In our study, we applied a multiparameter evaluation of the ALNs based on B-mode US features (shape, cortex, capsule), vascularity (CD and MFI) and stiffness (SWE, E mean and E max value in the cortex) in patients with BC before surgery (Fig. 2, Fig. 3 and Fig. 4). We found that a cortical thickness greater than 3 mm or a noncircumscribed capsular margin of the ALN was the single most suspicious feature suggesting metastasis, with good accuracy and a high PPV. Similar results were published by Abe H et al. and Choi YJ et al. regarding patients with BC before surgery (22,23). Abe H et al. reported cortical thickening in 63 (79%) of 80 metastatic nodes (22). Choi YJ et al revealed in their study that in the multivariate logistic regression analysis, a cortical thickness greater than 3 mm was the most accurate indicator, with a 4.14-fold increased risk of the presence of axillary lymph node metastasis compared to a cortical thickness less than 3 mm (23). In our study, in the cases of nodes with a cortical thickness greater than 3 mm, the vast majority, 83.1%, were characterized by the presence of metastasis; however, in 10/59 ALNs, no metastatic cells were detected (false positive results). On the other hand, in a group of ALNs with a cortex < 3 mm (false negative cases) and in 10/103 patients after surgery, metastases were identified. Of these, micrometastases were verified in 3 patients; in one node, the focus of the metastasis was 3 mm, and in 4 patients, these were ILC metastases. In the remaining two patients, the metastasis was classified as BC macrometastasis. When analysing this group of patients, it should be noted that the presence of micrometastases cannot be detected in the ALNUS examination, and their presence does not change the therapeutic procedure. However, in one patient with ILC, failure to visualize the lymph nodes resulted in the patient undergoing another procedure in the axilla. According to the results published by other authors regarding metastatic LNs in BC patients with ILC, Morrow E et al., in a retrospective study using ALNUS in 209 such patients, reported a sensitivity of 32.1% and a very high false negative rate (24). The authors suggested that in a preoperative assessment of ALNs, biopsy should be performed even if the LN presents as normal in the US examination. Histologically, ILC cells are small and scattered throughout the LN, so at early stages, they do not destroy the architecture of the LN; therefore, it is challenging to detect the metastasis. Kurst et al., who evaluated bladder cancer patients with biopsy-proven metastasis in 219 ALNs, reported that asymmetrical cortical thickening, as a single parameter, was an easy feature to assess via ALNUS, with a sensitivity of 88.7%, specificity of 54.3%, PPV of 77.5%, and NPV of 73.1% (25). We did not assess these parameters in our study.
In our study, differences in vascularity and stiffness assessed by imaging of the LN were also found as features differentiating their characteristics. Both of these features were characterized by a high PPV but low NPV.
In our study, perforating vessels or mixed vascularity patterns were more common in metastatic LNs than in normal LNs (31/59 vs. 5/44). However, this pattern of vascularity occurred only in approximately half of the LNs with metastases. The absence of vessels was observed only in 16 LNs (8 with metastases in 8 without metastases), so a lack of visible vessels in the LN is not a feature that could differentiate their characteristics.
According to the statistical analysis of the OR, the highest probability of malignancy was obtained for the CD mixed-vascularity pattern (9.2), followed by the perforating pattern (3.9) and the MFVI predominance pattern (3.7). These parameters independently increased the probability of developing lymph node metastases. However, the AUCs for these features in our study were 0.58 and 0.63, respectively. Interesting results were published by Kurt SA et al. on 219 metastatic ALNs. The authors assessed the superb microvascular imaging vascularization pattern of the LN and found that abnormal vascularization (defined as suspicious for metastasis: peripheral, penetrant, anarchic blood supply or avascularity) as a single feature in this metastatic LN had a sensitivity of 79.8, a specificity of 78.6, a PPV of 91.5, an NPV of 57.1, and an AUC of 0.792.
On the other hand, in a study published by Yang WT et al. (26,27) and Hussien (28), the authors reported that central vascularity patterns were most often observed in benign reactive or inflammatory LNs. Compared with benign axillary LNs, malignant LNs presented greater total and peripheral vessel counts (sensitivity and specificity, respectively). In a study of the measured vascularity index (VI), Uslu et al. used MVFI parameters. The VI is the ratio between the pixels of the Doppler signals and the pixels of the total lesion (21). The authors reported, based on the 58 LNs (35/58 were metastatic), that this index allows a quantitative assessment of the blood flow richness in the LN. For the cut-off VI of 8.55, the authors reached a sensitivity of 82, a specificity of 100, a PPV of 100, an NPV of 85, and an AUC of 0.91.
In reactively stimulated or normal LNs, Doppler examination revealed visible vessels from the hilum side towards the capsule, with a reduction in their number and diameter (vascular tree). In metastatic nodes, lymphangiogenesis occurs from the lymphatic channels penetrating the node from the capsule side, and peripheral or penetrating vascularization is visible. On the other hand, assessing small vessels using CD and MVFI is limited in deeper organs, and it can be clinically difficult to assess whether the lack of flow detection is because of this depth or whether the flow is not visible in small foci of metastases, which requires further investigation (29). Nathanson et al. reported that lymphatic endothelial cells (LECs) express distinctive lymphatic endothelial markers, such as vascular endothelial growth factor-3, and cytokines that stimulate them (30). Therefore, malignant cells invade the lymphatic lumen, move with the lymph from the subcapsular sinus to the cortex and proliferate. Tumour cells express VEGF-C and VEGF-D, which are connected with the proliferation of new vessel capillaries (lymphangiogenesis). However, not all tumour cells express chemokine receptors.
Current studies have reported low interobserver variability for SWE and significant improvement in differentiating ALN metastasis from benign ALN metastasis (31). We also found a significant difference in sonoelastography between benign and metastatic LNs in our study. The best cut-off value was achieved for an E max > 26.5 kPa, which was characterized by a high PPV and specificity but a low NPV and sensitivity. We placed the ROI in the cortex because metastatic cells invade the LNs from the capsula through the afferent lymphatic vessels. In this area of the LNs, the increased density of pathological cells likely increased the stiffness. In the meta-analysis published by Wang R.Y. et al., both shear-wave elastography (SWE) and strain elastography (SE) were used for the assessment of ALNs (32). Both SE and SWE had relatively high pooled AUCs of 0.85 and 0.94, respectively. The SWE values were significantly greater in malignant LNs. Similar to our study, SWE (maximum stiffness) had a greater AUC than SWE (mean stiffness), and the cut-off values ranged from 20.9 kPa to 38.6 kPa. Seo M et al evaluated B-mode features and SWE in 54 LNs in BC patients and found significantly greater SWE values for metastatic LNs (33). The mean values for metastatic LNs were E max = 79.8 kPa and E mean = 55.99 kPa; for benign LNs, the mean values were E max = 13.35 kPa and E mean = 9.38 kPa. However, these results could be explained by LN selection. SWE was performed in LNs where the mean cortical thickness of the metastatic LNs was greater than that in other studies (8.99 mm), and the majority of the metastatic LNs were metastatic (34/54). These results revealed that elastography should be performed during the assessment of LNs and could aid in decision making before biopsy.
Finally, in the multiparametric analysis, which assessed the AUC as well as the OR, the combination of B-mode features (cortex larger than 3 mm, noncircumscribed margin of the LN capsule) and SWE (E max > 26 kPa) for one model and the combination of B-mode features (noncircumscribed margin of the LN capsule), vascular analysis on CD (mixed or penetrating pattern) and SWE (E max > 26 kPa) were superior for determining a single approach. Similar results were reported by Seo M et al., who reported a combined modality (cortical thickness > = 4.85 mm or maximum stiffness > = 20.9 kPa), with an AUC of 0.946.
In the future, we plan to assess patterns of enhancement via CEUS as a predictor of malignancy and deep learning methods to improve the preoperative diagnosis of metastatic LNs in patients with BC.
Certainly, our study has some limitations:
First, this study was conducted in women with breast cancer, and consequently, the number of metastatic LNs was high. Second, it was not possible to determine whether the nodes that qualified during US examination to biopsy were exactly the same nodes assessed by the pathologist. We targeted the most suspicious LNs with a cortex larger than 3 mm for biopsy. Another limitation of our study involves the false-negative results, which could be explained by deep LNs being located behind the pectoralis muscles or the small size of the LNs.