Abstract
BACKGROUND:
Ultrasound (US) and computed tomography (CT) are common diagnostic imaging methods for detecting and diagnosing papillary thyroid microcarcinoma (PTMC). However, single-source dual-energy spectral computed tomography (spectral CT) reduces beam hardening artefacts and optimizes contrast, which may add value in detecting PTMC.
OBJECTIVE:
To investigate values of applying single-source dual-energy spectral CT for diagnosing PTMCs, in comparison with high frequency ultrasound and conventional polychromatic images.
METHODS:
Thirty-one patients with suspected PTMC underwent contrast-enhanced dual-energy spectral CT. The images were analyzed by two experienced radiologists. Noise and contrast-noise-ratio (CNR) were compared between conventional CT and spectral CT. Ultrasonography was also performed by an experienced radiologist with a 7 to 12-MHz linear array transducer. Detection and diagnostic sensitivity were determined and compared.
RESULTS:
Forty-six pathologically-confirmed PTMC lesions were detected in 31 patients. Spectral CT had lower noise and higher CNR than conventional CT (P < 0.05). US detected more tumors (45/46 [97.8%] than conventional CT images (40/46 [87.0%]) or spectral CT images (44/46 [95.7%]). Among them, 30 (65.2%), 36 (78.3%), and 40 (87.0%) lesions were diagnosed correctly by conventional CT, spectral CT and US, respectively. Spectral CT had higher sensitivity than conventional CT (P = 0.031). However, there was no significant difference between spectral CT and US diagnostic sensitivities (P = 0.125).
CONCLUSION:
Single-source dual-energy spectral CT was superior to conventional polychromatic images and similar to high frequency ultrasound in detecting and diagnosing for PTMCs. CT had advantages in detecting level VI and VII lymph nodes. Spectral CT and US provided good results for PTMC, and aid preoperative diagnosis.
Keywords
Introduction
Papillary thyroid microcarcinoma (PTMC) is a subtype of papillary thyroid carcinoma measuring 1 cm or less in its maximal diameter [1]. Many PTMCs are usually asymptomatic and identified incidentally. In recent years, the incidence of thyroid cancer has increased [2, 3]. According to the literature, PTMC is detected in up to 35.6% of autopsy specimens as a latent carcinoma, and up to 11% of PTMCs are associated with local invasion and lymph node, as well as distant metastasis [4]. Along with the development of imaging technology, the detection rate of PTMC has been increasing. High frequency ultrasound is most commonly used because it has high resolution, is non-invasive, and has low costs and has been accepted in the detection and diagnosis of PTMC [5]. Computed tomography (CT) is also a good method for evaluating thyroid lesions because it shows not only primary tumours, but also the invasion of adjacent structures such as the retrosternal area, esophagus, and trachea, and is also able to detect lymph nodes in the central node (level VI) or upper mediastinal area [6]. However, the conventional CT has shortcomings in the detection and diagnosis of the microcarcinoma due to the beam hardening artefacts and poor contrast to the surrounding thyroid tissue.
In recent years, single-source dual-energy spectral computed tomography has been introduced. It has the capability for fast switching between two peak voltage settings (140 kVp and 80 kVp) and allows the reconstruction of the conventional polychromatic images corresponding to 140 kVp and monochromatic images with energies ranging from 40 to 140 keV which provides the ability to reduce beam hardening artefacts and optimize contrast with selectable monochromatic energy (keV) [7–10]. Therefore, spectral CT imaging theoretically could increase the detection of small tumours, especially PTMCs. However, to the best of our knowledge, there are few studies on the value of spectral CT in the detection and diagnosis for PTMCs. The purpose of this study was to evaluate the detection and diagnostic sensitivity of spectral CT for PTMCs in comparison with conventional polychromatic images and high frequency ultrasound.
Patients and methods
Patients
This study was approved by our institutional ethics committee, and written informed consent was obtained from all patients. From January 2016 to October 2016, 31 patients who were suspected of papillary thyroid microcarcinoma underwent neck CT using a single-source dual-energy spectral CT mode. Within 3 days, a high frequency ultrasound was also carried out at the same location. All the surgical treatments were decided based on the results of cytological analysis of the fine needle aspiration biopsy. The detection and diagnostic sensitivity for the primary tumours and metastatic cervical lymph node of the 2 methods were determined by comparing the results to the pathology reports after surgery. Solitary or multiple lesions, which were 1 cm or less in maximal diameter, were collected to objectively reflect the primary tumours and metastatic cervical lymph nodes of PTMCs.
CT scanning and imaging analysis
All patients underwent contrast enhanced neck CT scan on a high-definition CT scanner (Discovery CT750 HD scanner, GE Medical Systems, Milwaukee) using a single-source dual-energy spectral imaging mode. The scanning volume covered the area from the skull base to the inferior edge of the aortic arch. The CT scanning parameters for the fast kVp-switching between 80 and 140 kVp were: fixed tube current, 260 mA; helical pitch, 0.984:1; and rotation speed, 0.7 s/r. All patients received a total of 90 mL of non-ionic contrast material (300 mgI/mL, Ultravist, Bayer Germany) at a rate of 3 mL/s administered by power injector. The scan delay was 45 s after injection. The spectral CT allowed the reconstruction of both the polychromatic images corresponding to 140 kVp and 101 sets of contiguous 1.25mm-thick monochromatic images with energies ranging from 40 keV to 140 keV. Both types of images were transferred to a GEAW4.6 workstation.
Spectral analysis was then performed using the Gemstone Spectral Imaging (GSI) Viewer imaging analysis software on the GEAW4.6 workstation. For image analysis purposes, CT images were divided into conventional CT images (polychromatic images corresponding to 140 kVp) and spectral CT images (monochromatic image-sets with energy levels from 40 keV to 140 keV and iodine-based images). Using the normal thyroid image as background, the mean CT number and its standard deviation were measured by placing an oval region of interest (ROI) in the lesion and normal thyroid area. The contrast-noise-ratio (CNR) was calculated by using the following equation: CNR = (CTthyroid-CTlesion)/SDthyroid, where CTthyroid and CTlesion were the mean CT numbers of normal thyroid and lesion in the ROIs, and SDthyroid was the standard deviation of the normal thyroid. For the monochromatic image-sets, GSI Viewer automatically calculates the CNR curve as function of photon energy from 40 keV to 140 keV. From the CNR curve, the optimal energy level of monochromatic image for getting the highest CNR could be determined.
Objective evaluation
The standard deviations of the average values from a circular or ovoid ROI drawn in neck muscles were used to compare the noise of the polychromatic images and the monochromatic images.
Subjective evaluation
Two radiologists (L.L and D.H.L) with 16 and 28 years of experience, respectively, in the diagnosis of head and neck neoplasm, both with fellowship training, analyzed all CT images independently. Polychromatic or monochromatic images were presented in a random order with the patients’ information and CT parameters removed from the images. Agreement was achieved between the 2 radiologists. First, we used a five-point scale to determine the detection ability for the small lesions of the 2 groups: 4 points, excellent for clear lesions and satisfactory details; 3 points, good for clear lesions and details; 2 points relatively clear lesions, decreased confidence in details; 1 point, confidence in details decreased, diagnosis questionable; and 0 points, no lesion can be displayed.
Second, we determined the qualitative diagnostic capability by observing the position, shape, boundary, density, calcification, and lymph node metastasis of the lesion in the images. According to the literature and clinical experience [11], we made the CT diagnostic criteria for PTMC as follows: irregular shape, blurred boundaries, fine granular calcifications internally, heterogeneous enhancement, and lymph node metastasis. The standard for micro-calcifications was diameter≤2 mm [12]. We divided the cervical lymph nodes into the internationally-accepted 7 levels [13]: Level I, the submental and submandibular nodes; Level II, the upper internal jugular nodes; Level III, the middle jugular nodes; level IV, the low jugular nodes; level V, the nodes in the posterior triangle; Level VI, the upper visceral nodes (central node); and level VII, the superior mediastinal nodes. Lymph nodes were characterized as highly suggestive of metastasis on the basis of any of the following criteria: the minor axis diameter of a node in the jugular chain was equal or greater than 6 mm, the minor axis diameter of a node in level VI was equal or greater than 5 mm, or a lymph node with typical signs of thyroid carcinoma metastasis like marginal enhancement, cystic changes, cyst papillary nodule, and micro-calcifications [14–18].
Ultrasonography scan and imaging analysis
The patients were made to lie in the supine position and were instructed to extend the neck as far as possible to facilitate imaging of the thyroid. Ultrasonography of the thyroid was performed with a 7 to 12 MHz linear array transducer (GE Logiq 9) by an experienced radiologist (Y. W). After obtaining a satisfactory image of the thyroid gland, we observed and recorded the number, size, shape, location, boundaries, internal echo, calcification, and blood flow distribution characteristics of nodules in the gland and cervical lymph nodes. Based on the information in the literature and our clinical experience [5], we made the ultrasound diagnostic criteria for PTMC as follows: irregular shape, blurred boundaries, uneven echo, hyperechoic spot, and cervical lymph node metastasis. The diagnostic criteria for cervical lymph node metastasis detection by ultrasound were similar to those of the CT scanning. Lacking contrast compared to CT, flow signal was only a reference to radiologists.
Statistical analysis
Statistical analyses were performed using SPSS software 19.0 (IBM Corporation, Armonk, NY, USA). Count data were expressed as a percentage and measurement data were expressed as the mean±standard deviation. The postoperative pathology result was considered as the gold standard. The detection rate and the capability of qualitative diagnosis to the PTMC were calculated to evaluate the diagnostic performance of each CT (conventional CT and spectral CT) and ultrasound finding. The detection rate = (the number of detected lesions / the number of pathological diagnosis lesions)×100%. For the diagnosis of lymph nodes, sensitivity = (the levels of lymph nodes diagnosed consistent with pathology/ the levels of pathological diagnosis metastatic lymph nodes)×100%. Specificity could not be calculated in our study because the non-metastatic lymph nodes diagnosis by CT and ultrasound images cannot be confirmed by the gold standard method. We used the paired t test to compare the subjective evaluation and objective evaluation. Statistical differences between CT and ultrasound were analyzed using the frequency distribution of χ2 test. P < 0.05 was considered statistically significant.
Results
Forty-six lesions in 31 patients (9 men, 22 women; age range, 22–73 years; mean age, 39.4 years) were confirmed as PTMC by postoperative pathologic examination. Pathological analysis revealed that single lesions were found in 19 patients (41.3%), bilateral malignancy was found in 9 patients (19.6%), and 3 tumours were found in 3 patients (6.5%). Twenty-eight lesions (60.9%) were located in the right lobe and 18 lesions (39.1%) were located in the left lobe. The mean size of all lesions was 0.69±0.04 cm (range, 0.2–1.0 cm).
Comparison of image quality between polychromatic and monochromatic images
The optimal energy level for obtaining the best lesion-to-thyroid contrast-noise-ratio (CNR) in spectral CT was at 65.83±2.01 keV. To conduct analyses expediently, all monochromatic images of group B were assessed at 65 keV. Image quality comparisons between the polychromatic images and monochromatic images are summarized in Table 1. Compared to the polychromatic images (conventional CT), monochromatic images (spectral CT) had a lower image noise and a higher CNR. The image quality scores for polychromatic images and monochromatic images were 2.48±1.35 and 3.09±1.17, respectively, and the difference was statistically significant (t = –5.147, P < 0.05). This result indicated that the image quality of monochromatic images was superior to that of polychromatic images.
Comparison of image quality between polychromatic images and monochromatic images
Comparison of image quality between polychromatic images and monochromatic images
*6 cases which unclear showed on polychromatic images were not included in the contrast-noise-ratio(CNR) calculation.
The polychromatic images, monochromatic images combined with iodine-based images, and high frequency ultrasound showed 40, 44, and 45 lesions with detection rates of 87.0%, 95.7%, and 97.8%, respectively. The monochromatic images combined with iodine-based images were superior to the polychromatic images in terms of the detection rate of lesions (x2 = 13.94, P = 0.000). However, the detection rates of lesions by spectral CT and ultrasound imaging were similar, and the finding was not statistically significant (x2= 0.046, P = 0.829). These findings indicated that, in terms of the lesion detection rate, spectral CT was superior to conventional CT and similar to ultrasound.
Comparison of manifestations for microcarcinoma, including shape, boundaries, density (echo), and microcalcifications between spectral CT and ultrasound
Among the 46 lesions, 2 cases were displayed unclearly in spectral CT. One case was unclear in ultrasound. There were no significant differences in the manifestations using the 2 different diagnostic methods to display the shape, boundaries, density (echo), and microcalcifications of microcarcinoma (P > 0.05) (Table 2 and Figs. 1, 2).
Polychromatic, monochromatic, iodine-based, and high frequency ultrasound images of a papillary thyroid microcarcinoma of the right lobe of the thyroid gland. A. Conventional polychromatic image shows a low-density tumor (black arrow) in the right lobe of the thyroid. B. Monochromatic image (65 keV) shows a more conspicuous lesion (black arrow). C. Iodine-based image shows a nodule (white arrow) in the right lobe of the thyroid with an irregular shape and blurred boundary, characteristic of the local iodine deficiency. D. High-frequency ultrasound shows a nodule (black arrow) in the right lobe of the thyroid with irregular shape and blurred boundaries; microcalcifications can be seen.
Polychromatic, monochromatic, and high-frequency ultrasound images of papillary thyroid microcarcinomas. A. On the conventional polychromatic image, no obvious lesion is visible. B. Monochromatic imaging (65 keV) shows a low-density tumor (black arrow) in the right lobe of thyroid with irregular shape and blurred boundary; microcalcifications (white arrow) can be seen. C. High frequency ultrasound showsa nodule (black arrow) in the right lobe of thyroid with irregular shape and blurred boundary; microcalcifications (white arrow) can be seen.
Comparison of diagnostic performance for microcarcinoma between spectral CT and ultrasound
*2 cases which missed by spectral CT were not included in the data. +Statistical differences were analyzed using McNemar exact test. A two-tailed p < 0.05 was considered significant.
All patients underwent central level dissection. Lateral level dissection was performed selectively on patients with nodes preoperatively diagnosed as lymph node metastasis. Among 31 patients with PTMC, 27 patients were diagnosed with lymph node metastasis by spectral CT and ultrasound before surgery. Sixty-two central levels and 53 lateral levels were surgically resected. Based on the pathological findings, 25 patients (80.6%) were confirmed to have lymph node metastasis involving 83 levels in total (II = 13, III = 19, IV = 18, V = 1,VI = 31, and VII = 1; left and right side were calculated separately). The minor axis diameter of the lymph nodes ranged from 0.4 cm to 2.3 cm. Preoperative spectral CT and ultrasound were able to detect lymph node metastases in 68 and 63 levels, respectively. When compared with the pathological results, it was noted that spectral CT and ultrasound had missed diagnoses in 15 and 20 levels and misdiagnoses in 2 and 3 levels, respectively. The sensitivity of ultrasound to diagnose cervical lymph node metastasis was 75.9% and the sensitivity of spectral CT to diagnose cervical lymph node metastases was 80.7% (Table 3). The sensitivity of spectral CT imaging was higher than that of ultrasound imaging in level VI. Additionally, there was 1 upper mediastinal lymph node that CT detected while ultrasound did not.
Comparisons of diagnostic sensitivity of cervical lymph node metastasis (LNM) between spectral CT and ultrasound
Comparisons of diagnostic sensitivity of cervical lymph node metastasis (LNM) between spectral CT and ultrasound
*Statistical differences were analyzed using McNemar exact test.
Among the 46 lesions confirmed by pathology, 30, 36, and 40 lesions were diagnosed correctly by conventional CT, spectral CT, and ultrasound, respectively, and the sensitivities were 65.2%, 78.3%, and 87.0%, respectively. Compared to conventional CT, spectral CT had higher sensitivity (P = 0.031). However, there was no significant difference in the sensitivity of spectral CT and ultrasound imaging (P = 0.125).
Discussion
Since typical clinical manifestations are not seen in most patients with PTMCs, preoperative diagnosis of PTMCs relies heavily on imaging findings. Currently, ultrasound and CT are the 2 main, commonly used diagnostic methods in the detection and characterization of thyroid disease in the clinical setting. Ultrasound, an excellent, non-invasive, and cost-effective diagnostic method, operates conveniently and displays the size, shape, and boundaries of the thyroid clearly. It can also clearly show small lesions and microcalcifications. Thus, ultrasound becomes the first choice and an important method of preoperative imaging examinations of the thyroid [5]. However, CT scanning has advantages over ultrasound in showing lymph nodes in the retrosternal area, upper mediastinal area, or tracheoesophageal groove [6, 11]. Thus, CT scanning has become an important preoperative method in the diagnosis of thyroid cancer [11, 18].
Although CT images have high resolutions, the beam hardening artefacts caused by the clavicle and contrast agents by intravenous injection in the conventional polychromatic images influence the detection of small lesions to a certain extent. Compared to conventional CT, spectral CT can offer monochromatic images at different energy levels, which can provide more diagnostic information and reduce hardening artifacts [8]. Therefore, spectral CT imaging theoretically can increase the detection rate of small tumours. Regarding image quality and lesion detection, our research showed that the spectral CT images were superior to the conventional polychromatic images in both the subjective and objective evaluations. On the CT image, when the difference in density between the lesion and surrounding normal tissue is small, the lesion cannot be displayed clearly. Due to the change in iodine content and formation of low-density areas, iodine-based images can detect lesions with local iodine deficiency, even though the lesion is very small. We conducted this study to compare the diagnostic sensitivity of ultrasound with that of spectral CT in preoperative evaluation of primary tumours and cervical lymph nodes in patients with PTMC and to determine whether spectral CT can add diagnostic value in the evaluation of these patients. In our study, spectral CT misdiagnosed 2 primary tumours because of their small size (the pathological sizes of the two specimens were 2 mm and 3 mm, respectively), and 1 case was misdiagnosed by ultrasound because it was iso-echoic. Ultrasound can visualize thyroid nodules in enough detail for characterizing lesions and identifying calcifications. However, the results depend on the skill of the operators. In our study, there were no significant differences in the results obtained using the 2 different diagnostic methods to display the shape, boundaries, density (echo), and microcalcifications of a microcarcinoma.
For successful surgical management of PTMC, accurate preoperative determination of the presence of lymph node metastasis is important. In our research, 83 levels in total (both sides were calculated separately) were confirmed with lymph node metastasis based on pathological findings, which were mainly distributed in level III, IV, and VI. Referring to the preoperative diagnostic criteria, preoperative spectral CT and ultrasound diagnosed lymph node metastases in 67 and 63 levels, respectively. The sensitivity of spectral CT images or ultrasound for diagnosing cervical lymph node metastasis was 80.7% and 75.9%, respectively. Spectral CT demonstrated a slight advantage in sensitivity. In our study, we also found that the sensitivity of spectral CT was similar to ultrasound for detection of lymph nodes in the jugular chain. However, for lymph nodes in level VI, the sensitivity of CT was superior to ultrasound. Additionally, there was 1 upper mediastinal lymph node that CT detected while ultrasound did not. These findings indicate that CT has an advantage over ultrasound in displaying lymph nodes of the 2 levels. The main reasons for misdiagnosis using CT were due to the diameter of metastatic lymph nodes being less than the preoperative diagnostic criteria and without typical specific performances in enhanced scanning. As for ultrasound, the main reasons for misdiagnosis were due to the small lymph nodes and the existence of blind spots in the exploration region. Choi’s research [18] on detecting lymph node metastasis of papillary thyroid carcinomas showed that no significant difference was observed between ultrasound and CT, which is similar to the findings of this study.
Spectral parameters of thyroid diseases reported in the literature include iodine content, standardized iodine content, spectral HU curve slope, etc. Comprehensive analysis of those parameters contributes to the quantitative diagnosis of thyroid diseases [19–21]. However, besides the underlying disease of the thyroid and thyroid function, enhanced scan delay time and individual differences influence the iodine content. In addition, there is an overlap between benign and malignant lesions in reference value of iodine content. Hence, in our study, we did not measure the iodine content. Instead, we focused mainly on the qualitative aspect, i.e., the use of imaging features primarily to make diagnoses. It is worth mentioning that we adopted a tube current of 260 mA in this study, which was far lower than the 550–600 mA used in previous studies [19–21]. Furthermore, the application of single-phase enhanced scanning reduced the radiation dose to the patients.
Forty-six lesions in 31 patients were confirmed to be PTMC based on the findings of the postoperative pathological examination. Combined with primary tumour and metastatic lymph node performances, 31, 36, and 40 lesions were diagnosed as PTMC by conventional CT, spectral CT, and ultrasound, respectively. The diagnostic sensitivities of conventional CT, spectral CT, and ultrasound were 67.4%, 78.3%, and 87%, respectively. Compared to the conventional CT, the sensitivity for PTMC of spectral CT for PTMC was significantly improved and was close to that of ultrasound.
There are also a number of limitations to this study. For example, a few patients with small lesions in the thyroid gland chose observation rather than surgical treatment, which led to the inclusion of only small number of subjects in our study. In our further studies, we should expand the sample size and add benign cases as a control group.
In conclusion, spectral CT images were superior to conventional polychromatic images and similar to high frequency ultrasound in terms of the detection and diagnosis rates of PTMC. Ultrasound has the advantage of displaying primary tumours while spectral CT has the advantage of displaying the lymph nodes in level VI and upper mediastinum. Both spectral CT and ultrasound were able to produce good images, and hence may help in the preoperative diagnosis of PTMC.