44 |
| (100; 62; 74; 100) | (100; 64; 76; 100) | (100;0; 53; ND) | ||
| 9 | 0.59 | 2.43 | 0.42 | 0.51 |
| (100; 62; 74; 100) | (100; 62; 74; 100) | (100; 64; 76; 100) | (100;0;53;ND) | |
| 10 | 0.60 | 2.53 | 0.40 | 0.43 |
| (100; 62; 74; 100) | (100; 62; 73; 100) | (100; 71; 80; 100) | (100;0; 53; ND) |
ND = Not determined.
Table 4—Diagnostic yield of spectral Doppler waveform analysis of intranodal blood flow in 40 abnormally large superficial lymph nodes in 33 dogs (2 lymph nodes in 5 dogs and 3 lymph nodes in 1 dog were evaluated; 1 node was examined in all other dogs). The endpoint was neoplastic disease. Data represent the most accurate cutoff value and that with highest specificity at 100% sensitivity as calculated from the first recorded tracing for each Doppler index (only 1 value is reported for N:S ratio because both selection criteria yielded the same cutoff value).
| Index | Cutoff value | Sensitivity (% [95% CI]) | Specificity (% [95% CI]) | PPV %[95%CI])( | NPV %[95% CI]) | Diagnostic accuracy (%) |
|---|---|---|---|---|---|---|
| RI | 0.69 | 83(68–97) | 94(83–100) | 95(85–100) | 80(62–98) | 87.5 |
| 0.62 | 100 (NA) | 59(35–83) | 82(68–97) | 100 (NA) | 82.5 | |
| S:D ratio | 3.22 | 91 (80–100) | 100 (NA) | 100 (NA) | 89(75–100) | 95.0 |
| 2.62 | 100 (NA) | 47(23–71) | 79 (62–95) | 100 (NA) | 77.5 | |
| N:S ratio | 0.45 | 100 (NA) | 67(42–91) | 81 (66–97) | 100 (NA) | 86.5 |
| D:N ratio | 0.87 | 50(33–67) | 67(42–91) | 69(45–92) | 48(24–71) | 59.5 |
| 1.38 | 100 (NA) | 0(NA) | 0(NA) | 56(34–77) | 59.5 |
CI = Confidence interval. NA = Not applicable.
The calculation methods yielding the highest overall diagnostic accuracy and the highest specificity at 100% sensitivity via the 1 patient–1 node approach underwent receiver operating characteristic analysis. The highest diagnostic accuracy yielded by RI (91%) was obtained for a cutoff value of 0.69 by use of only the first recorded tracing (Table 2). At 100% sensitivity, the highest diagnostic accuracy (82%) was obtained as the mean of the RI values from the 2 tracings with the highest RIs and by use of a cutoff value of 0.60 (Table 3).
The S:D ratio yielded the highest diagnostic accuracy (97%) when only the first recorded tracing was analyzed (cutoff value, 3.22; Table 2). Multiple calculation methods (methods 1 [first recorded tracing], 2 [mean S:D ratio derived from all recorded tracings], 4 [lowest RI], 5 [fastest systolic velocity], 6 [lowest diastolic velocity], and 7 [from the 2 tracings with the highest RIs]) yielded the highest diagnostic accuracy (76%) after maximization of sensitivity (Table 3). The corresponding cutoff values were 2.62, 2.48, 2.53, 2.49, 2.43, and 2.53, respectively.
The N:S ratio yielded the highest diagnostic accuracy (87%) when the mean value from all recorded tracings (cutoff value, 0.33) or from the 2 tracings with the lowest diastolic velocities (cutoff value, 0.32) was calculated (Table 2). The highest diagnostic accuracy with 100% sensitivity was obtained when the N:S ratio was calculated from the first recorded tracing (cutoff value, 0.45), from the tracing with the lowest RI (0.46), or as a mean value from the 2 tracings with the lowest diastolic velocities (0.40; Table 3). In all instances, diagnostic accuracy was 87%.
Dot plot of S:D and N:S ratio values for 40 large superficial lymph nodes affected with neoplasia (N) or inflammation (I) in 33 dogs (2 lymph nodes in 5 dogs and 3 lymph nodes in 1 dog were evaluated; 1 node was examined in all other dogs). The horizontal lines indicate the suggested cutoff values to be used to obtain the highest diagnostic accuracy (S:D ratio, 3.22; N:S ratio, 0.45).
Citation: American Journal of Veterinary Research 69, 4; 10.2460/ajvr.69.4.478
The D:N ratio yielded the highest diagnostic accuracy (67%) when calculated from the tracing with the lowest diastolic velocity (cutoff value, 1.03) or as a mean value from the 2 tracings with the lowest diastolic velocities (1.00; Table 2). At 100% sensitivity, the highest diagnostic accuracy (57%) was obtained on calculation of the D:N ratio from either the tracing with the fastest systolic velocity (cutoff value, 0.39) or lowest diastolic velocity (0.39; Table 3). All other calculation methods yielded a specificity of 0%.
The S:D ratio calculated from the first recorded tracing (cutoff value, 3.22) was the most accurate index and was selected for further analysis. At 100% sensitivity, the N:S ratio calculated from the first recorded tracing (cutoff value, 0.45) was selected for further analysis; several other calculation methods yielded the same specificity but because they required the analysis of 3 tracings and were therefore more time consuming, the latter were deemed less suitable to clinical use and were excluded from further analysis. The diagnostic yield of the selected cutoff values in the population of 40 abnormally large lymph nodes was also assessed to determine their accuracy in a clinical setting (Table 4; Figure 3).
Discussion
Lymph nodes may become abnormally large because of inflammation, primary neoplastic processes, or metastatic involvement. Multiple ultrasonographic techniques including B-mode imaging,14,15,d spectral Doppler analysis of intranodal blood flow,1–9,16 color and power Doppler ultrasonography (either fundamental or contrast-enhanced),17–20 and harmonic ultrasonography21–24 have been evaluated in humans as a method to determine the etiology of lymphadenopathy. In veterinary medicine, color and power Doppler ultrasonography,10,25 spectral waveform analysis,10 and contrast-enhanced harmonic and power Doppler ultrasonography26 have been used to characterize the angioarchitecture and hemodynamics of lymph nodes.
In humans with cervical lymphadenopathy, the lymph nodes affected with malignant changes typically have a higher resistance flow pattern (with higher pulsatility and RI) than nodes affected with non-neoplastic changes.1 In 1 study,4 resistance was low in nodes with neoplasia, but that finding was not corroborated in later investigations. Lower resistance in reactive nodes has been attributed to increased hilar perfusion and vasodilation, whereas the higher values in nodes affected with neoplastic changes could be a result of distortion and compression of nodal vessels by tumor cells and by inadequately successful tumor-induced angiogenesis.8,9,27 Compared with lymphomatous nodes, significantly higher Doppler indices in nodes affected by metastases have been reported by some authors.2,3,28 This finding can be explained by the fact that the normal intranodal architecture in lymphomas has not been damaged by the growth of neoplastic tissue, which is lodged in a mainly intact nodal fibroepithelial skeleton.1 In contrast, metastatic involvement results in distortion of the intranodal architectural network, along with damage or compression of blood vessels, leading to a considerable increase in resistance to blood flow.2,5
The higher resistance blood flow pattern in lymph nodes affected by neoplasia in humans has been partially confirmed in dogs by Nyman et al,10 who evaluated, among other ultrasonographic findings, the RI and pulsatility in a population of canine lymph nodes. For pulsatility, a significant difference was detected among apparently normal (1.06) and reactive (1.07) nodes, lymphomatous nodes, (mean pulsatility, 1.27), and nodes with metastasis (mean pulsatility, 1.72).10 Resistive index was significantly higher in the nodes with neoplastic changes, compared with the other lymph nodes; this finding was attributed to the difference between the metastatic and the other groups of nodes (normal, reactive, and lymphomatous).10 The results of the present study, in which lymph nodes with neoplasia had significantly higher RI and S:D ratio values, compared with inflammatory lymph nodes, support the suggestion that there is a higher resistance blood flow pattern in nodes with neoplastic changes. An attempt to test the differences between lymphomatous lymph nodes and those with metastasis was not done in the present study because of the low number of metastasisaffected nodes (n = 5). However, because of the overlap of the Doppler indices between lymphomatous and metastasis-affected lymph nodes, our results seem to suggest that spectral Doppler analysis is not a reliable diagnostic method to distinguish between them.
The main goal of many studies is to determine cutoff levels of resistivity and pulsatility indices in lymph nodes that are affected by neoplasia or inflammation.1–9,16 Unfortunately, the methods for calculation of velocity indices reported in the human medical literature are inconsistent.9 They include calculation of the indices from the highest velocity vessel3; data collection from 8 sites and calculating the indices from the 3 vessels with highest velocity8 or highest RIs9; selection of the tracing with the highest RI,2,7 lowest RI,4 or the fastest or next fastest arterial signal5,6 for analysis; or data collection from 3 sites from which the mean indices are calculated.16 In the report by Nyman et al,10 the method used for determination of RI and pulsatility values in dogs is not stated, making comparison of their results with those of the present study difficult. Nevertheless the RI cutoff value proposed in the aforementioned study10 (0.68) is exceedingly close to the one determined in the present study (0.69).
In a series of lymph nodes from dogs, the combination of pulsatility with the size and the distribution of vascular flow within the lymph node proved to be the most accurate ultrasonographic parameter in discriminating between neoplasia and inflammation in lymph nodes (classification error of 11%10). Although pulsatility was not analyzed in the present study, the diagnostic accuracy yielded by the S:D ratio (5% classification error because of 2 false-negative results) was comparable to the accuracy yielded by pulsatility index in the study by Nyman et al.10
In the present study, comparison of the diagnostic accuracy provided by S:D ratio and RI values via the different calculation methods revealed a slightly higher classification error in RI for most of the calculation methods; both the RI and N:S ratio, had an overall diagnostic accuracy that was slightly lower than the S:D ratio. The N:S ratio proved to be the most accurate index when sensitivity was maximized; a cutoff value of 0.45 (calculated from the first recorded tracing) yielded a PPV of 81% and a classification error of 13.5% (5 false-positive results), compared with a classification error of 17.5% for the RI, 22.5% for S:D ratio, and 40.5% for D:N ratio. Despite the high variability of diastolic notch velocity and detection of several patterns of D:N ratios, this index was not correlated with the type of lymphadenopathy; consequently, the D:N ratio was considered of no diagnostic value in the differentiation of nodes with inflammatory changes from those with neoplastic changes.
With regard to lymph nodes, different methods of Doppler spectral parameter calculation provide conflicting results from the same set of nodes.9 This finding supports the need for a standardized method and underscores the importance of applying the exact method of analysis used in a reference technique in other studies if the latter data are to be applied to a specific population for comparison.9 Another difficulty is that lymph nodes that are only partly replaced by neoplastic tissue may be classified in 3 possible categories of vascularity.9 Metastasis within lymph nodes has the hemodynamic profile associated with malignant neovascularization, which typically results in higher than normal resistance.9 The part most remote from the metastasis could even have apparently normal flow with normal hemodynamics.9 A third pattern, characterized by lower than normal resistance, could be evident in any reactive part of the node.2 Given these possibilities, it can be appreciated that on the basis of data collection from only 1 area of an abnormal lymph node, the vessels with characteristic hemodynamics could be undetected.9 Furthermore, examination of multiple vessels and calculation of the mean value from all selected samples is likely to blur significant discriminatory values.9 Consequently, it would be theoretically indicated to examine multiple vessels to decrease the risk of overlooking the ones with the high-resistance features of neoplasia.
In agreement with the results of the study in humans by Ho et al,9 examination of different vessels in the same lymph node in the dogs of the present study resulted in different values of each index; also, the use of different calculation methods resulted in the selection of different cutoff values that were characterized by variable diagnostic accuracies. However, our results do not support the need to examine multiple vessels; among the analyses of single indices, the calculation method that provided the lowest classification error was that involving the first recorded spectral tracing. Nevertheless, it cannot be completely excluded that, at 100% sensitivity, a minimally increased diagnostic accuracy could have been obtained by selecting the tracing with the lowest diastolic velocity or calculating the mean value from the 2 tracings with the lowest diastolic velocities. The reasons why the first recorded tracing yielded the best diagnostic accuracy are speculative. In most instances, the first tracing represented the largest intranodal arterial vessel or the vessel with the most prominent blood flow; these features made them the easiest to detect and examine. Compared with smaller vessels or vessels with less prominent blood flow, data collection from larger vessels or those with prominent blood flow might have been more rewarding because of a lower incidence of operator error. The lack of a clear increase in diagnostic accuracy yielded by examination of multiple areas of the same node might also be associated with the low number of lymph nodes affected with metastases in the sample population of the present study and with a relatively uniform involvement of most lymph nodes by the disease process (either lymphoma or inflammation). Furthermore, it cannot be excluded that metastatic lymphadenopathy in dogs is characterized by more diffuse nodal infiltration than it is in humans. Because of the small sample size in the present study, it cannot be stated with absolute certainty that the selected cutoff values and calculation methods would be the most accurate for use in a larger population. Because the analysis of > 1 spectral tracing did not increase diagnostic accuracy, recording of a single tracing in a clinical setting may be adequate but only if future investigations involving a considerably higher number of metastasis-affected lymph nodes confirm the results of our study.
Findings of lymph node ultrasonography, even if supported by data derived by use of Doppler techniques, do not replace a cytologic or histologic diagnosis of affected lymph nodes. In our opinion, the N:S ratio is the index most suitable to clinical use, and an N:S ratio value > 0.45 should prompt tissue sample collection to determine whether an individual lymph node is affected with neoplastic changes. The use of maximally sensitive cutoff values as biopsy selection criteria is driven by the need to avoid false-negative diagnoses, which could possibly delay further diagnostic assessments and treatment. A major drawback associated with use of cutoff values that are characterized by a relatively low specificity (67%) is the resultant increased number of false-positive ultrasonographic results and, consequently, negative results of biopsy specimen examination.
Doppler spectral tracings from most lymph nodes can be obtained without the aid of contrast medium administration, making the procedure safe and inexpensive to perform. In most instances, the Doppler ultrasonographic examination is relatively quick to perform (in all the dogs of the present study, the 3 tracings were recorded in < 10 minutes). Nevertheless, 15% of dogs were excluded from the study population because 3 Doppler tracings could not be recorded as a result of motion artifacts that lead to poor image quality or Doppler signal. However, the analysis of a single spectral tracing, as supported by the results of our study, appears to reduce examination time and probably also minimizes the number of technically inadequate examinations. The criteria used in the selection of the tracings used for calculation of the Doppler indices strongly influence the results of waveform analysis.9 Appropriate clinical use of the provided cutoff values relies on the assumption that spectral Doppler indices are calculated according to the calculation method applied in the present study.
Because all ultrasonographic examinations were performed by the same operator, interobserver variability was excluded. Although the examiner was not aware of the final diagnosis at the moment of the examination, the study might be biased to some extent by that person's knowledge of clinical and B-mode ultrasonographic findings, which can never be avoided.
The results of the present study suggest that spectral waveform analysis of intranodal blood flow may be a reliable technique to distinguish neoplastic from inflammatory peripheral lymphadenopathy in dogs. Determination of Doppler indices from 1 spectral tracing is at least as accurate as calculating those indices from more tracings. The single parameter with the overall highest diagnostic accuracy was the S:D ratio (diagnostic accuracy, 95%; PPV, 100%; and NPV, 89%). In dogs, lymph nodes with an N:S ratio > 0.45 (diagnostic accuracy, 86%; PPV, 81%; and NPV, 100%) should be considered highly suspicious for neoplasia and sample collection of node tissue for diagnostic assessment should be undertaken. Further studies are necessary to determine whether other applications of spectral waveform analysis might benefit veterinary medicine. Assessment of the response to chemoor radiotherapy and noninvasive evaluation of centrally located nodes are examples of possible applications requiring further investigation.
ABBREVIATIONS
| RI | Resistive index |
| PSV | Peak systolic velocity |
| EDV | End diastolic velocity |
| DNV | Diastolic notch velocity |
| S:D | Peak systolic velocity-to-end diastolic velocity |
| N:S | Diastolic notch velocity-to-peak systolic velocity |
| D:N | End diastolic velocity-to-diastolic notch velocity |
| PPV | Positive predictive value |
| NPV | Negative predictive value |
Toshiba Core Vision Pro SSA-350, Toshiba Medical System s.r.l., Rome, Italy.
ImageJ, US National Institute of Health, Bethesda, Md.
Number Cruncher statistical software 2004, NCSS, Kaysville, Utah.
Solbiati L, Rizzatto G, Bellotti E, et al. High resolution sonography of cervical lymph nodes in head and neck cancer: criteria for differentiation of reactive versus malignant lymph nodes (abstr). Radiology 1988;169;113.
References
- 1.↑
Brnic Z, Hebrang A. Usefulness of Doppler waveform analysis in differential diagnosis of cervical lymphadenopathy. Eur Radiol 2003;13:175–180.
- 2.↑
Na DG, Lim HK, Byun HS, et al. Differential diagnosis of cervical lymphadenopathy: usefulness of colour Doppler sonography. Am J Roentegenol 1997;168:1311–1316.
- 3.↑
Steinkamp HJ, Maurer J, Cornehl M, et al. Recurrent cervicallymphadenopathy: differential diagnosis with colour duplex sonography. Eur Arch Otorhinolaryngol 1994;251:404–409.
- 4.↑
Chang DB, Yuan A, Yu CJ, et al. Differentiation of benign and malignant cervical lymph nodes with colour Doppler sonography. Am J Roentegenol 1994;162:965–968.
- 5.
Choi MY, Lee JW, Jang KJ. Distinction between benign and malignant causes of cervical, axillary and inguinal lymphadenopathy: value of Doppler spectral waveform analysis. Am J Roentegenol 1995;165:981–984.
- 6.
Adibelli ZH, Ünal G, Gül E, et al. Differentiation of benign and malignant cervical lymph nodes: value of B-mode and color Doppler sonography. Eur J Radiol 1998;28:230–234.
- 7.
Wu CH, Chang YL, Hsu WC, et al. Usefulness of Doppler spectral analysis and power Doppler sonography in the differentiation of cervical lymphadenopathies. Am J Roentegenol 1998;171:503–509.
- 8.↑
Ahuja AT, Ho SS, Leung SF, et al. Metastatic adenopathy from nasopharyngeal carcinoma: successful response to radiation therapy assessed by colour duplex sonography. AJNR Am J Neuroradiol 1999;20:151–156.
- 9.↑
Ho SS, Metreweli C, Ahuja AT. Does anybody know how we should measure Doppler parameters in lymph nodes? Clin Radiol 2001;56:124–126.
- 10.↑
Nyman HT, Kristensen AT, Skovgaard IM, et al. Characterization of normal and abnormal canine superficial lymph nodes using gray-scale B-mode, color flow mapping, power, and spectral Doppler ultrasonography: a multivariate study. Vet Radiol Ultrasound 2005;46:404–410.
- 11.↑
Evans DH, McDicken WN. Waveform analysis and pattern recognition. In: Evans DH, McDicken WN, eds. Doppler ultrasound: physics, instrumentation and signal processing. 2nd ed. Chichester, West Sussex, England: Wiley, 2000;204–207.
- 12.
Rao JNK, Scott AJ. A simple method for the analysis of clustered binary data. Biometrics 1992;48:577–585.
- 13.
Genç Y, Gökmen D, Tüccar E, et al. Estimation of sensitivity and specificity for clustered data. Turk J Med Sci 2005;35:21–24.
- 14.
Rubaltelli L, Proto E, Salmaso R, et al. Sonography of abnormal lymph nodes in vitro: correlation of sonographic and histologic findings. Am J Roentgenol 1990;155:1241–1244.
- 15.
Vassallo P, Wernecke K, Roos N, et al. Differentiation of benign from malignant superficial lymphadenopathy; the role of high-resolution US. Radiology 1992;183:215–220.
- 16.↑
Ying M, Ahuja A, Brook F. Accuracy of sonographic vascular features in differentiating different causes of cervical lymphadenopathy. Ultrasound Med Biol 2004;30:441–447.
- 17.
Tschammler A, Ott G, Schang T, et al. Lymphadenopathy: differentiation of benign from malignant disease—color Doppler US assessment of intranodal angioarchitecture. Radiology 1998;208:117–123.
- 18.
Moritz JD, Ludwig A, Oestmann JW. Contrast-enhanced color Doppler sonography for evaluation of enlarged cervical lymph nodes in head and neck tumors. Am J Roentgenol 2000;174:1279–1284.
- 19.
Tschammler A, Beer M, Hahn D. Differential diagnosis of lymphadenopathy: power Doppler vs color Doppler sonography. Eur Radiol 2002;12:1794–1799.
- 20.
Ahuja A, Ying M. Sonographic evaluation of cervical lymphadenopathy: is power Doppler sonography routinely indicated? Ultrasound Med Biol 2003;29:353–359.
- 21.
Dybiec E, Brodzisz A, Pietka M, et al. The application of ultrasound contrast, 3D imaging and tissue harmonic imaging in the differential diagnosis of lymph node enlargement in children. Ann Univ Mariae Curie Sklodowska [Med] 2002;57:131–142.
- 22.
Kubota K, Hisa N, Ogawa Y, et al. Evaluation of tissue harmonic imaging for breast tumors and axillary lymph nodes. Oncol Rep 2002;9:1335–1338.
- 23.
Kubota K, Ogawa Y, Nishigawa T, et al. Tissue harmonic imaging sonography of the axillary lymph nodes: evaluation of response to neoadjuvant chemotherapy in breast cancer patients. Oncol Rep 2003;10:1911–1914.
- 24.
Rubaltelli L, Khadivi Y, Tregnaghi A, et al. Evaluation of lymph node perfusion using continuous mode harmonic ultrasonography with a second generation contrast agent. J Ultrasound Med 2004;23:829–836.
- 25.
Nyman HT, Lee MH, McEvoy FJ, et al. Comparison of B-mode and Doppler ultrasonographic findings with histologic features of benign and malignant superficial lymph nodes in dogs. Am J Vet Res 2006;67:978–984.
- 26.↑
Salwei RM, O'Brien RT, Matheson JS. Characterization of lymphomatous lymph nodes in dogs using contrast harmonic and power Doppler ultrasound. Vet Radiol Ultrasound 2005;46:411–416.
- 27.
Ariji Y, Kimura Y, Hayashi N, et al. Power Doppler sonography of cervical lymph nodes in patients with head and neck cancer. AJNR Am J Neuroradiol 1998;19:303–307.
- 28.
Tschammler A, Gunzer E, Reinhart E, et al. Dignitätsbeurteilung vergrösserter Lymphknoten durch qualitative und semiquantitative Auswertung der lymphknoten perfusion mit der farbkodierten Duplexsonographie. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 1991;154:414–418.
Appendix
Calculation methods and selected tracings used in the determination of Doppler ultrasonographic indices.
