Breast cancer: comparison of quantitative dual-layer spectral CT and axillary ultrasonography for preoperative diagnosis of metastatic axillary lymph nodes

Breast cancer is the most common cancer affecting women and the second leading cause of cancer-related deaths [1]. The axillary lymph nodes are the most common site for metastatic disease [2]. Correct identification of axillary lymph node metastases is important for both determining the patient’s prognosis and for planning the patient’s treatment [3], as the presence of axillary lymph node metastases has a negative impact on overall survival [4].

Ultrasonography (US) of the breasts and axillae with fine needle aspiration (FNA) from morphologically suspicious axillary lymph nodes is often used for preoperative staging of the axilla. Sensitivity for detecting axillary lymph node metastases has been reported to be around 80% with specificity near 100% [5, 6]. US is usually followed by sentinel lymph node biopsy (SNB), which, if negative, rules out axillary lymph node metastases with high accuracy [7]. Patients with < 3 positive nodes on SNB can often be managed with SNB as the only surgical procedure in the axilla [8, 9]. However, patients with clinically positive lymph nodes, ≥ T3 tumour, ≥ 3 positive nodes on SNB, or planned mastectomy will often receive axillary lymph node dissection (ALND) [10]. Since both SNB and ALND are invasive surgical procedures (which may cause immediate complications such as infection or, in a few cases, long-term effects such as lymphedema, shoulder impairment, or pain [11‐13]), non-invasive methods to reduce the need for SNB and ALND are highly desired.

Dual-energy computed tomography (DECT) is an emerging technique [14]. Where conventional CT describes the tissues relative x-ray attenuation properties measured in HU, the dual-energy approach provides quantitative measures of material concentrations and decompositions (e.g., iodine and uric acid) as well as the ability to suppress materials such as iodine, water, or calcium [15, 16].

Most dual-energy techniques acquire a spectral CT dataset by scanning the same area twice with different kilovolt peak. This can be achieved either by dual-source CT or by rapid tube potential switching (single-source) [14, 16, 17]. The improvement in iodine visualisation and quantification offered by DECT has shown promising results in the diagnosis of lymph node metastases from a range of different cancer diseases [18‐22], including breast cancer [23]. However, this approach previously required that patients were selected prospectively for DECT and that a special DECT protocol was selected prior to scanning, limiting its clinical usefulness.

The newest generation of DECT systems uses a dual-layer detector [24] for separating the x-ray spectrum, with the top row detecting low-energy photons and the bottom layer detecting high-energy photons [14, 16, 17]. This removes the need for changing the tube kilovolt peak or other scan parameters whilst still having all spectral CT parameters available for analyses using a spectral workstation. In addition, the conventional CT images have been shown to not differ from single-energy CT in dose [25] or image quality [26]

The aim of the present study was to assess the diagnostic performance of quantitative spectral CT parameters from dual-layer (detector) spectral CT (DLSCT) to preoperatively diagnose axillary lymph node metastases in breast cancer patients undergoing definitive surgery. Quantitative spectral parameters were compared to axillary US using the results from the histopathological examination of the surgical specimen as reference.

The best diagnostic performance was found for Δ contrast enhancement using a cutoff value of 17 HU, which resulted in a sensitivity, specificity, and AUC of 0.79, 0.92, and 0.87, respectively. In a previous study, Tawfik et al. [18] found a sensitivity of 0.90, specificity of 0.78, and an AUC of 0.90 for Δ contrast enhancement to detect cervical lymph node metastases from squamous cell carcinoma using dual-source DECT. The differences, albeit minor, in sensitivity and specificity between our study and the study by Tawfik et al. could be due to them scanning the lymph nodes in the venous phase. Since Δ contrast enhancement was the difference in HU between contrast-enhanced images and VNC images, it could potentially also be obtained by scanning the patients before and after contrast administration.

Compared to the histopathology report, Δ contrast enhancement missed five metastatic lymph nodes whilst axillary US missed six, three of them being from the same patient. The three lymph node metastases that were detected by Δ contrast enhancement but not by axillary US all had a short axis < 10 mm with morphologically benign appearance. Δ contrast enhancement only had one false positive in a high attenuating lymph node (Fig. 5). The histopathology report showed silicone content from a leaked breast implant and the lymph node was correctly diagnosed as non-metastatic by axillary US. However, in contrast to DLSCT, the breast radiologists performing the axillary US had access to the patient records and were aware of the leaking breast implant. DECT is able to visualise silicone and has been shown to be able to detect leaks from ruptured breast implants [31, 32]. Since the patient’s records would be available in a clinical setting, it is likely that a clinical reading of the DLSCT would correctly diagnose the silicone uptake in the lymph node. The high specificity and PPV found in our study (0.92 and 0.95, respectively) are particularly promising since, if the high specificity and PPV could be replicated in a larger patient cohort, > 3 positive nodes on DLSCT could potentially mean that SNB could be avoided and the patient could proceed directly to surgery and ALND. Because DLSCT, in contrast to axillary US, exposes the patient to ionising radiation, it is unlikely to become a standard part of the diagnostic workup in all women with loco-regional breast cancer. However, quantitative DLSCT could potentially have a place in the diagnostic workup of women with a high a priori risk of metastatic disease where contrast-enhanced CT is already indicated (e.g., patients with palpable lymph nodes).

Whilst iodine density had lower diagnostic performance than Δ contrast enhancement, it still showed sensitivity comparable to axillary US. The optimal cutoff for distinguishing metastatic and non-metastatic lymph nodes were calculated to 1.5 mg/mL. In a recent study by Volterrani et al. [33], they found a cutoff value of 1.7 mg/mL iodine in the late arterial phase for differentiating malignant and benign breast lesions using a rapid tube potential switching DECT protocol. Based on these results, it seems that adopting a threshold of 1.5–2.0 mg/mL iodine could be useful both for assessing breast lesions as well as lymph nodes.

Significant differences in iodine density, spectral slope, Z effective, conventional CT HU, and Δ contrast enhancement were observed between non-metastatic contralateral lymph nodes in arterial phase and non-metastatic inguinal lymph nodes in portal-venous phase. This was expected as the spectral CT parameters all depend on the iodine uptake in the lymph nodes which is phase dependent. These findings are in concordance with previous studies which have shown that iodine density, spectral slope, and Z effective are lower in arterial phase compared with venous phase for both metastatic and non-metastatic lymph nodes [19, 23]. In a study by Zhang et al. [23] where DECT was performed both in arterial and venous phase, spectral parameters from axillary lymph nodes in the venous phase showed higher diagnostic performance than in the arterial phase.

Some previous studies have tried to compensate for phase and differences in the patients’ circulation by normalising the iodine density or Z effective to the internal carotid artery [19] or aorta [20, 23]. We did not normalise our data to the arterial attenuation, since our aim was to assess the diagnostic performance of the available images in a clinical setting. Instead, all our axillary lymph node parameters were obtained in the arterial phase 15 s after a threshold of 150 HU was measured in the descending aorta to minimise differences in circulation. The one spectral CT parameter that does not depend on iodine uptake—the VNC—was not significantly different between phases. This shows that the phase has to be considered when quantifying spectral CT parameters and is a challenge for adopting diagnostic thresholds since DECT protocols vary amongst institutions.

The strengths of our study include the prospective patient enrolment and the use of histopathological examination of the lymph node as reference standard. The lymph nodes were semi-automatically segmented using tumour tracking software by two radiologists in consensus to reduce the subjective nature of manual drawing of regions of interest. We used a DLSCT scanner that acquires spectral CT datasets without the need for changing the scan protocol, where previous studies assessing quantitative spectral CT parameters for detecting lymph node metastases have used dual-source CT or single-source CT with rapid tube potential switching. The reconstruction of the spectral base images took less than five minutes after which all spectral CT datasets were available making it feasible to implement in the clinic.

Our study has some limitations. First, we did not correlate lymph nodes from DLSCT with the histopathology report on a node-to-node basis but used a region-to-region correlation instead. Second, all axillary lymph nodes were scanned in arterial phase. It would have been interesting to compare the diagnostic performance between contrast phases using DLSCT. Finally, our study population was small, and only 13/36 axillae had benign histopathology. Since the patients were referred for contrast-enhanced CT of the chest, abdomen, and pelvis on a suspicion of disseminated disease, the a priori risk of having axillary lymph node metastases was increased in this patient cohort compared to the general patient population. In order to generalise the results, multi-centre studies with larger patient cohorts are needed.

In conclusion, DLSCT improved the sensitivity to detect lymph node metastases compared to conventional CT HU with similar sensitivity and specificity as US with FNA and had a very high PPV. DLSCT is a promising quantitative technique for identifying lymph node metastases and could potentially reduce the need for SNB in patients with positive findings. Future studies should aim to validate the diagnostic performance of DLSCT on a node-to-node basis in a larger cohort.