A Deep Learning Model for Predicting Molecular Subtype of Breast Cancer by Fusing Multiple Sequences of DCE-MRI From Two Institutes.

RATIONALE AND OBJECTIVES: To evaluate the performance of deep learning (DL) in predicting different breast cancer molecular subtypes using DCE-MRI from two institutes. MATERIALS AND METHODS: This retrospective study included 366 breast cancer patients from two institutes, divided into training (n = 292), validation (n = 49) and testing (n = 25) sets. We first transformed the public DCE-MRI appearance to ours to alleviate small-data-size and class-imbalance issues. Second, we developed a multi-branch convolutional-neural-network (MBCNN) to perform molecular subtype prediction. Third, we assessed the MBCNN with different regions of interest (ROIs) and fusion strategies, and compared it to previous DL models. Area under the curve (AUC) and accuracy (ACC) were used to assess different models. Delong-test was used for the comparison of different groups. RESULTS: MBCNN achieved the optimal performance under intermediate fusion and ROI size of 80 pixels with appearance transformation. It outperformed CNN and convolutional long-short-term-memory (CLSTM) in predicting luminal B, HER2-enriched and TN subtypes, but without demonstrating statistical significance except against CNN in TN subtypes, with testing AUCs of 0.8182 vs. [0.7208, 0.7922] (p=0.44, 0.80), 0.8500 vs. [0.7300, 0.8200] (p=0.36, 0.70) and 0.8900 vs. [0.7600, 0.8300] (p=0.03, 0.63), respectively. When predicting luminal A, MBCNN outperformed CNN with AUCs of 0.8571 vs. 0.7619 (p=0.08) without achieving statistical significance, and is comparable to CLSTM. For four-subtype prediction, MBCNN achieved an ACC of 0.64, better than CNN and CLSTM models with ACCs of 0.48 and 0.52, respectively. CONCLUSION: Developed DL model with the feature extraction and fusion of DCE-MRI from two institutes enabled preoperative prediction of breast cancer molecular subtypes with high diagnostic performance.

A Novel DNA Aptamer Probe Recognizing Castration Resistant Prostate Cancer in vitro and in vivo Based on Cell-SELEX.

Background: Early recognition of castration-resistant state is of significance for timely adjustment of treatment regimens and improvement of prognosis. Purpose: This study aims to screen new aptamers CRda8 and CRda21 which recognize castration resistant prostate cancer (CRPC) cells with high affinity and specificity by SELEX technology. Methods: The enrichment of specific aptamer candidates was monitored by flow cytometric analysis. The affinity and specificity of aptamer candidates were evaluated by flow cytometry and immunofluorescence assay. MR imaging of CRda21-conjugated polyethylene glycol (PEG)-Fe3O4 nanoparticles to CRPC was further explored in vivo. Results: Both aptamers showed high specificity to target cells with dissociation constants in the nanomolar range, and did not recognize other tested cells. The staining of clinical tissue sections with fluorescent dye labeled aptamers showed that sections from CRPC exhibited stronger fluorescence while sections from benign prostatic hyperplasia and androgen dependent prostate cancer did not exhibit notable fluorescence. In vivo MRI demonstrated that CRda21-conjugated PEG-Fe3O4 had good affinity to CRPC and produced strong T2WI signal intensity reduction distinguished from peritumoral tissue. Conclusion: The high affinity and specificity of CRda8 and CRda21 make the aptamer hold potential for early recognition of castration-resistant state and diagnosis of CRPC at the cellular level.

Synthetic MRI and amide proton transfer-weighted MRI for differentiating between benign and malignant sinonasal lesions.

OBJECTIVES: To explore the value of the synthetic MRI (SyMRI), combined with amide proton transfer-weighted (APTw) MRI for quantitative and morphologic assessment of sinonasal lesions, which could provide relative scale for the quantitative assessment of tissue properties. METHODS: A total of 80 patients (31 malignant and 49 benign) with sinonasal lesions, who underwent the SyMRI and APTw examination, were retrospectively analyzed. Quantitative parameters (T1, T2, proton density (PD)) and APT % were obtained through outlining the region of interest (ROI) and comparing the two groups utilizing independent Student t test or a Wilcoxon test. Receiver operating characteristic curve (ROC), Delong test, and logistic regression analysis were performed to assess the diagnostic efficiency of one-parameter and multiparametric models. RESULTS: SyMRI-derived mean T1, T2, and PD were significantly higher and APT % was relatively lower in benign compared to malignant sinonasal lesions (p < 0.05). The ROC analysis showed that the AUCs of the SyMRI-derived quantitative (T1, T2, PD) values and APT % ranged from 0.677 to 0.781 for differential diagnosis between benign and malignant sinonasal lesions. The T2 values showed the best diagnostic performance among all single parameters for differentiating these two masses. The AUCs of combined SyMRI-derived multiple parameters with APT % (AUC = 0.866) were the highest than that of any single parameter, which was significantly improved (p < 0.05). CONCLUSION: The combination of SyMRI and APTw imaging has the potential to reflect intrinsic tissue characteristics useful for differentiating benign from malignant sinonasal lesions. CLINICAL RELEVANCE STATEMENT: Combining synthetic MRI with amide proton transfer-weighted imaging could function as a quantitative and contrast-free approach, significantly enhancing the differentiation of benign and malignant sinonasal lesions and overcoming the limitations associated with the superficial nature of endoscopic nasal sampling. KEY POINTS: • Synthetic MRI and amide proton transfer-weighted MRI could differentiate benign from malignant sinonasal lesions based on quantitative parameters. • The diagnostic efficiency could be significantly improved through synthetic MRI + amide proton transfer-weighted imaging. • The combination of synthetic MRI and amide proton transfer-weighted MRI is a noninvasive method to evaluate sinonasal lesions.

Automatic deep learning method for detection and classification of breast lesions in dynamic contrast-enhanced magnetic resonance imaging.

Background: The purpose of this study was to develop a deep learning-based system with a cascade feature pyramid network for the detection and classification of breast lesions in dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). Methods: This retrospective study enrolled 191 consecutive patients with pathological confirmed breast lesions who underwent preoperative magnetic resonance imaging (MRI) at the Second Affiliated Hospital of Xi'an Jiaotong University. Patients were randomly divided into a training set comprising 153 patients with 126 malignant and 27 benign lesions and a validation set containing 38 patients with 31 malignant and 7 benign lesions under 5-fold cross-validation. Two radiologists annotated the location and classification of all lesions. After augmentation with pseudo-color image fusion, MRI images were fed into the developed cascade feature pyramid network system, feature pyramid network, and faster region-based convolutional neural network (CNN) for breast lesion detection and classification, respectively. The performance on large (>2 cm) and small (≤2 cm) breast lesions was further evaluated. Average precision (AP), mean AP, F1-score, sensitivity, and false positives were used to evaluate different systems. Cohen's kappa scores were calculated to assess agreement between different systems, and Student's t-test and the Holm-Bonferroni method were used to compare multiple groups. Results: The cascade feature pyramid network system outperformed the other systems with a mean AP and highest sensitivity of 0.826±0.051 and 0.970±0.014 (at 0.375 false positives), respectively. The F1-score of the cascade feature pyramid network in real detection was also superior to that of the other systems at both the slice and patient levels. The mean AP values of the cascade feature pyramid network reached 0.779±0.152 and 0.790±0.080 in detecting large and small lesions, respectively. Especially for small lesions, the cascade feature pyramid network achieved the best performance in detecting benign and malignant breast lesions at both the slice and patient levels. Conclusions: The deep learning-based system with the developed cascade feature pyramid network has the potential to detect and classify breast lesions on DCE-MRI, especially small lesions.

Establishment of Prognosis Model in Acute Myeloid Leukemia Based on Hypoxia Microenvironment, and Exploration of Hypoxia-Related Mechanisms.

Acute myeloid leukemia (AML) is a highly heterogeneous hematologic neoplasm with poor survival outcomes. However, the routine clinical features are not sufficient to accurately predict the prognosis of AML. The expression of hypoxia-related genes was associated with survival outcomes of a variety of hematologic and lymphoid neoplasms. We established an 18-gene signature-based hypoxia-related prognosis model (HPM) and a complex model that consisted of the HPM and clinical risk factors using machine learning methods. Both two models were able to effectively predict the survival of AML patients, which might contribute to improving risk classification. Differentially expressed genes analysis, Gene Ontology (GO) categories, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis were performed to reveal the underlying functions and pathways implicated in AML development. To explore hypoxia-related changes in the bone marrow immune microenvironment, we used CIBERSORT to calculate and compare the proportion of 22 immune cells between the two groups with high and low hypoxia-risk scores. Enrichment analysis and immune cell composition analysis indicated that the biological processes and molecular functions of drug metabolism, angiogenesis, and immune cell infiltration of bone marrow play a role in the occurrence and development of AML, which might help us to evaluate several hypoxia-related metabolic and immune targets for AML therapy.

Selection of DNA Aptamers Recognizing EpCAM-Positive Prostate Cancer by Cell-SELEX for in vitro and in vivo MR Imaging.

BACKGROUND: The sensitive and specific detection of pathogenic cells is important in tumor diagnosis at an early stage. Aptamers are short single-stranded oligonucleotides evolved from systematic evolution of ligands by exponential enrichment (SELEX). It has been proved that aptamers can interact with cognate target molecules with high affinity and specificity and have great potential in the development of medical imaging at molecular level. PURPOSE: To select epithelial cell adhesion molecule (EpCAM) specific aptamers targeting prostate cancer and further to conjugate aptamers with GoldMag nanoparticles (a typical iron oxide core/gold shell structure) to construct magnetic molecular probes for medical imaging. METHODS: EpCAM-specific aptamers were selected by Cell-SELEX. The enrichment of specific aptamer candidates was monitored by flow cytometric analysis. Aptamers were further conjugated with GoldMag nanoparticles to construct magnetic molecular probes. The affinity and specificity of aptamer candidates and aptamer-conjugated GoldMag nanoparticles were evaluated. The MR imaging of aptamer-conjugated GoldMag nanoparticles to prostate cancer was further explored in vitro and in vivo. RESULTS: After 12 rounds of selection, aptamer candidates Eppc6 and Eppc14 could specifically target three types of prostate cancer cells, revealing a high affinity of Eppc6 and Eppc14. Moreover, aptamer-conjugated GoldMag nanoparticles not only exhibited good affinity to different prostate cancer cells but also produced strong T2WI signal intensity reduction distinguished from peritumoral tissue in MRI, indicating that the molecular probes possess both the affinity properties of EpCAM-specific aptamer and the superparamagnetic features of iron oxide. CONCLUSION: Our study indicates that aptamer Eppc6 and Eppc14 can recognize prostate cancer cells and tissues. The aptamer-conjugated GoldMag nanoparticles constructed in the study can be used as a molecular imaging agent for detection of PCa in MRI.

Human Amniotic Mesenchymal Stem Cells Inhibit aGVHD by Regulating Balance of Treg and T Effector Cells.

BACKGROUND: Acute graft versus host disease (aGVHD) remains a leading cause of transplant-related mortality following allogeneic haematopoietic cell transplantation (allo-HCT). Human amniotic mesenchymal stem cells (hAMSCs) are a novel mesenchymal stem cells (MSCs), which have stronger proliferation and immunomodulatory ability compared with bone marrow mesenchymal stem cells (BM-MSCs). Besides, as the amniotic membrane is often treated as medical waste after delivery, hAMSCs can be obtained conveniently and noninvasively. The aim of this study was to explore the therapeutic efficacy and underlying mechanisms of hAMSCs transplantation for the humanized aGVHD mouse model. METHODS: We established a humanized aGVHD mouse model by transplanting human peripheral blood mononuclear cells (PBMCs) into NOD-PrkdcscidIL2rγnull (NPG) mice, human amniotic membrane collected from discarded placenta of healthy pregnant women after delivery and hAMSCs were extracted from amniotic membrane and expanded in vitro. Mice were divided into untreated group (Control), aGVHD group (aGVHD), and hAMSCs treatment group (aGVHD+hAMSCs), the hAMSCs labeled with GFP were administered to aGVHD mice to explore the homing ability of hAMSCs. T effector and regulatory T cells (Tregs) levels and cytokines of each group in target organs were detected by flow cytometry and cytometric bead array (CBA), respectively. RESULTS: We successfully established a humanized aGVHD mouse model using NPG mice. The hAMSCs have the ability to inhibit aGVHD in this mouse model through reduced villous blunting and lymphocyte infiltration of the gut while reducing inflammatory edema, tissue destruction and lymphocyte infiltration into the parenchyma of the liver and lung. hAMSCs suppressed CD3+CD4+ T and CD3+CD8+ T cell expression and increased the proportion of Tregs, and besides, hAMSCs can reduce the levels of IL-17A, INF-γ, and TNF in aGVHD target organs. CONCLUSION: The NPG murine environment was capable of activating human T cells to produce aGVHD pathology to mimic aGVHD as in humans. The hAMSCs controlled aGVHD by decreasing inflammatory cytokine secretion within target organs by modulating the balance of Tregs and T effector cells in humanized mice.

Immunomodulating functions of human leukocyte antigen-G and its role in graft-versus-host disease after allogeneic hematopoietic stem cell transplantation.

Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is a potentially curative therapeutic strategy to treat several hematological malignancies and non-hematological malignancies. However, graft-versus-host disease (GVHD) is a frequent and serious transplant-related complication which dramatically restrains the curative effect of allo-HSCT and a significant cause of morbidity and mortality in allogeneic HCT recipients. Effective prevention of GVHD mainly depends on the induction of peripheral immune tolerance. Human leukocyte antigen-G (HLA-G) is a non-classical MHC class I molecule with a strong immunosuppressive function, which plays a prominent role in immune tolerance. HLA-G triggers different reactions depending on the activation state of the immune cells and system. It also exerts a long-term immune tolerance mechanism by inducing regulatory cells. In this present review, we demonstrate the immunomodulatory properties of human leukocyte antigen-G and highlight the role of HLA-G as an immune regulator of GVHD. Furthermore, HLA-G could also serve as a good predictor of GVHD and represent a new therapeutic target for GVHD.

Comparison of Urologist Satisfaction for Different Types of Prostate MRI Reports: A Large Sample Investigation.

OBJECTIVE: To evaluate urologist satisfaction on structured prostate MRI reports, including report with tumor-node-metastasis (TNM) staging (report B) and with Prostate Imaging Reporting and Data System (PI-RADS) score with/without TNM staging (report C, report with PI-RADS score only [report C-a] and report with PI-RADS score and TNM staging [C-b]) compared with conventional free-text report (report A). MATERIALS AND METHODS: This was a prospective comparative study. Altogether, 3015 prostate MRI reports including reports A, B, C-a, and C-b were rated by 13 urologists using a 5-point Likert Scale. A questionnaire was used to assess urologist satisfaction based on the following parameters: correctness, practicality, and urologist subjectivity. Kruskal-Wallis H-test followed by Nemenyi test was used to compare urologists' satisfaction parameters for each report type. The rate of urologist-radiologist recalls for each report type was calculated. RESULTS: Reports B and C including its subtypes had higher ratings of satisfaction than report A for overall satisfaction degree, and parameters of correctness, practicality, and subjectivity (p < 0.05). There was a significant difference between report B and C (p < 0.05) in practicality score, but no statistical difference was found in overall satisfaction degree, and correctness and subjectivity scores (p > 0.05). Compared with report C-b (p > 0.05), report B and C-a (p < 0.05) showed a significant difference in overall satisfaction degree and parameters of practicality and subjectivity. In terms of correctness score, neither report C-a nor C-b had a significant difference with report B (p > 0.05). No statistical difference was found between report C-a and C-b in overall satisfaction degree and all three parameters (p > 0.05). The rate of urologist-radiologist recalls for reports A, B, C-a and C-b were 29.1%, 10.8%, 18.1% and 11.2%, respectively. CONCLUSION: Structured reports, either using TNM or PI-RADS are highly preferred over conventional free-text reports and lead to fewer report-related post-hoc inquiries from urologists.

A DNA aptamer recognizing MMP14 for in vivo and in vitro imaging identified by cell-SELEX.

A key challenge for the management of various types of cancer, including pancreatic cancer and hepatocellular carcinoma, is accurate diagnosis at an early stage. Matrix metalloproteinase 14 (MMP14) is overexpressed in numerous types of cancer and is associated with poor prognosis. Therefore, MMP14-specific imaging probes have potential use in the diagnosis of MMP14-positive cancer. Aptamers are short oligonucleotide sequences that can bind to molecular targets with a high specificity and affinity. Aptamers are typically obtained from an in vitro library; this process is usually termed systematic evolution of ligands by exponential enrichment (SELEX). In the present study, a DNA aptamer targeting MMP14 was obtained by cell-SELEX and termed M17, which specifically recognizes MMP14-positive cells. Aptamer M17 selectively binds to membrane proteins of MMP14-transfected 293T cells (Kd, 4.98±1.26 nM). Pancreatic cancer cell imaging suggested that aptamer M17 can bind to the cell membranes of two pancreatic cancer cell lines (MIA PaCa-2 and PANC-1). In vivo tumor imaging demonstrated that the targeting recognition of MIA PaCa-2 tumor cells in mice could be visualized using Cy5-labeled aptamer M17. Aptamer M17-conjugated polyethylene glycol-Fe3O4 can specifically bind to MIA PaCa-2 and PANC-1 cells, and reduce MRI T2-weighted imaging signal intensity. The DNA aptamer M17 has the advantages of simplicity of synthesis, small size, low immunogenicity, high penetrability and high affinity. Therefore, aptamer M17 is a potential molecular probe for the diagnosis and treatment of MMP14-positive cancer.

Changes in magnetic resonance T2-weighted imaging signal intensity correlate with concurrent chemoradiotherapy response in cervical cancer.

PURPOSE: This study is aimed to compare magnetic resonance imaging (MRI) parameters and clinical pathological factors (CPF) of residual tumor group with non-residual tumor group in cervical cancer (CC) patients during concurrent chemoradiotherapy (CCRT), and thus to establish a biomarker for individualized treatment strategy. MATERIAL AND METHODS: From May 2014 to November 2015, 164 CC patients were included in this retrospective study. T2-weighted MRI was performed at pre-treatment (week-0), the completion of external radiotherapy (RT) (week-4), and one month after the completion of CCRT, using 3.0T MR scanner with regular pelvic coil. Mean signal intensity and tumor size on T2WI images were measured and calculated for each tumor, and lumbar 4-5 intervertebral disc at week-0 and week-4. All patients subsequently underwent routine follow-up, including periodic clinical and imaging examinations when necessary. Receiver operator characteristics (ROC) analysis were conducted to determine cut-off values. RESULTS: The residual tumor group showed a higher Δ tumor-to-disc signal intensity ratio (ΔTDR) than non-residual tumor group (0.78 ± 0.30 vs. 0.48 ± 0.19, t = 3.42, p < 0.05). The biomarker of combined MRI parameter and CPF showed the highest diagnostic performance than single MRI parameter or CPF alone. CONCLUSIONS: MRI parameter ΔTDR may be an independent prognostic factor for predicting residual tumor occurrence in CC after CCRT treatment. The combination of MRI parameter and CPF can serve as a valuable biomarker to distinguish CC with higher possibility of residual tumor occurrence.

Quantitative parameters of intravoxel incoherent motion diffusion weighted imaging (IVIM-DWI): potential application in predicting pathological grades of pancreatic ductal adenocarcinoma.

BACKGROUND: The aim of this study was to compare intravoxel incoherent motion diffusion weighted imaging (IVIM-DWI) parameters such as standard apparent diffusion coefficient (ADCstandard), pure diffusion coefficient (Dslow), pseudodiffusion coefficient (Dfast) and perfusion fraction (ƒ) for differentiating pancreatic ductal adenocarcinoma (PDAC) with different pathological grades. METHODS: Institutional Review Board of our hospital approved this study protocol. Subjects comprised 38 PDACs confirmed by pathology. Pancreatic multiple b values DWI with 15 b values of 0, 10, 20, 40, 60, 80, 100, 150, 200, 400, 800, 1,000, 1200, 1,500, and 2,000 s/mm2 was performed using GE Discovery MR750 3.0T scanner. ADCstandard, Dslow, Dfast and ƒ values of all PDACs were calculated using mono- and bi-exponential models. Parameters of well/moderately differentiated and poorly differentiated PDAC were compared using Independent Sample t-test. P values <0.05 were considered significant. RESULTS: Mean Dslow value of well/moderately differentiated PDAC was significantly lower than that of poorly differentiated PDAC (0.540×10-3vs. 0.676×10-3 mm2/s, P<0.001). Mean ƒ value of well/moderately differentiated PDAC was significantly higher than that of poorly differentiated PDAC (60.3% vs. 38.4%, P<0.001). The area under curve value of ƒ in differentiating well/moderately differentiated PDAC from poorly differentiated PDAC was slightly higher than that of Dslow (0.894>0.865). When the Dslow value was less than or equal to 0.599×10-3 mm2/s, the sensitivity and specificity were 100% and 84.6% respectively. When ƒ value was greater than 49.6%, the sensitivity and specificity were 92.0% and 84.6% respectively. CONCLUSIONS: Dslow and ƒ derived from IVIM-DWI model can be used to distinguish well/moderately differentiated PDAC from poorly differentiated PDAC. And to serve this purpose, Dslow and ƒ have high diagnostic performance. IVIM-DWI is a promising and non-invasive tool for predicting pathological grade of PDAC.

DWI as a Quantitative Biomarker in Predicting Chemotherapeutic Efficacy at Multitime Points on Gastric Cancer Lymph Nodes Metastases.

The purpose of the hypothesis testing is to determine that apparent diffusion coefficient (ADC) as an early biomarker can predict the metastatic lymph nodes' (LNs) response to neoadjuvant chemotherapy in advanced gastric cancer (GC) in early stage. From March 2011 to June 2015, 106 patients with advanced GC were enrolled in the study. Patients underwent conventional magnetic resonance imaging and functional diffusion weighted imaging before and 3 days, 7 days, 30 days, and 60 days following the standard chemotherapy. After surgery, among 3034 detected LNs, the positive group was divided into complete response (CR) group, partial response (PR) group, and stable disease (SD) group in accordance to the Response Evaluation Criteria in Solid Tumors (RECIST) 1.1. Mean ADCs, short/long diameters of LNs before chemotherapy between the whole positive and the negative LNs were compared by t test. Changes of mean ADCs in 3 groups were analyzed by 1-way ANOVA. The mean ADC of the whole positive LNs was (1.145 ±â€Š0.014) × 10⁻³ mm²/s, which was significantly lower than that of the whole negative LNs ([1.491 ±â€Š0.010] × 10⁻³ mm²/s; P < 0.05). The means of both short/long diameters in the whole positive LNs were significantly longer than those in the whole negative LNs (P < 0.05). In CR, PR, and SD groups, the mean ADC of metastatic LNs on the 3rd day, 7th day, 13th day, and 16th day following the chemotherapy were all higher than that of LNs before chemotherapy, respectively (all P < 0.05). In addition, significant difference was found between mean ADCs in any 2 time points (all P < 0.05), except between mean ADCs in the 3rd day and in the 7th day of the chemotherapy. In conclusion, ADC can be used as an early biomarker to predict the metastatic LNs' response to neoadjuvant chemotherapy in advanced GC in early stage.

Lymph node metastasis in patients with gastric cancer: a multi-modality, morphologic and functional imaging study.

The aim of the study was to investigate the value of computed tomography (CT), magnetic resonance imaging (MRI) and diffusion weighted imaging (DWI) for diagnosing lymph nodes metastasis before treatment in gastric cancer. Eighty-two patients with proven gastric cancer underwent CT, morphological MRI (T2WI) and DWI examinations. Two radiologists independently assessed these images for the presence of lymph nodes involvement. Pathologic findings were considered as "gold standard". Independent samples t-test was used for the comparisons of short diameters and ADC values between the positive lymph nodes and the negative lymph nodes. Diagnostic accuracy of these three imaging modalities was evaluated by area under the receiver operating characteristics (ROC) curve (AUC). The ADC value of the positive lymph nodes was (1.15 ± 0.01) × 10-3 mm2/s, which was significantly lower than that of the negative lymph nodes (1.48 ± 0.01) × 10-3 mm2/s (t = 18.70, P < 0.0001). The short diameter of the positive lymph nodes (1.54 ± 0.38 cm) was significantly greater than that of the negative lymph nodes (0.95 ± 0.12 mm) (t = 19.03, P < 0.001). The AUC for all imaging modalities combined (0.893) was significantly larger than that for each imaging modality alone (P < 0.05), and the AUC of DWI (0.797) was significantly larger than (P < 0.05) that of morphological MRI (0.733). There was no statistically significant difference between the AUCs of CT and morphological MRI (P = 0.462). In conclusion, CT, MRI and DWI combined present significantly higher accuracy than each imaging modality alone in the detection of lymph nodes involvement.