Diagnostic efficacy of ultrasonography, Doppler ultrasonography and elastography in the evaluation of suspected malignant lymph nodes
Article Category: Research paper
Pubblicato online: 28 feb 2023
Pagine: 1 - 9
Ricevuto: 26 giu 2022
Accettato: 10 ott 2022
DOI: https://doi.org/10.15557/JoU.2023.0001
Parole chiave
© 2023 Hülya Çetin Tunçez, Ali Murat Koç, Zehra Hilal Adıbelli, Fatma Zeynep Arslan, Asuman Argon, Gülşen Yücel Oğuzdoğan, published by Sciendo
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Ultrasound elastography (USE) is a modality that makes it possible to interpret the degree of stiffness of body tissues, similarly to the findings obtained by palpation. Malignant tissues are usually harder due to the desmoplastic reaction. Accordingly, a harder tissue pattern and higher velocity values are detected on USE in malignant lesions(1). A recent study found that combining USE with the US examination serves to increase the accuracy of detecting malign lymph nodes(2). In a recent meta-analysis evaluating the diagnostic performance of USE for axillary lymph node metastasis, the authors reported a huge variety among studies. They found that the most important reason for such heterogeneous results included study design and measurements(3). We believe that a reference value and standard measurement methods are necessary. However, recent literature data suggest being careful in the interpretation of USE values in view of great heterogeneity among the researchers(3).
In this prospective study, the efficiency of imaging findings was investigated by comparing the histopathological results of lymph nodes with Doppler and US features and elasticity scores.
Patients over 18 years of age, presenting with lymph nodes suspected of malignancy in the neck and axilla on US, were evaluated prospectively. In B-mode US, long axis/short axis ratio >2 in the lymph node, >1 cm in the short axis, increased cortex thickness (>2.5 mm), absence of fatty hilus were regarded as a suspicion of malignancy(4, 5, 6). The lymph nodes which were suspicious for malignancy were followed-up for 12 weeks with an appropriate antibiotic therapy. During antibiotic treatment, 875 mg of amoxicillin and 125 mg of clavulanic acid were administered to the patients for 10 days. Subsequently, the reduction in size was monitored, and the lymph nodes suspected of malignancy were re-evaluated 12 weeks later.
The study was conducted prospectively between 2019 and 2020. Patients who did not agree to participate in the study, patients under the age of 18, patients with biopsy contraindications/refusal to biopsy, patients in whom US/biopsy could not be performed, and pregnant or lactating patients were excluded from the study. A total of 10 patients whose biopsy results were unsatisfactory and who met the exclusion criteria were excluded from the study.
Oral and written consents were obtained from the parents of all patients who participated in our study, as necessary. Ethical approval was obtained from the local scientific research ethics committee (decision No. 05, dated 17.01.2019). The study was conducted on the basis of ethical standards outlined in the Declaration of Helsinki of the World Medical Association. All sonographic evaluations were performed by two radiologists who are specialized in this field of radiology.
The lymph nodes of the patients were evaluated using a 6–13 Mhz linear probe with an ultrasound device (Hitachi HI VISION) with B-mode US, Doppler US and strain USE features. All radiological examinations were performed by a radiologist with 8 years’ experience in the field of sonographic evaluation. The patients were placed in the supine position, and mild manual compression was applied at short intervals to the lymph node in the perpendicular position on the strain elastographic examination. In B-mode US, the features of lymph nodes suspicious for malignancy (i.e. irregular shape, increased size, pronounced hypoechogenicity, presence of micro/macro calcification, short axis/long axis ratio >2, increased size of the short axis, increased cortex thickness, obliterated hilus or increased cortex thickness >3.5 mm) were described as abnormal(5,6). Resistivity index (RI), pulsatility index (PI), and acceleration rate and time were evaluated for intranodal arterial structures on color Doppler US, and the strain ratio value and elasticity scores were recorded on USE. The strain ratio value was calculated as the ratio of the lymph node examined to the elasticity values of the adjacent muscle tissue for the lymph nodes in the neck and axilla(5,6). In the axilla, it was accepted as the ratio of the elasticity values of the adipose tissue adjacent to the lymph node(5). In the calculation of the strain ratio value, an examination area (region of interest, ROI) was as large as possible to represent that part. Reference ROI values were obtained by measuring from the reference tissue in the same image as the study area. The ratio of the reference ROI to the first ROI was calculated. The size of the ROI cursor was adjusted for each lesion according to the lesion size. The ROI cursor was placed on the stiffest part according to the color map. The mean strain ratio value obtained by measuring the lesions at least twice was included in the study(6). After the sonographic examination, the patients underwent an ultrasound-guided fine needle aspiration cytology or tru-cutting needle biopsy. For fine-needle aspiration biopsy of the lymph node, a 21-gauge needle and a 10 ml disposable injector were preferred. Histopathological examination results were used as reference standards. In order to evaluate the adequacy of the gray-scale findings, vascularity features and elasticity of the lymph nodes in predicting the pathology results, the cases were classified as benign or malignant by two radiologists who were blind to the histopathology results of the patients, based on the current literature on the subject (Tab. 1, Fig. 1). A short axis diameter ≥1 cm, or a long/short axis ratio was <2, round shape and presence of abnormal hilus were considered as malignant features. On Doppler US, detecting blood supply outside of the hilus (peripheral or both central and peripheral blood supply), RI >0.7 and PI >1.4 were regarded as malignant (Fig. 2)(5,6).
Fig. 1
Schematic display of elasticity patterns with USE
Fig. 2
Elasticity scoring
| No stiff area within the lesion | |
| The percentage of stiff area is less than 45% | |
| The percentage of stiff area is greater than 45% | |
| The lesion is mostly stiff | |
| The lesion is completely stiff |
Elasticity patterns were determined on USE. Pattern 1 was described as no stiff area within the lesion. If the percentage of stiff area was less than 45%, it was considered as Pattern 2. If the percentage of stiff area was greater than 45%, it was classified as Pattern 3. If the lesion is mostly stiff and containing minimal soft area, it was considered as Pattern 4. If the lesion was completely stiff, it was considered as Pattern 5. Patterns 1 and 2 were considered as benign, while patterns 3,4 and 5 were classified as malignant. When the elastographic measurements were evaluated quantitatively, the lymph nodes with an SR value of >1.5 were recognized as malignant, in accordance with the threshold values found in similar studies in the literature. The USE results were compared with the sonographic interpretations of benign-malignant.
Statistical analysis of the data was performed using IBM SPSS Statistics version 24. Pearson chi-squared test and Fisher’s exact test were applied for comparison of categorical data between the groups; since continuous data were not normally distributed (Kolmogorov-Smirnov
In this study, a total of 100 lymph nodes of 96 patients aged between 32 and 86 years were evaluated ultrasonographically. The pathology results of 10 patients were described as non-diagnostic pathology, two patients were re-biopsied, but they were not included in the study because of insufficient results. Of all the patients, 79 (92%) were female, and 7 (8%) were male with a mean age of 53.7 years. Of the lymph nodes, 76 (84.5%) were located in the axilla and 14 (15.5%) were located in the cervical region. Three of 14 cervical lymph nodes were found to be malignant, and 51 of 76 axillary lymph nodes were found to be malignant. A statistically significant difference was found between the groups according to localization (
The morphological difference (round, ovoid) between the benign and malign lymph nodes was statistically significant (
Ultrasonographic features of the examined lymph nodes according to their pathological diagnosis
| Pathological diagnosis | X2 | ||||||
|---|---|---|---|---|---|---|---|
| Benign | Malignant | ||||||
| % | % | ||||||
| Round | 2 | 5.6 | 13 | 24.1 | |||
| Ovoid | 34 | 94.4 | 41 | 75.9 | 5.333 | ||
| Normal fatty hilum | 17 | 47.2 | 13 | 24.1 | 5.213 | ||
| Abnormal obliterated hilum | 19 | 52.8 | 41 | 75.9 | |||
| No | 28 | 77.8 | 46 | 85.2 | |||
| Yes | 8 | 22.2 | 8 | 14.8 | 0.811 | 0.368 | |
| Regular | 36 | 100.0 | 50 | 92.6 | 2.791 | 0.147 | |
| Irregular | 0 | 0.0 | 4 | 7.4 | |||
| No | 34 | 94.4 | 53 | 98.1 | 0.92 | 0.561 | |
| Yes | 2 | 5.6 | 1 | 1.9 | |||
| Hypoechoic | 34 | 94.4 | 54 | 100.0 | 2.900 | 0.163 | |
| Isoechoic | 2 | 5.6 | 0 | 0.0 | |||
The long axis lengths of the examined lymph nodes ranged from 8 mm to 50 mm, while the short axis lengths varied from 5 mm to 27 mm. The ratio of long axis to short axis was between 0.88 and 4. The mean of the short axis was 9.36 ± 2.93 mm in the benign group and 13.45 ± 6.15 mm in the group of malignant lymph nodes. The average long axis to short axis ratio was 2.17 ± 0.59 in the benign group and 1.86 ± 0.53 in the malignant group, and the differences between the short axis length and the long axis/short axis ratio were statistically significant (
When predicting the pathology outcome by ROC analysis for the power of B-mode US data, the threshold value calculated for the short axis value, with 43.4% sensitivity and 88.9% specificity, was 13 mm. The Area Under Curve (AUC) value was found to be 69.7%, and it was statistically significant (
On Doppler ultrasound examination, 55 (61.1%) lymph nodes showed central (hilar) blood supply, 28 (31.1%) showed peripheral blood supply, and 7 (7.8%) had both central and peripheral (anarchic) blood supply. Of the 36 lymph nodes that were found to be histopathologically benign, 28 (77.8%) had central blood, 7 (19.4%) had peripheral blood, and 1 (2.8%) had both central and peripheral blood supply. Of the 54 malignant lymph nodes, 27 (50%) showed central blood supply, 21 (38.9%) had peripheral, and 6 (11.1%) had both central and peripheral blood supply. Sensitivity was 50%, specificity was 77.78%, and accuracy was calculated as 61.1%. There was a statistically significant difference between the groups in terms of blood supply patterns on Doppler US (
On Doppler examination, RI was found to be in the range of 0.52– 1.79, while PI was shown to be between 0.70 and 87. The acceleration rate ranged from 0.10 to 3.55 ms, while the acceleration time ranged between 32 and 421 ms. The RI value was 0.75 ± 0.15 and 0.78 ± 0.28 in benign and malignant cases, respectively, while the PI value was 3.75 ± 14.28 and 1.56 ± 0.64, respectively. The acceleration rate was 0.58 ± 0.58 and 0.62 ± 0.41 in the benign and malignant groups, respectively, and the acceleration time was 160.64 ± 57.36 and 160.78 ± 60.39, respectively. No statistically significant difference was found in terms of RI, PI, acceleration rate and acceleration time, and differentiation between benign and malignant lymph nodes (
Based on the threshold values found in the ROC analysis for the power of Doppler data in estimating the pathology result, the AUC values calculated to predict the pathology outcome of RI, PI, acceleration rate and acceleration time were not found to be statistically significant (
In the elastographic examination, the strain ratio ranged between 0.42 and 5.9. While the mean strain ratio value was 1.52 ± 0.98 in patients with benign histopathological results, it was 2.04 ± 1.21 in malignant patients, and the difference was found to be statistically significant (
In the ROC analysis performed for the power of USE data (SR value) to predict the pathology outcome, the threshold value was calculated as 1.21, with a sensitivity of 77.8% and a specificity of 52.8%. The AUC value was found to be 66.3% and was statistically significant (
In addition, elasticity patterns were evaluated by USE. Patterns 1 and 2 were considered benign, while patterns 3, 4 and 5 were considered malignant. A total of 47 (87%) of 54 patients who were found to be malignant histopathologically had patterns 3, 4 and 5, while 20 (55%) of 36 patients found to be benign had patterns 1 and 2. The accuracy rate was 77.8%, and the
When the imaging methods and histopathological results were compared, 24 (44.4%) of 54 pathologically malignant patients were interpreted as malignant by B-mode US, and 32 (58.5%) were evaluated as malignant by Doppler US; 37 (68.5%) of these patients on USE, 35 (64.2%) of them on B-mode and Doppler US, 48 (88.7%) of them on Doppler US and USE, and 49 (90.4%) of them on all three imaging modalities were interpreted as malignant. In the Kappa fit analysis, the consistency between the histopathology results and the results determined by oppler US, USE, double and triple methods was found to be statistically significant (
Fig. 3
Visual representation of the distribution of pathology results and interpretations for different imaging methods
Kappa analysis results for the compatibility of pathology results with B-mode, Doppler US and US elastography results
| Pathological diagnosis | Total | Kappa value | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Benign | Malignant | ||||||||
| % | % | % | |||||||
| Benign | 27 | 75.0 | 30 | 55.6 | 57 | 63.3 | 0.177 | 0.061 | |
| Malignant | 9 | 25.0 | 24 | 44.4 | 33 | 36.7 | |||
| Total | 36 | 40.0 | 54 | 60.0 | 90 | 100.0 | |||
| Benign | 28 | 77.8 | 22 | 41.5 | 50 | 56.2 | |||
| Malignant | 8 | 22.2 | 32 | 58.5 | 40 | 43.8 | 0.341 | ||
| Total | 36 | 40.0 | 54 | 60.0 | 90 | 100.0 | |||
| Benign | 21 | 58.3 | 17 | 31.5 | 38 | 42.2 | 0.266 | ||
| Malignant | 15 | 41.7 | 37 | 68.5 | 52 | 57.8 | |||
| Total | 36 | 40.0 | 54 | 60.0 | 90 | 100.0 | |||
| Benign | 26 | 72.2 | 19 | 35.8 | 45 | 50.0 | |||
| Malignant | 10 | 27.8 | 35 | 64.2 | 45 | 50.0 | 0.350 | ||
| Total | 36 | 40.0 | 54 | 60.0 | 90 | 100.0 | |||
| Benign | 18 | 50.0 | 6 | 11.3 | 24 | 27.0 | |||
| Malignant | 18 | 50.0 | 48 | 88.7 | 66 | 73.0 | 0.409 | ||
| Total | 36 | 40.0 | 54 | 60.0 | 90 | 100.0 | |||
| Benign | 18 | 50.0 | 5 | 9.6 | 23 | 26.1 | 0.428 | ||
| Malignant | 18 | 50.0 | 49 | 90.4 | 67 | 73.9 | |||
| Total | 36 | 40.0 | 54 | 60.0 | 90 | 100.0 | |||
US – ultrasonography
When the individual and combined effects of B-mode, Doppler US and USE in differentiating lymph nodes were evaluated, the method with the highest sensitivity was the combination of all three imaging methods (90.4%), while the modality with the highest specificity was Doppler US (77.8%). The examinations with the highest overall accuracy were the combination of all three methods (73.9%), while the examination with the lowest accuracy was B-mode US (56.7%) (Tab. 4).
ROC analysis results for the power of imaging methods in predicting pathological diagnosis
| Sensitivity | Specificity | Positive predictive value | Negative predictive value | Area under the curve | Confidence interval (95%) | Accuracy | ||
|---|---|---|---|---|---|---|---|---|
| 44.4 | 75.0 | 72.7 | 47.4 | 0.597 | 0.489–0.699 | 0.108 | 56.7 | |
| 58.5 | 77.8 | 79.5 | 56.0 | 0.681 | 0.574–0.776 | 66.3 | ||
| 68.5 | 58.3 | 71.2 | 55.3 | 0.634 | 0.526–0.733 | 64.4 | ||
| 64.2 | 72.2 | 77.3 | 57.8 | 0.682 | 0.575–0.777 | 67.4 | ||
| 88.7 | 50.0 | 72.3 | 75.0 | 0.693 | 0.587–0.787 | 73.0 | ||
| 90.4 | 50.0 | 72.3 | 78.3 | 0.702 | 0.595–0.795 | 73.9 |
A ROC analysis for the power of imaging methods in predicting the outcome of pathological results was performed. The AUC value calculated for the predictive power of the pathology results of the B-mode US results was not statistically significant (
The criteria that were found to be statistically significant among the B-mode findings were the short axis length and the long axis/short axis ratio. When the effects of their elastographic strain ratio value and their dual combinations were evaluated, the association of short axis and strain ratio was found in 48 (88.9%) of 54 malignant patients, while the association of the long axis/short axis and strain ratio was found in 45 (83.3%). The findings were statistically significant (
The threshold values found in the ROC analysis for the power of imaging methods in predicting the pathology results of the cervical lymph nodes; 33 mm for the long axis, 14 mm for the short axis, and 2.21 for the long axis/short axis ratio were calculated (Tab. 5). While the sensitivity of short axis and USE is 100% in predicting the pathological outcome in the cervical region, the rate is lower (88%) in the axilla. The accuracy of the short axis length-long axis/short axis ratio and their strain ratio as well as their binary combinations were found to be higher in the axillary region than in the cervical region.
ROC analysis results for the power of imaging methods in predicting the pathology result of cervical lymph nodes
| Threshold value | Sensitivity | Specificity | predictive Positive value | predictive Negative value | Area under the curve | Confidence interval (95%) | ||
|---|---|---|---|---|---|---|---|---|
| >33 | 33.3 | 100.0 | 100.0 | 84.6 | 0.515 | 0.242–0.781 | 0.947 | |
| >14 | 66.7 | 90.9 | 66.7 | 90.9 | 0.742 | 0.447–0.932 | 0.232 | |
| ≤2.21 | 100.0 | 54.6 | 37.5 | 100.0 | 0.788 | 0.494–0.954 | ||
| >0.76 | 100.0 | 54.6 | 37.5 | 100.0 | 0.742 | 0.447–0.932 | 0.102 | |
| >1.34 | 100.0 | 54.6 | 37.5 | 100.0 | 0.697 | 0.402–0.906 | 0.180 | |
| >0.41 | 100.0 | 54.6 | 37.5 | 100.0 | 0.727 | 0.432–0.923 | 0.106 | |
| >181 | 0.0 | 63.6 | 0.0 | 70.0 | 0.53 | 0.255–0.793 | 0.848 | |
| >1.96 | 66.7 | 81.8 | 50.0 | 90.0 | 0.576 | 0.292–0.826 | 0.758 |
The threshold values found in the ROC analysis for the power of imaging methods in predicting the pathology results in the axilla; long axis: 24 mm, short axis: 10 mm, long axis/short axis ratio: 1.46, RI: 0.76, PI: 1.26, acceleration speed: 0.41, acceleration time: 136, SR value calculated as 1.75; the AUC values calculated for the long axis, short axis, PI and USE (SR value) were found to be statistically significant (
Fig. 4
ROC analysis results for the power of imaging methods in predicting the pathology result of axillar lymph nodes
| Threshold value | Sensitivity | Specificity | Positive predictive value | Negative predictive value | Area under the curve | Confidence interval (95%) | ||
|---|---|---|---|---|---|---|---|---|
| >24 | 40.0 | 88.0 | 87.0 | 42.3 | 0.642 | 0.523–0.749 | ||
| >10 | 54.0 | 84.0 | 87.1 | 47.7 | 0.73 | 0.615–0.826 | ||
| ≤1.46 | 24.0 | 100.0 | 100.0 | 39.7 | 0.624 | 0.504–0.733 | 0.060 | |
| >0.76 | 52.9 | 80.0 | 84.4 | 45.5 | 0.621 | 0.503–0.730 | 0.055 | |
| >1.26 | 66.7 | 60.0 | 76.2 | 48.4 | 0.639 | 0.518–0.748 | ||
| >0.41 | 66.7 | 52.0 | 72.7 | 44.8 | 0.573 | 0.452–0.688 | 0.295 | |
| ≤136 | 39.6 | 84.0 | 82.6 | 42.0 | 0.531 | 0.410–0.649 | 0.656 | |
| >1.75 | 47.1 | 88.0 | 88.9 | 44.9 | 0.688 | 0.571–0.789 |
Benign axillary lymph nodes are typically smaller than 2 cm in maximal size, and their parenchymal cortical thickness is less than 2.5 mm(7). Unlike axillary lymph nodes, the normal short-axis in benign cervical nodes varies according to the levels in the neck; in general, the short axis diameter in axial plane should be <10 mm for cervical lymph nodes. Even the presence of lymph nodes in some parts of the neck, such as the medial retropharyngeal area, is pathological(8).
Although the specific criteria distinguishing benign lymph nodes from malignant lymph nodes have not been determined precisely, various threshold values have been reported in the literature for the short axis of the lymph node: 5 mm, 8 mm, and 10 mm(9). In our study, the threshold value of 13 mm for the short axis of all lymph nodes showed 88.9% specificity in predicting malignant lymph nodes.
When the lymph nodes are evaluated as round and ovoid according to their shape, the threshold value for the long axis/short axis ratio in malignant lymph nodes was found to be 2 in several studies(10). In our study, malignant lymph nodes were found to have more rounded morphology and were found to be significant in accordance with the literature (
In studies performed according to vascular distribution patterns on Doppler US, hilar blood supply was observed in normal and reactive lymph nodes, while peripheral and mixed (hilar + peripheral) blood supply was seen in malignant lymph nodes(11, 12, 13, 14). However, the study by Ying
In some studies on lymph nodes performed with USE, the strain ratio value was measured, while in others the elasticity score was calculated. In our study, the cut-off value was calculated as 1.21, with 77.8% sensitivity and 52.8% specificity, in the ROC analysis performed for the strain ratio value measured with USE, and it was found to be statistically significant (
In the scoring system; in pattern 1; 80% or more of the lymph node in the section is red or green (soft), in pattern 2; 50–80% is red or green, pattern 3; 50–80% is blue, pattern 4; 80% or more of the lymph node is blue (stiff). However, a threshold pattern for sensitivity and specificity between benign and malignant lymph nodes was not defined in this scoring system(20). Alam
Regarding the limitations of the study, the reason for the high number of female patients and axillary lymph nodes was thought to be the fact that patients with suspected breast cancer were referred to our clinic and included in the study. Another disadvantage is that the compression load applied with free-hand USE is not standardized and may, therefore, cause some inter-observer and inter-patient variability in the same observer.
Another important limitation of our study was that we examined the lymph nodes in both cervical and axillary regions within the same study. There are significant differences in lymph nodes located in both regions in terms of size and adjacent fatty and muscle tissues. Histopathological results were obtained with tru-cut or fine-needle biopsy. This may result in imprecise results for processes that are actually malignant, such as some lymphomas.
In our study, USE was evaluated semiquantitavely. In addition to USE, performing B-mode US and Doppler US enabled a comparison of elastographic findings with sonographic findings. Shape, short axis, long axis/short axis ratio and hilum status were found to be statistically significant in B-mode US.
USE is a highly sensitive imaging method in the differentiation of benign and malignant lymph nodes, and when combined with B-mode US and Doppler US findings, it makes a significant contribution to the diagnosis. With an increased use of USE, the diagnostic accuracy will improve, and the number of unnecessary biopsies will decrease.