Circulating Tumor Cells Correlate with Clinicopathological Features and Outcomes in Differentiated Thyroid Cancer

Although the incidence of differentiated thyroid cancer (DTC) has increased worldwide, it has an excellent prognosis, with a ∼90% 10-year overall survival rate [1] However, for DTC patients with distant metastases (DMs), quality of life and survival time are seriously diminished [2, 3]. DTC with DM is commonly treated with radioiodine (¹³¹I), but about 30% of such patients have radioiodine-refractory DTC (RR-DTC) [4], which has a 10-year survival rate of only ∼10% from DM detection [5]. Obviously, patients with RR-DTC do not significantly benefit from ¹³¹I treatment, and may suffer adverse effects from high serum TSH stimulation after thyroxin withdrawal [6]. Although detection of serum thyroglobulin (Tg) levels was a reliable predictor of DTC DMs, this method is limited by metastatic lesions’ ability to release detectable amounts of Tg, loss of the ability of secreting Tg with preserved capability of ¹³¹I trapping, structural changes in Tg, or Tg level decreased by elevated circulating Tg antibodies (TgAb) [7, 8] ]. Also, Tg levels do not predict RR-DTC [9] . Therefore, tumor markers that can help detect DTC DMs and predict RR-DTC are sorely needed.

Circulating tumor cells (CTCs) are rare tumor cells that escape from solid tumors, travel into peripheral blood circulation, and seed DMs. They have been widely used for early DM detection, monitoring prognoses and treatment efficacy, and risk stratification for some solid tumors [10, 11] . The CellSearch system is the most widely used assay for CTCs, and depends on expression of CTC-surface epithelial cell adhesion molecule (EpCAM) [12]. However, the CellSearch system has a low detection rate for DTC CTCs, probably due to their low EpCAM expression during epithelial–mesenchymal transition (EMT) [13]. Negative enrichment (NE) combined with immunofluorescence and in situ chromosomal hybridization (NE-iFISH) is reportedly effective in detecting CTCs from gastric or pancreatic cancers [14] [15], but its use in detecting CTCs from metastatic DTC has not been reported.

We hypothesized that CTCs in DTC patients are associated with DMs, response to ¹³¹I treatment, and prognosis. In this study, we detected the numbers of CTCs in DTC patients using the NE-iFISH and explored whether CTC levels were related to clinicopathological factors or outcomes in patients with DM⁺ DTC.

This study was performed in accordance with the ethical standards of our hospital.

Informed consents were obtained from all individual participants was used in the study.

The study was performed at the Shanghai Sixth People’s Hospital, a major ¹³¹I treatment center in China, and was approved by its ethics committee. We enrolled 42 patients with DM⁺ DTC treated between August 2014 and August 2017, and 30 patients with DM^- DTC patients treated during August 2017. The inclusion criteria were (a) patients with histologically proven DTC (follicular thyroid cancer [FTC] or papillary thyroid cancer [PTC]); and (b) treatment with total or near-total thyroidectomy and postoperative ¹³¹I ablation. Patients with histories of other malignancies or other adjuvant therapy were excluded. The DTC patients were classified as having intermediate high risk [16]. We also included 30 healthy volunteers, with no illness or past cancer history, as a healthy control (HC) group. All participants gave informed written consent.

Baseline data were recorded when peripheral blood samples were collected for CTC detection, including age, sex, maximum tumor diameter, DTC pathology, extrathyroidal invasion, numbers of primary tumors, N stage, serum stimulated thyroglobulin (sTg), and TgAb (Table 1); after each ¹³¹I therapy and during the follow-up, results of sTg, TgAb, ¹³¹I whole-body scan (¹³¹I-WBS) combined with SPEC/CT, other medical imaging such as CT and ¹⁸F-FDG PET/CT, courses ¹³¹I treatment, cumulative ¹³¹I activity, sites and numbers of DMs and follow-up time were recorded.

Diagnosis of DMs from DTC depended on Tg, TgAb under T4 withdrawal for 2–3 weeks, ¹³¹I-WBS combined with SPECT/CT after ¹³¹I therapy, 18F-FDG PET/CT, CT, magnetic resonance imaging (MRI) or fine needle biopsy results; and was based on the following three criteria: (a) pathological results showed DMs; (b) ¹³¹I-WBS demonstrated distant lesions with high ¹³¹I uptake, with at least one imaging method (CT, MRI, or ¹³¹I-SPECT/CT) showing DMs; or (c) ¹³¹I-WBS reveals no ¹³¹I uptake, but the 18F-FDG-PET/CT scans are positive for DM lesions with elevated serum Tg.

Each patient stopped taking L-T4 and undertook a low-iodine diet for 2–3 weeks before ¹³¹I therapy. Subsequently, the patients received oral ¹³¹I after routine measurements that included FT3, FT4, TSH, Tg, TgAb, neck ultrasonography (US), and CT scans. The first oral dose of 3.7 GBq (100 mCi) of ¹³¹I was administered to ablate thyroid remnants. Oral doses of ¹³¹I with standard activities of 5.55–7.4 GBq (150– 200 mCi) were administered for subsequent treatment of DTC DMs. Patients underwent ¹³¹I-WBS combined with ¹³¹I-SPECT/CT imaging 3–5 days after ¹³¹I oral administration. Treatment intervals ranged from 4–6 months; treatments were repeated 1–6 times. Median follow-up period was 1.2 years (range: 0.2–3 years).

According to the 2015 American Thyroid Association management guideline [16], radioiodine-refractory structurally evident DTC in patients with DMs who were under appropriate TSH stimulation and iodine preparation is classified by 4 mechanisms: (i) metastatic lesions do not ever uptake ¹³¹I after successful remnant thyroid ablation; (ii) tumor tissue loses the ability to concentrate ¹³¹I after previous evidence of RAI-avid disease (in the absence of stable iodine contamination), (iii) ¹³¹I is concentrated in some lesions but not in others; and (iv) metastatic disease progresses despite significant concentration of ¹³¹I in the course of ¹³¹I treatment.

Progressive disease (PD), stable disease (SD), partial response (PR), and complete response (CR) to ¹³¹I treatment for DTC lung and bone metastases were evaluated based on the Response Evaluation Criteria in Solid Tumors (RECIST 1.1) [17] and MDA criteria [18] respectively.

RECIST 1.1: complete response (CR), disappearance of all target lesions, any pathological lesions (target or non-target) must have a reduction in short axis to < 10 mm; partial response (PR), ≥30% decrease in the diameters of target lesions; progressive disease (PD), ≥20% increase in the diameters of target lesions, combined with an absolute increase of ≥5 mm in the sum of diameters (the appearance of one or more new lesions was also considered progression); stable disease (SD), neither sufficient shrinkage to qualify for PR nor sufficient increase to qualify for PD.

MDA criteria: 1) CR: complete sclerotic fill-in of lytic lesions on X-ray or CT; normalization of bone density on X-ray or CT; normalization of signal intensity on MRI. 2) PR: development of a sclerotic rim or partial sclerotic fill-in of lytic lesions on X-ray or CT; interval visualization of lesions with sclerotic rims or new sclerotic lesions in the setting of other signs of PR and absence of progressive bony disease; ≥50% decrease in measurable lesions on X-ray, CT, or MRI; ≥50% subjective decrease in the size of ill-defined lesions on X-ray, CT, or MRI. 3) SD: no change; < 25% increase or < 50% decrease in size of measurable lesions; < 25% subjective increase or < 50% decrease in size of ill-defined lesions; no new bone metastases. 4) PD: ≥25% increase in size of measurable lesions on X-ray, CT, or MRI; ≥25% subjective increase in the size of ill-defined lesions on X-ray, CT, or MRI; new bone metastases.

PFS was measured from the date of baseline blood sample to the date of confirmed clinical progression or death was censored at last follow-up.

Enrichment of CTCs was performed according to NE-iFISH kit instruction (Majorbio, Shanghai, China) and according to a previously published protocol [14, 15]. In brief, peripheral blood samples (7.5 ml) were collected into ACD anticoagulant tubes when all patients ceased thyroid hormone medication and began low-iodine diets, more than 2–3 weeks before the first ¹³¹I therapy. The supernatant was discarded after centrifuging the tubes within 24 hours after sample collection. Samples were transferred to centrifuge tubes containing 3 ml of lymphocyte separation media (Majorbio). After centrifuging for 10 min at 450 g, the cell suspension was collected from the buffy-coat layer. Immunomagnetic particles conjugated anti-CD45 antibody (Majorbio) were added into the cell suspension, which was incubated at room temperature for 10 min and then placed on a magnetic stand (Majorbio) until the liquid became clear. The supernatant was pipetted off the magnetic field to remove leukocytes by centrifuging at 900 g for 5 min.

The identification of CTCs was performed according to mi-FISH kit instruction (Majorbio). To identify aneuploid CTCs, fluorescence in situ hybridization (FISH) and immunocytochemistry were used in combination. The supernatant was rinsed with wash buffer, and then monoclonal antibody anti-CD45 conjugated to Alexa Fluor 594 (Majorbio) and anti- anti-EpCAM Antibody conjugated to Alexa Fluor 488 (Majorbio) were added before incubation at room temperature for 30 min. The supernatant was rinsed with wash buffer. To the obtained cellular precipitation, we immediately added cell fixatives-1 (Majorbio) before making smears. The cell smears were dried at 32 ℃ overnight. We added 200 μl of cell fixatives-2 (Majorbio) to the cell smears and let stand for 10 min. After rinsing with wash buffer, the mixture was put into in a 100% alcohol after washing and let stand for 1 min. After the slide was air-dried, 10 μl of probe solution containing fluorescence-labeled alpha-satellite probes for Centromere Probe 8 (CEP8) SpectrumOrange (Vysis, Abbott Laboratories, Abbott Park, IL, USA) was added and then covered with a coverslip and sealed with neutral resin (rubber). The hybridization procedure was as follows: degeneration at 75 °C for 5 min, followed by hybridization at 37 °C for 4 hours. After rinsing with PBS, the slides were mounted with mounting medium containing DAPI and photographed with a fluorescence microscope (Zeiss, German). CTCs were confirmed to be negative for CD45 and either positive for EpCAM staining or aneuploidy chromosome 8. CTCs were defined as EpCAM+/CD45–/DAPI+/CEP8≥2 or EpCAM–/CD45–/DAPI+/CEP8> 2, whereas EpCAM–/ CD45+/DAPI+/CEP8=2 was defined as white blood cells (WBCs) [14, 15]. All CTC numbers in this paper are the numbers of CTCs counted in 7.5-ml samples of subjects’ peripheral venous blood.

Student’s t-test was used to test for significant differences in numbers of CTCs among the DM+ DTC, DM– DTC and HC groups. Receiver operating characteristic (ROC) curves were constructed, and area under curve (AUC) was calculated to evaluate specificity and sensitivity of predicting DMs from DTC or RR-DTC. The chi-square test or Fisher’s exact test and multivariate logistic regression analyses were used to evaluate CTC numbers with respect to clinicopathological factors. The Kaplan–Meier method was used for PFS analysis; differences in PFS were examined using the log-rank test. Multivariate analysis was performed to assess relationships between PFS and several variables simultaneously using Cox’s proportional hazards model. P< 0.05 was considered significant. MedCalc software version 17.0 (MedCalc, Mariakerke, Belgium) and SPSS17.0 (SPSS Inc., Chicago, IL, USA) were used for statistical analyses.

The CTC counts for the HC, DM⁻ DTC and DM⁺ DTC groups are shown in Fig. 1 and Fig. 2a.

The CTCs were defined as EpCAM⁺/CD45⁻/DAPI⁺/CEP8≥2, which weren’t be detected in the 7.5ml blood. The numbers of detected CTCs (per 7.5 ml blood) in the DTC patients were median: 4; range: 0–72; mean: 7.31, total: 527; and were found in 62 of the 72 patients (86.1%). In the DM+ DTC group, CTCs were detected in 92.9% (39/42) of patients (mean: 10.69; range: 0–72). In the DM- DTC group, CTCs were in 76.7% (23/30) of patients (mean: 2.6; range: 0–10). In the HC group, CTCs were found in 26.7% ( 8/30) of subjects (mean: 0.4; range: 1–3, including 1 CTC: n=5; 2 CTCs: n=1; 3 CTCs: n= 2). Mean CTC numbers were significantly higher in the DM⁺ DTC than in the HC group (P< 0.001) and DM⁻ DTC group (P=0.0016); and higher in the DM⁻ DTC group than in the HC group (P< 0.001).

We plotted a ROC curve to compare the sensitivity and specificity of CTC numbers with respect to DM⁺ DTC and DM^- DTC, using 3.5 CTCs (71.4% sensitive; 73.3% specific), 4.5 CTCs (64.3% sensitive; 83.8% specific) and 5.5 CTCs (61.9% sensitive; 90% specific) as possible cut-off values (Youden’s index). We therefore defined the cutoff at 4.5 CTCs (AUC: 0.789; 95% CI: 0.685–0.893; Fig. 2c). In the DM⁺ DTC group, 27 (64.3%) had more, and 15 had fewer, than 4.5 CTCs, whereas in the DM^- DTC group, 5 (16.7%) had more CTCs and 25 had fewer (P< 0.001). Therefore, CTCs ≥ 5 was considered predictive of DMs.

Patient age at baseline, sex, DTC pathology, tumor size, extrathyroidal invasion, multifocality and N stage were not significantly associated with CTCs ≥ 5. In univariate regression, CTCs ≥ 5 was significantly associated with sTg > 200 ng/ml (odds ratio [OR]: 3.667, 95% CI: 1.370–9.810) and DM⁺ DTC (OR: 0.111, 95% CI: 0.035–0.351) in binary logistic regression, only DM+ DTC was independently associated with CTCs ≥ 5 (P=0.009; OR compared with DM^- DTC: 0.028 (95% CI: 0.002–0.403, Table 2).

Clinical characteristics of the DM⁺ DTC group (n=42) are shown in Table 3 and Table 4. Their median age was 43.2 years (range 12–71 years); they included 17 males and 25 females; 33 had PTC and 11 had FTC; 5 had only bone metastases, 24 had only lung metastases, 11 had synchronous bone and lung metastases; 1 had synchronous bone and brain metastases, and 1 had synchronous bone and kidney metastases.

We classified 19 DM⁺ DTC patients as RR-DTC and 23 as non-RR-DTC (2015 ATA guidelines). Mean numbers of CTCs in the RR-DTC group (16.8) were significantly higher than in the non-RR-DTC group (5.60; P=0.0049; Fig. 2b).

To determine the CTC count cutoff that would be most indicative of RR-DTC, we tested 4.5 CTCs (sensitivity: 84.2%; specificity: 52.2%), 5.5 CTCs (sensitivity: 78.9%; specificity: 52.2%) and 6.5 CTCs (sensitivity: 73.7%; specificity: 69.6%), and finally used 6.5 CTCs as the cut-off number (AUC=0.781; 95%CI 0.635–0.928; Fig. 2d). In the DM⁺ RR-DTC group (n=19), 14 (73.6%) patients had more CTCs and 5 had fewer than the 6.5 CTC cut-off; whereas in the DM^- non-RR-DTC group (n=23), 7 (30.4%) had more CTCs and 16 had fewer than the cut-off (P< 0.001). Therefore, CTCs ≥7 was considered indicative of RR-DTC.

Univariate analysis of factors considered likely to be associated with RR-DTC (or specifically, CTCs ≥7) are shown in Table 4. Only RR-DTC was significantly associated with CTCs ≥7 (P=0.005; OR compared with non-RR-DTC: 0.156, 95% CI: 0.040–0.605). In multivariate logistic regression analysis, patient age at baseline, sex, DTC pathology, tumor size, extrathyroidal invasion, multifocality, site of metastases, and presence of DM at diagnosis were not associated with CTCs ≥7; whereas CTCs ≥7 was associated with RR-DTC (P=0.007; OR: 0.0451, 95% CI: 0.0047–0.4345) and sTg level ≤ 500 ng/ml (P=0.044; OR: 0.087, 95% CI: 0.008–0.932; Table 4).

Three-year PFS rates and prognostic factors in the DM⁺ DTC group (n=42) are listed in Table 5 and Table 6. All 42 DM⁺ DTC patients had completed at least one course of ¹³¹I therapy. At the time of analysis, 19 (45.2%) of the 42 patients had PD, 2 achieved CR (4.8%), 4 PR (9.5%), 17 (4.0%) SD. Median PFS was 1.2 years (range: 0.2–3 years). At the time of follow-up, 4 patients had died of metastatic lesions, and the remaining 38 patients were alive.

In univariate analysis, CTCs≥7 and RR-DTC were strong predictors of worse 3-year PFS in all DM⁺ DTC patients compared with CTCs< 7 and non-RR-DTC. In this study, the 3-year PFS rates were CTCs ≥7: 33.3%; CTCs < 7: 76.2% (Fig. 3a); non-RR-DTC group: 78.3%; RR-DTC group: 26.3% (Fig. 3b). In multivariate Cox regression analyses, PFS was significantly associated with CTCs ≥7 (relative risk [RR]: 5.108, 95% CI: 1.334–19.550), sTg level ≤ 500 ng/ml (0.216, 95% CI: 0.050–0.936) and non-RR-DTC (RR: 0.093, 95% CI: 0.015–0.575).

Circulating biomarkers can potentially provide early and minimally invasive diagnoses help assess response to therapy and predict prognosis of cancer patients. Studies on the utility of circulating biomarkers such as cell-free DNA, miRNA, and mRNA as novel biomarkers for various human diseases have increased exponentially in the last decade [6, 19]. Recent studies have shown CTCs to be prognostic and predictive biomarkers for several solid cancers [20, 21]. CTCs are malignant cells that have escaped from primary or metastatic tumor sites [22]. CellSearch is the only FDA-approved technology for CTC detection, which is based on detecting EpCAM expression [23, 24]. Few studies of the prognostic significance of CTCs from DTC are available. Nonetheless, Xu et al. found that only 1 of 14 patients with DM+ DTC had detectable CTCs using CellSearch [12]. Therefore, detecting CTCs in DTC patients requires other approaches.

In the present study, we applied a new capture technique (NE–iFISH) to detect DTC CTCs and to assess their relationships with clinicopathological factors, therapeutic repose to ¹³¹I and prognosis of DTC. Compared with conventional CellSearch, NE–iFISH has demonstrated high sensitivity for CTCs from other cancer types [25, 26].

NE–iFISH has several advantages. On one hand, it applies anti-multiple WBC markers antibodies to ensure both minimum hypotonic injuries to CTCs and maximal removal of WBCs; on the other hand, it can enrich CTCs independently of EpCAM expression and tumor size [27]. We were also able to combine CEP 8, EpCAM, CD45 and DAPI to detect CTCs. Aneuploidy, which includes both numerical and structural chromosomal abnormalities, is a hallmark of cancer, and might be exploited for CTC detection. Numerical abnormalities of CTCs can be divided into different subtypes, such as triploidy, tetroploidy and multiploidy. Previous studies have shown that CEP 8 reflects numerical chromosomal abnormalities, which have been reported in thyroid cancer, and lung, esophageal, gastric and colon cancers [27-29]. Thus, aneuploidy of CEP 8 examined by CEP8-iFISH can be used to detect CTCs. In our study, of 72 DTC patients, EpCAM^-/CD45^-/DAPI⁺/CEP8> 2 cells were detected in 86.11% (62/72) patients (mean: 7.14; range: 0–72 CTCs) compared with the HC group, in whom only 26.7% (8/30) subjects had EpCAM^-/CD45^-/DAPI⁺/CEP8> 2 cells, and none had more than 3 such cells. Obviously, DTC patients had significantly more CTCs than did the HC group. However, EpCAM⁺/CD45^-/DAPI⁺/CEP8≥2 weren’t detected in all DTC patients and HC people. Although EpCAM expression has reportedly been found in DTC specimens, those CTCs have not been successfully isolated by means of anti-EpCAM Ab. The reason may be that down-regulated membrane expression and up-regulated nuclear accumulation of EpCAM are associated with EMT during DTC progression and metastasis [13]. This is consistent with the observation of Xu et al. using CellSearch [13].

Patients with breast, prostate, or colorectal cancers with hematogenous metastases reportedly have high incidences of CTCs; CTCs ≥5 in 7.5 ml peripheral venous blood has been widely used as a cut-off to predict DMs from these cancers [23, 30, 31]To predict DMs from DTC we plotted a ROC curve (Yourdon’s index), which gave an optimal cutoff for CTCs at 4.5 (as measured by NE-iFISH) which was 64.3% sensitive and 83.8% specific for predicting DTC DMs. We therefore used CTCs ≥5 to predict DTC DMs, which is consistent with the above studies. Although we also found that CTCs ≥5 were associated with sTg level ≥ 200 ng/ml in univariate analysis, in multivariate analysis, only DM+ DTC was associated with CTCs ≥5.

For RR-DTC patients, ¹³¹I therapy has very limited benefits. Such patients should not undergo unnecessary ¹³¹I treatment, which would needlessly expose them to the risk of high serum TSH stimulation after thyroxin withdrawal. Identification of accurate predictive biomarkers is critical to develop novel ¹³¹I therapeutic strategies. Specifically, noninvasive biomarkers to detect RR-DTC patients are urgently needed before ¹³¹I therapy. However, circulating tumor biomarkers that can predict response to ¹³¹I therapy in patients with DM+ DTC have not been available. Only our previous study has shown that four circulating long non-coding RNAs have potential utility as diagnostic markers for non-¹³¹I-avid lung metastases from DTC [6]. In a review of CTCs, only Winkens et al. reported that ¹³¹I therapy could decrease the number of circulating epithelial cells in the DTC patients [32].

To our knowledge, the present study is the first in which CTCs have been analyzed or compared between non-RR-DTC and RR-DTC patients with DMs. We compared numbers of CTCs between non-RR-DTC and RR-DTC patients using ROC curves and observed whether CTC numbers could predict DM⁺ RR-DTC. Our study found that the optimal cutoff for CTCs was 6.5, which was 73.7% sensitive and 69.6% specific in predicting DM⁺ RR-DTC. Therefore, CTCs in 7.5 ml blood serve as a predictor of DM⁺ RR-DTC. Univariate and multivariate analyses associated CTCs ≥7 in 7.5 ml blood with DM⁺ RR-DTC, and (in multivariate analysis only) with sTg level ≥500 ng/ml.

Radioiodine-refractory character is associated with poor outcome in DM+ DTC. Durante et al. reported that the 10-year survival rate can be 56% in patients with intense radioiodine accumulation, but as low as 10% in those who lose radio-iodine avidity [5]. Previously our group showed that DTC patients with non-¹³¹I-avid disease had a 10-year survival rate of 38.1% compared with 69.2% for those with ¹³¹I-avid metastases [2]. As CTCs ≥7 was associated with DM⁺ RR-DTC, we speculate that CTCs ≥7 may also correlate with DM⁺ DTC prognosis and progression. Nevertheless, due to short follow-up times in our study (median: 1.2 years, range: 0.2-3 years), we did not evaluate overall survival, and only analyzed PFS of DM⁺ DTC patients. In view of this, we compared 3-year PFS rates between non-RR-DTC and RR-DTC patients and found that CTCs ≥7 was significantly associated with lower 3-year PFS than CTCs < 7 (33.3% vs 76.2%). Multivariate Cox-regression analyses confirmed that CTC≥7 is a strong independent predictor of DM⁺ RR-DTC.

This study had some limitations. First, the sample size was relatively small, especially for DM⁺ DTC patients. Second, dynamic monitoring of CTCs wasnt performed in the course of ¹³¹I treatment; which may have missed some measurement deviation. Third, overall survival wasn’t evaluated in these DM⁺ DTC patients because of short follow-up time. Fourth, as our center specializes in ¹³¹I treatment for DTC patients, primary thyroid surgeries often occurred at other institutions. Therefore, N stages for 16 DTC patients were unclear for us. We also didn’t compare sensitivity and specificity of the CTC detection between NE-iFISH and CellSearch technology in these patients.

For the first time, we have developed a strategy to identify CTCs in DTC patients using NE-iFISH technology. The CTC detection rate was 86.1% in our DTC patients; CTCs ≥5 in 7.5 ml blood is correlated with DM+ DTC (sensitivity: 64.3%; specificity: 83.8%). Further, CTCs ≥7 may predict DM⁺ RR-DTC (sensitivity: 73.7%; specificity 69.6 %), and is associated with worse prognosis in DM⁺ DTC. We found that assessing CTCs can help predict metastasis, response to ¹³¹I treatment and prognosis in DTC patients. However, our results need further validation in a larger study.

The authors thank Shanghai Meiji Bio-tech Inc. for their technical support in the CTCs detection and thank Marla Brunker, from Liwen Bianji, Edanz Group China (wwwliwenbianji. cn/ac), for polishing the English text of a draft of this manuscript. This study was sponsored by the National Natural Science Foundation of China (No: 81771865, 81271611), Shanghai Outstanding Youth Doctor Training Program and Shanghai Key Discipline of Medical Imaging (NO: 2017ZZ02005).

The authors have declared that no competing interest existed.

Zhong-Ling Qiu, Wei-Jun Wei and Zhen-Kui Sun contributed equally to this work.