Yan Chen

Objectives: To evaluate a multiparametric MRI protocol encompassing intravoxel incoherent motion diffusion-weighted imaging, diffusion kurtosis imaging, and dynamic contrast-enhanced MRI for discriminating aggressive (AEC) from non-aggressive type endometrial carcinoma (NAEC) according to the FIGO 2023 staging system. Methods: This study involved retrospective analyses of a prospective dataset. A total of 112 consecutive patients (77 NAEC and 35 AEC) underwent multiparametric MRI. Intratumoral and peritumoral quantitative MRI parameters were calculated. A multivariate logistic regression model comprising clinical data, conventional MRI features, and quantitative MRI metrics was constructed. Model performance was evaluated using receiver operating characteristic analysis, calibration curves, and bootstrap resampling (n = 1000). Results: Aggressive type endometrial carcinoma demonstrated significantly lower perfusion fraction (f) and mean diffusivity (MD), but higher pseudo diffusion coefficient (D*) and peritumoral D* (D*_peri) compared to NAEC (all P < .05). Multivariate analysis identified f, peritumoral mean kurtosis (MK_peri), and peritumoral maximum slope of increase (MaxSlope_peri) as independent predictors of AEC (AUC = 0.791, 95% CI, 0.692-0.891). Integration of menopause status, tumour location and extension beyond the corpus, and quantitative MRI parameters yielded a combined model with a stratified bootstrap AUC of 0.830 (95% CI, 0.800-0.846), particularly for FIGO 2023 stage I-II patients (AUC = 0.832, 95% CI, 0.742-0.922). Significant differences in D*, D*_peri, f, and MD were observed among NAEC with and without squamous differentiation, and AEC groups. Conclusions: Multiparametric MRI, incorporating advanced quantitative sequences and conventional MRI features, could effectively predict AEC before surgery. Advances in knowledge: The study bridges advanced imaging with the updated FIGO 2023 staging system, potentially offering a non-invasive assessment tool for endometrial carcinoma management.

Objective: To investigate the utility of multi-sequence magnetic resonance imaging (MRI) and whole-tumor apparent diffusion coefficient (ADC) histogram metrics in preoperatively differentiating p53 abnormal (p53abn) from non-p53abn endometrial carcinoma (EC).Methods: This retrospective study included 146 EC patients (29 p53abn cases and 117 non-p53abn cases) who underwent preoperative MRI scans. MRI features were analyzed. Whole-tumor ADC histogram analysis was conducted by delineating regions of interest (ROIs) on diffusion-weighted imaging (DWI) scans. Receiver operating characteristic (ROC) curve analysis with the area under the curve (AUC) was used for diagnostic performance evaluation.Results: Extrauterine extension (p=0.004) and lymphadenopathy (p=0.005) were more frequently observed in p53abn EC compared to non-p53abn EC. p53abn EC exhibited significantly lower value of minADC (p=0.001), meanADC (p=0.005), P10 (p=0.009), P50 (p=0.007), and P90 (p=0.013) ADC and higher value of kurtosis (p=0.008), compared to non-p53abn EC. MinADC demonstrated the highest discrimination ability in differentiating p53abn from non-p53abn EC [AUC 0.70(0.60;0.80)].Conclusion: Preoperative multi-sequence MRI findings and whole-tumor ADC histogram metrics are conducive to differentiating p53abn from non-p53abn EC.

Background: This study aims to construct predicting models using radiomic and clinical features in predicting first-line vascular endothelial growth factor receptor-tyrosine kinase inhibitor (VEGFR-TKI) early resistance in metastatic clear cell renal cell carcinoma (mccRCC) patients. We also aim to explore the correlation of predicting models with short and long-term survival of mccRCC patients. Materials and methods: In this retrospective study, 110 mccRCC patients from 2009 to 2019 were included and assigned into training and test sets. Radiomic features were extracted from tumor 3D-ROI of baseline enhanced CT images. Radiomic features were selected by Lasso method to construct a radiomic score. A combined nomogram was established using the combination of radiomic score and clinical factors. The discriminative abilities of the radiomic, clinical and combined nomogram were quantified using ROC curve. Cox regression analysis was used to test the correlation of nomogram score with progression-free survival (PFS) and overall survival (OS). PFS and OS were compared between different risk groups by log-rank test. Results: The radiomic, clinical and combined nomogram demonstrated AUCs of 0.81, 0.75, and 0.83 in training set; 0.79, 0.77, and 0.88 in test set. Nomogram score ≥ 1.18 was an independent prognostic factor of PFS (HR 0.22 (0.10, 0.47), p < 0.001) and OS (HR 0.38 (0.20, 0.71), p = 0.002), in training set. PFS in low-risk group were significantly longer than high-risk group in training (p < 0.001) and test (p < 0.001) set, respectively. OS in low-risk group were significantly longer than high-risk group in training (p = 0.003) and test (p = 0.009) set, respectively. Conclusion: A nomogram combining baseline radiomic signature and clinical factors helped detecting first-line VEGFR-TKI early resistance and predicting short and long-term prognosis in mccRCC patients.

Objectives: To investigate whether reduced field-of-view (rFOV) diffusion-weighted imaging (DWI) with deep learning reconstruction (DLR) can improve the accuracy of evaluating muscle invasion using VI-RADS. Methods: Eighty-six bladder cancer participants who were evaluated by conventional full field-of-view (fFOV) DWI, standard rFOV (rFOV_STA) DWI, and fast rFOV with DLR (rFOV_DLR) DWI were included in this prospective study. Tumors were categorized according to the vesical imaging reporting and data system (VI-RADS). Qualitative image quality scoring, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and ADC value were evaluated. Friedman test with post hoc test revealed the difference across the three DWIs. Receiver operating characteristic analysis was performed to calculate the areas under the curve (AUCs). Results: The AUC of the rFOV_STA DWI and rFOV_DLR DWI were higher than that of fFOV DWI. rFOV_DLR DWI reduced the acquisition time from 5:02 min to 3:25 min, and showed higher scores in overall image quality with higher CNR and SNR, compared to rFOV_STA DWI (p < 0.05). The mean ADC of all cases of rFOV_STA DWI and rFOV_DLR DWI was significantly lower than that of fFOV DWI (all p < 0.05). There was no difference in mean ADC value and the AUC for evaluating muscle invasion between rFOV_STA DWI and rFOV_DLR DWI (p > 0.05). Conclusions: rFOV DWI with DLR can improve the diagnostic accuracy of fFOV DWI for evaluating muscle invasion. Applying DLR to rFOV DWI reduced the acquisition time and improved overall image quality while maintaining ADC value and diagnostic accuracy. Critical relevance statement: The diagnostic performance and image quality of full field-of-view DWI, reduced field-of-view (rFOV) DWI with and without DLR were compared. DLR would benefit the wide clinical application of rFOV DWI by reducing the acquisition time and improving the image quality. Key Points: Deep learning reconstruction (DLR) can reduce scan time and improve image quality. Reduced field-of-view (rFOV) diffusion-weighted imaging (DWI) with DLR showed better diagnostic performances than full field-of-view DWI. There was no difference of diagnostic accuracy between rFOV DWI with DLR and standard rFOV DWI. Graphical

Purpose To investigate the influence of deep learning reconstruction (DLR) on bladder MRI, specifically examination time, image quality, and diagnostic performance of vesical imaging reporting and data system (VI-RADS) within a prospective clinical cohort. Methods Seventy participants with bladder cancer who underwent MRI between August 2022 and February 2023 with a protocol containing standard T2-weighted imaging (T2WI_S), standard diffusion-weighted imaging (DWI_S), fast T2WI with DLR (T2WI_DL), and fast DWI with DLR (DWI_DL) were enrolled in this prospective study. Imaging quality was evaluated by measuring signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and qualitative image quality scoring. Additionally, the apparent diffusion coefficient (ADC) of bladder lesions derived from DWI_S and DWI_DL was measured and VI-RADS scoring was performed. Paired t-test or paired Wilcoxon signed-rank test were performed to compare image quality score, SNR, CNR, and ADC between standard sequences and fast sequences with DLR. The diagnostic performance for VI-RADS was assessed using the area under the receiver operating characteristic curve (AUC). Results Compared to T2WI_S and DWI_S, T2WI_DL and DWI_DL reduced the acquisition time from 5:57 min to 3:13 min and showed significantly higher SNR, CNR, qualitative image quality score of overall image quality, image sharpness, and lesion conspicuity. There were no significant differences in ADC and AUC of VI-RADS between standard sequences and fast sequences with DLR. Conclusions The application of DLR to T2WI and DWI reduced examination time and significantly improved image quality, maintaining ADC and the diagnostic performance of VI-RADS for evaluating muscle invasion in bladder cancer. Graphical abstract

Background: The Vesical Imaging-Reporting and Data System (VI-RADS) has demonstrated effectiveness in predicting muscle invasion in bladder cancer before treatment. The urgent need currently is to evaluate the muscle invasion status after neoadjuvant chemotherapy (NAC) for bladder cancer. This study aims to ascertain the accuracy of VI-RADS in detecting muscle invasion post-NAC treatment and assess its diagnostic performance across readers with varying experience levels. Methods: In this retrospective study, patients with muscle-invasive bladder cancer who underwent magnetic resonance imaging (MRI) after NAC from September 2015 to September 2018 were included. VI-RADS scores were independently assessed by five radiologists, consisting of three experienced in bladder MRI and two inexperienced radiologists. Comparison of VI-RADS scores was made with postoperative histopathological diagnosis. Receiver operating characteristic curve analysis (ROC) was used for evaluating diagnostic performance, calculating sensitivity, specificity, and area under ROC (AUC)). Interobserver agreement was assessed using the weighted kappa statistic. Results: The final analysis included 46 patients (mean age: 61 years ± 9 [standard deviation]; age range: 39–70 years; 42 men). The pooled AUC for predicting muscle invasion was 0.945 (95% confidence interval (CI): 0.893–0.977) for experienced readers, and 0.910 (95% CI: 0.831–0.959) for inexperienced readers, and 0.932 (95% CI: 0.892–0.961) for all readers. At an optimal cut-off value ≥ 4, pooled sensitivity and specificity were 74.1% (range: 66.0–80.9%) and 94.1% (range: 88.6–97.7%) for experienced readers, and 63.9% (range: 59.6–68.1%) and 86.4% (range: 84.1–88.6%) for inexperienced readers. Interobserver agreement ranged from substantial to excellent between all readers (k = 0.79–0.92). Conclusions: VI-RADS accurately assesses muscle invasion in bladder cancer patients after NAC and exhibits good diagnostic performance across readers with different experience levels.

This study aimed to explore the association between the quantitative characteristics of dual-energy spectral CT and cytoreduction surgery outcome in patients with advanced epithelial ovarian carcinoma (EOC). In this prospective observational study, patients with advanced EOC (federation of gynecology and obstetrics stage III–IV) treated in the Department of Gynecological Oncology at our Hospital between June 2021 and March 2022 were enrolled. All participants underwent dual-energy spectral computed tomography (DECT) scanning 2 weeks before cytoreductive surgery. The quantitative data included peritoneal cancer index (PCI) determined by DECT, CT value at 70 keV, normalized iodine concentration, normalized water concentration, effective atomic number (effective-Z), and slopes of the spectral attenuation curves (slope λ Hounsfield unit). Fifty-five participants were included. The patients were 57.2 ± 9.8 years of age, and 72.7% were menopausal. The maximal diameter of tumors was 8.6 (range, 2.9–19.7) cm, and 76.4% were high-grade serous carcinomas. Optimal cytoreduction was achieved in 43 patients (78.2%). Compared with the optimal cytoreductive group, the suboptimal cytoreductive group showed a higher PCI (median, 21 vs 6, P < .001), higher 70 keV CT value (69.5 ± 16.6 vs 57.1 ± 13.0, P = .008), and higher slope λ Hounsfield unit (1.89 ± 0.66 vs 1.39 ± 0.60, P = .015). The multivariable analysis showed that the PCI (OR = 1.74, 95%CI: 1.24–2.44, P = .001) and 70 keV CT value (OR = 1.07, 95%CI: 1.01–1.13, P = .023) were independently associated with a suboptimal cytoreductive surgery. The area under the receiver operating characteristics curve of PCI and 70 keV CT value was 0.903 (95%CI: 0.805–1.000, P = .000) and 0.740 (95%CI: 0.581–0.899, P = .012), respectively. High PCI and 70 keV CT value are independently associated with suboptimal cytoreductive surgery in patients with advanced EOC. The PCI determined by DECT might be a better predictor for suboptimal cytoreduction.

Purpose To explore the diagnostic efficacy of MR-based texture analysis in differentiation of small (≤ 4 cm) and very small (≤ 2 cm) renal cell carcinoma subtypes. Methods One hundred and eight patients with pT1a (≤ 4 cm) renal cell carcinoma and pretreatment MRI were enrolled in this retrospective study. Histogram and gray-level co-occurrence matrix (GLCM) parameters were extracted from whole-tumor images. Among subtypes, patient age, tumor size, histological grading and texture parameters were compared. Diagnostic model using combination of texture parameters was constructed using logistic regression and validated using fivefold cross-validation. AUC with 95% CI, accuracy, sensitivity and specificity for subtype differentiation are reported. Further we explored the distinguishing ability of texture parameters and diagnostic model in very small (≤ 2 cm) RCC subgroups. Results Significant texture parameters among RCC subtypes were identified. For small (≤ 4 cm) renal cell carcinoma subtyping, combining models based on texture parameters achieved good AUCs for differentiating ccRCC vs. non-ccRCC, chRCC vs. non-chRCC and ccRCC vs. chRCC (0.79, 0.74 and 0.81). Further, in subgroups of very small (≤ 2 cm) RCCs, diagnostic models had better differentiating performances, achieving AUCs of 0.88, 0.99, 0.96 in differentiating ccRCC vs. non-ccRCC, chRCC vs. non-chRCC and ccRCC vs. chRCC. Conclusion MR texture analysis may help to differentiate small (≤ 4 cm) and very small (≤ 2 cm) RCC subtypes. This non-invasive method can potentially provide additional information for localized RCC treatment and surveillance strategy. Graphical abstract

Purpose: It is challenging for radiologists and gynecologists to identify the type of ovarian lesions by reading magnetic resonance (MR) images. Recently developed convolutional neural networks (CNNs) have made great progress in computer vision, but their architectures still need modification if they are used in processing medical images. This study aims to improve the feature extraction capability of CNNs, thus promoting the diagnostic performance in discriminating between benign and malignant ovarian lesions.Methods: We introduce a feature fusion architecture and insert the attention models in the neural network. The features extracted from different middle layers are integrated with reoptimized spatial and channel weights. We add a loss function to constrain the additional probability vector generated from the integrated features, thus guiding the middle layers to emphasize useful information. We analyzed 159 lesions imaged by dynamic contrast‐enhanced MR imaging (DCE‐MRI), including 73 benign lesions and 86 malignant lesions. Senior radiologists selected and labeled the tumor regions based on the pathology reports. Then, the tumor regions were cropped into 7494 nonoverlapping image patches for training and testing. The type of a single tumor was determined by the average probability scores of the image patches belonging to it.Results: We implemented fivefold cross‐validation to characterize our proposed method, and the distribution of performance matrics was reported. For all the test image patches, the average accuracy of our method is 70.5% with an average area under the curve (AUC) of 0.785, while the baseline is 69.4% and 0.773, and for the diagnosis of single tumors, our model achieved an average accuracy of 82.4% and average AUC of 0.916, which were better than the baseline (81.8% and 0.899). Moreover, we evaluated the performance of our proposed method utilizing different CNN backbones and different attention mechanisms.Conclusions: The texture features extracted from different middle layers are crucial for ovarian lesion diagnosis. Our proposed method can enhance the feature extraction capabilities of different layers of the network, thereby improving diagnostic performance.