OBJECTIVES:To develop and test a convolutional neural network model for automated segmentation of complicated cystic renal masses (cCRMs) on MRI. METHODS:This multicenter retrospective study analysed 210 cCRMs between October 2019 and May 2021, divided into training/internal validation (n = 150, Institution 1) and test sets (n = 60, Institutions 2-4). Comparative 3D V-Net and U-Net models were developed across 7 MRI sequences (T2-weighted, diffusion-weighted, apparent diffusion coefficient maps, unenhanced T1-weighted, and enhanced corticomedullary, nephrographic, and excretory phases images). A total of 14 models were developed, and 7 pairwise comparisons were performed between the 3D V-Net and U-Net models. Segmentation performance was evaluated using Dice similarity coefficient (DSC) and Hausdorff distance (HD), with subgroup analysis of small cCRMs (≤40 mm). RESULTS:In the test set, the excretory-phase V-Net (EPV-Net model) showed the highest DSC, and perform better than the corresponding U-Net (EPU-Net model) across all cCRMs (DSC: 0.74 ± 0.05 vs 0.70 ± 0.06, P < .001; HD: 27.41 ± 7.44 mm vs 39.18 ± 11.07 mm, P < .001) and the 35 small cCRMs subgroup (DSC: 0.74 ± 0.05 vs 0.70 ± 0.06, P < .001; HD: 27.48 mm ± 6.32 vs 38.72 ± 10.69 mm, P < .001). CONCLUSIONS:The 3D EPV-Net model demonstrated good segmentation accuracy, even for small lesions, supporting its clinical utility for cCRMs evaluation. ADVANCES IN KNOWLEDGE:This automated approach may streamline workflow compared to manual segmentation in cCRMs assessment.

PURPOSE:Robot-assisted nephrectomy and inferior vena cava (IVC) thrombectomy are technically demanding and high-risk procedures. The role of multidisciplinary team (MDT) management in robotic surgeries has not been reported previously. This study aimed to evaluate the safety and feasibility of MDT management in robot-assisted nephrectomy and IVC thrombectomy. PATIENTS AND METHODS:We retrospectively analyzed 209 patients who underwent robot-assisted nephrectomy and IVC thrombectomy for renal tumor with venous tumor thrombus (Mayo levels I-IV) in our center between June 2013 and December 2023. Since July 2018, a proactive, comprehensive, and full-process MDT management framework has been implemented perioperatively. Patients were divided into MDT management (n = 142) and non-MDT management (n = 67) groups, with propensity score matching (1:1) resulting in 67 patients in each group. Multivariable regression, survival analyses, and interrupted time series analysis were conducted to assess perioperative outcomes and survival. RESULTS:All procedures were completed without conversion. MDT management significantly improved perioperative outcomes compared with the non-MDT group. Specifically, MDT reduced estimated blood loss (p = 0.045), intraoperative blood transfusion (p = 0.009), blood transfusion volume (p = 0.005), postoperative intensive care unit stay (p = 0.002), and postoperative hospital stay (p < 0.001). Multivariable regression confirmed that these improvements were independent of other factors. MDT management was associated with a significant improvement in overall survival (hazard ratio [HR] = 0.27, p = 0.023) and a potential benefit in progression-free survival (HR = 0.40, p = 0.065), as indicated by survival analysis. CONCLUSIONS:MDT management in robot-assisted nephrectomy and IVC thrombectomy improves perioperative safety and enhances survival outcomes. This study highlights the critical role of the standardized MDT framework we proposed in optimizing high-risk robotic procedures.

Abstract Objectives To validate blood oxygen level-dependent MRI (BOLD-MRI) for non-invasive discrimination of diabetic nephropathy (DN) vs non-diabetic renal disease (NDRD) and prediction of end-stage renal disease (ESRD) in diabetic kidney disease (DKD). Materials and methods A prospective cohort of 133 biopsy-proven DKD patients underwent BOLD-MRI. The semi-automated 12-layer concentric-objects method was used to analyze BOLD-MRI variables. Prognostic markers for ESRD were identified using univariate and multivariate Cox regression. Feature importance was used to select key diagnostic variables and establish logistic regression and machine-learning differential diagnosis models. Results Among 133 patients (44 DN, 55 NDRD, 34 combined), 20 (15.5%) progressed to ESRD over a mean of 21.8 months. Higher renal medullary R2* (MR2*) (> 24 1/s) reduced ESRD risk by 52% (HR, 0.48) in DKD. Prognostic models integrating pathological grouping, hemoglobin levels, and cysC levels achieved a c-index of 0.90. For the DN and combined groups, MR2*, glomerular grading, interstitial lesions, interstitial fibrosis, and tubular atrophy were predictive of ESRD, with a c-index of 0.91. For differential diagnosis, the random forest (RF) model achieved an AUC of 0.901, with diabetic retinopathy, diabetes duration, albumin, blood urea nitrogen, MR2*, hypertension, and glycosylated hemoglobin as the most contributing factors. For the combined group classified as DN, the AUC of the RF model was 0.791; when classified as NDRD, the AUC was 0.856. Conclusion MR2* shows potential value as a non-invasive diagnostic and prognostic tool in the assessment of DKD. However, BOLD-MRI remains a promising yet exploratory technique that requires external validation and interventional studies before clinical implementation. Critical relevance statement Blood oxygen level-dependent-MRI-derived renal medullary R2* robustly predicts ESRD risk and distinguishes DN without biopsy, offering an immediately translatable, non-invasive biomarker for the precision management of DKD in routine nephrology practice. Trial registration ClinicalTrials.gov, NCT03865914. Key Points Blood oxygen level-dependent-MRI medullary R2*(MR2*) > 24 s− 1 halves DKD ESRD risk (HR 0.48). MR2* integrated with clinical variables drives c-index to 0.90 for ESRD prognosis. RF leveraging MR2* and clinical traits attains an AUC of 0.901 for diagnosing DN. Graphical Abstract

BACKGROUND:Lymph node metastasis is a well-established prognostic factor in renal cell carcinoma (RCC); however, the prognosis varies among patients with nodal metastases. Subclassification by T category may improve the prognostic stratification. METHODS:This study aimed to refine the prognostic stratification of advanced RCC by analyzing the overall survival and cancer-specific survival of patients classified as pT1-3N1M0. We retrospectively analyzed data from 364 patients with American Joint Committee on Cancer (AJCC) stage III and IV RCC who underwent nephrectomy with lymphadenectomy at multiple centers between 2010 and 2023. Five-year survival rates were estimated using Kaplan-Meier analysis, and the concordance index (C-index) was used to assess the stratification ability of the different staging systems. Multivariable analyses were conducted to adjust for potential confounders. Additionally, 194 and 1513 patients from The Cancer Genome Atlas and Surveillance, Epidemiology, and End Results databases were used to validate the prognostic values. RESULTS:Patients with pT1-2N1M0 had significantly better 5-year survival rates than those with pT3N1M0 (overall survival: 62.4% vs. 31.5%, P = 0.003; cancer-specific survival: 62.4% vs. 38.5%, P = 0.008). Reclassifying pT3N1M0 from AJCC stage III to IV improved prognostic discrimination, as reflected by higher C-index values compared to the AJCC staging (8th edition) (ΔC-index: 0.03-0.10, all P values <0.05). The modified staging system demonstrated comparable performance to the previous system while avoiding the over-staging of patients with pT1-2N1M0 (ΔC-index: 0-0.04, most P values >0.05), as confirmed across local and external cohorts. These findings were supported by the Multivariable analyses. CONCLUSION:Reclassifying pT3N1M0 as stage IV improves prognostic stratification in RCC.

Purpose To develop and evaluate a preoperative MRI-based model for predicting inferior vena cava (IVC) wall invasion in renal cell carcinoma (RCC) with IVC tumor thrombus (IVCTT) and to compare its performance with that of individual MRI features and radiologists' subjective assessments. Materials and Methods This single-center study with retrospective and prospective components included individuals who underwent or were scheduled to undergo surgery for RCC with IVCTT (retrospective training set, n = 173, January 2005-December 2023; prospective temporal validation set, n = 44, January 2024-September 2025). Histopathology served as the reference standard. Quantitative (tumor, vessel, and thrombus measurements) and qualitative (signal and morphologic characteristics) MRI features were assessed. Two fellowship-trained abdominal radiologists independently provided subjective assessments of IVC wall invasion, and interobserver agreement was assessed. Variables significant at univariable analysis were entered into multivariable logistic regression to identify predictors of IVC wall invasion. Diagnostic performance was compared using receiver operating characteristic (ROC) curve analysis and DeLong tests. Results A total of 217 individuals were included (mean age, 57 years ± 12 [SD], 166 male). Four independent predictors of IVC wall invasion were identified: bland thrombus (odds ratio [OR] = 3.32 [95% CI: 1.38, 8.03]), lumbar vein diameter (>5.25 mm) (OR = 2.64 [95% CI: 1.23, 5.69]), ipsilateral renal vein ostium diameter (>19.20 mm) (OR = 3.64 [95% CI: 1.73, 7.63]), and thrombus craniocaudal length (>46.95 mm) (OR = 3.08 [95% CI: 1.43, 6.63]). The multivariable model incorporating these predictors achieved area under the ROC curve (AUC) values of 0.81 (95% CI: 0.75, 0.88) and 0.84 (95% CI: 0.73, 0.96) in the training and validation sets, respectively, significantly outperforming the best individual MRI predictor (AUC = 0.71) and radiologists' subjective assessments (AUC = 0.66) (all P < .05). Conclusion The multiparametric MRI-based model demonstrated good discriminatory performance for predicting IVC wall invasion and outperformed individual MRI features and subjective radiologist assessment. Keywords: MR Imaging, Urinary, Kidney, Renal Cell Carcinoma, Magnetic Resonance Imaging, Inferior Vena Cava Tumor Thrombus, Venous Wall Invasion Supplemental material is available for this article. © RSNA, 2026.

Neoadjuvant immunotherapy-based combination holds promises in reducing surgical risk and improving survival for renal cell carcinoma (RCC) with venous tumor thrombus (VTT). However, its role in RCC-VTT has been less explored. To evaluate the efficacy and safety of neoadjuvant toripalimab plus axitinib in nonmetastatic RCC-VTT, we conducted a combined analysis of two Phase II trials with similar design. Thirty-four patients with nonmetastatic clear cell RCC (ccRCC) and Mayo Level 0-IV VTT were enrolled. Toripalimab plus axitinib was administered for up to 12 weeks before surgery. The primary endpoint was objective response rate (ORR). In this study, the ORR and disease control rates were 41% (14 out of 34) and 97% (33 out of 34), respectively. 47% (16 out of 34) patients experienced a reduction in VTT levels. Grade 3 treatment-related adverse events (TRAEs) occurred in 24% (eight out of 34) patients, and no Grade 4 or 5 TRAEs were observed. Thirty patients were eligible for surgery, and the surgical strategy was simplified in 53% (16 out of 30) patients. One-year disease-free survival and overall survival were 76.7% (95% CI, 59.1-88.2%) and 91.2% (95% CI, 77.0-97.0%), respectively. Multiomics analysis revealed the nonresponder group exhibited significant tumor heterogeneity and a stroma-characterized tumor microenvironment. In conclusion, neoadjuvant toripalimab plus axitinib was clinically active and safe in patients with nonmetastatic ccRCC-VTT.

BACKGROUND:Although the clear cell likelihood score (ccLS) v2.0 demonstrates high specificity for clear cell renal cell carcinoma (ccRCC), its performance to characterize general malignancy in small renal masses (SRMs) remains limited. PURPOSE:To develop and validate a modified clear cell likelihood score (m-ccLS) incorporating the pseudocapsule to improve malignancy detection in SRMs while preserving specificity for diagnosing ccRCC. STUDY TYPE:This study was retrospective in type. SUBJECTS:352 patients with pathologically proven SRMs were included: development (n = 235), internal validation (n = 60), and external validation (n = 57). FIELD STRENGTH/SEQUENCE:Imaging was performed at 3.0 and 1.5 T using fast spin-echo T2-weighted imaging, single-shot echo planar diffusion-weighted imaging, 3D spoiled gradient echo (GRE) T1-weighted dynamic contrast-enhanced imaging, and in- and opposed-phase using T1-weighted GRE. ASSESSMENT:14 radiologists blinded to histopathology independently evaluated each SRM using ccLS v2.0 and m-ccLS scores in separate reading sessions; four, five, and five readers interpreted the development, internal, and external cohorts, respectively. STATISTICAL TESTS:Random-effects logistic regression, receiver operating characteristic curve, DeLong test, net reclassification improvement (NRI), integrated discrimination improvement (IDI), and Fleiss Kappa test were used. The statistical significance level was p < 0.05. RESULTS:For malignancy detection, m-ccLS showed a significantly higher area under the curve (AUC) than ccLS v2.0 across the development (0.850 vs. 0.772), internal validation (0.856 vs. 0.779), and external validation (0.803 vs. 0.720) cohorts with improved classification (NRI = 0.270, 0.045, and 0.028) and discrimination (IDI = 0.132, 0.206, and 0.120). For diagnosing ccRCC, m-ccLS and ccLS v2.0 showed similar results (0.908 vs. 0.894, p = 0.250; 0.912 vs. 0.898, p = 0.134; 0.865 vs. 0.838, p = 0.065) in development, internal, and external validation cohorts, respectively. m-ccLS category 3 contained fewer ccRCCs (33.3% vs. 72.5%; 15.8% vs. 47.2%; 7.6% vs. 26.9%) and malignancies (79.2% vs. 88.7%; 71.6% vs. 73.0%; 55.4% vs. 63.9%) than ccLS v2.0 category 3. DATA CONCLUSION:m-ccLS improves malignancy detection in SRMs compared with ccLS v2.0 without impairing diagnostic performance for ccRCC. EVIDENCE LEVEL:4. TECHNICAL EFFICACY:Stage 2.

To develop a grading system integrating MRI and clinicopathological features for predicting positive surgical margin (PSM) following robotic-assisted laparoscopic prostatectomy (RALP) among patients with prostate cancer. Patients undergoing RALP were retrospectively included with consecutive MRI examinations collected from two centers (center 1 and center 2). The train cohort included patients at center 1 between January 2020 and December 2021, and the validation cohort comprised those between January 2022 and December 2022. Patients from center 2 were assigned to the test cohort. MRI and clinicopathological features associated with PSM were assessed. A logistic regression model was used to develop the grading system. The prediction and calibration performance were evaluated by area under the receiver operating characteristic curves (AUCs) and Hosmer-Lemeshow goodness-of-fit test. AUC values were compared by Delong test. A total of 396 patients and 29.2

To evaluate intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) for stratifying early response to neoadjuvant immune checkpoint inhibitor and tyrosine kinase inhibitor (ICI-TKI) therapy in renal cell carcinoma (RCC) and explore correlations between IVIM‑DWI parameters and tumor immune cell infiltration. This retrospective exploratory analysis utilized prospectively collected data from a single‑center study enrolling patients with advanced, metastatic or unresectable RCC scheduled for neoadjuvant ICI‑TKI therapy. Two radiologists independently segmented primary tumors on baseline IVIM‑DWI to obtain the true diffusion coefficient (D), pseudo‑diffusion coefficient (D*), perfusion fraction (f), and standard apparent diffusion coefficient (ADCstandard). Response was assessed per iRECIST. Baseline parameters were compared between responders and nonresponders; subgroup analysis was conducted for clear cell RCC (ccRCC). Diagnostic performance was evaluated using the area under the curve (AUC). Immune infiltration was analyzed from RNA‑sequencing data, and correlations were assessed with the Spearman’s test. Among 58 patients (44 ccRCC), 33 (56.9

RATIONALE AND OBJECTIVES: This study aimed to develop and validate a magnetic resonance imaging (MRI)-based habitat radiomics model to predict recurrence-free survival (RFS) in patients with nonmetastatic clear cell renal cell carcinoma (ccRCC) after surgical resection. MATERIALS AND METHODS:A retrospective cohort of 630 patients with nonmetastatic ccRCC who underwent surgical resection at the First Medical Center of Chinese PLA General Hospital (2011-2019) was included. Preoperative T2-weighted imaging (T2WI) and contrast-enhanced corticomedullary phase (CP) MRI were used to cluster tumor voxels into homogeneous habitats via K-means algorithm based on signal intensity. Radiomic features were extracted from habitats; after feature selection, these features were integrated with clinicopathological indicators to build a Cox proportional hazards regression model. Model performance was assessed via receiver operating characteristic (ROC) curves, concordance index (C-index), calibration curves, and decision curve analysis (DCA). RESULTS:Three distinct tumor habitat regions were identified through clustering, from which 13 recurrence-related radiomic features were selected to construct a Habitat Signature (HS). Multivariate Cox regression analysis demonstrated that age (HR = 1.039, 95% CI: 1.016-1.063, P<0.001), sex (male vs female, HR = 2.608, 95% CI: 1.291-5.270, P=0.008), and pathological T stage (T3 vs T1, HR = 4.284, 95% CI: 1.997-9.193, P < 0.001) served as independent predictors of postoperative recurrence. Constructed by combining clinicopathological predictors with the HS score, the clinical-habitat combined model yielded AUC values for 3-year and 5-year postoperative recurrence prediction of 0.80/0.81 in the training set and 0.85/0.81 in the test set, along with C-indices of 0.80 and 0.81, respectively. CONCLUSION:The predictive model constructed by combining MRI-based HS score and clinical-pathological features has predictive value for recurrence of nonmetastatic ccRCC.

To evaluate whether MRI characteristics of septa and walls—specifically enhancement patterns during the corticomedullary (CP), nephrographic (NP), and excretory (EP) phases—can improve malignancy risk stratification in Bosniak III cystic renal masses (CRMs). This single-center, retrospective study analyzed 120 patients with Bosniak III cystic renal masses who underwent renal MRI between January 2009 and December 2021. The cohort included two subcategories: III-WS (enhancing thick wall/septa ≥ 4 mm; n = 22) and III-OP (enhancing irregular wall/septa or convex protrusion ≤ 3 mm; n = 98). All lesions were confirmed either by histopathology (115 CRMs) or ≥ 5 year stability (5 CRMs). Four radiologists (2 senior, 2 junior), blinded to clinical and pathological data, independently assessed septal and wall enhancement (obvious vs. non-obvious) across CP, NP) and EP, a consensus was reached through discussion. Interobserver agreement was evaluated using Conger’s kappa, while diagnostic performance (AUC, sensitivity, specificity, accuracy) of obvious enhancement was assessed via ROC analysis, with AUC comparisons performed using DeLong’s test. The study included 120 patients (mean age: 48 ± 11 years; 94 male), with 95 (79.2

Abstract Objectives The objective of this study is to assess contralateral renal parenchymal volume (CRPV) changes in patients with renal cell carcinoma (RCC) after PN and identify the clinical factors associated with CRPV increase. Methods This retrospective study included 143 RCC patients (aged 24–76 years) who underwent PN between 2017 and 2018. Preoperative and postoperative CRPV were assessed using 3D Slicer software. Patients were categorized into CRPV‐increase and CRPV‐decrease groups based on volumetric changes. Univariate and multivariable logistic regression analyses were conducted to determine independent factors influencing CRPV increase. Results With a median follow‐up of 904 days (IQR: 605 – 1707), the overall median CRPV increase was 4.1% (IQR: −2.8% to 12.6%) relative to the preoperative volume, and 68.5% (98/143) of patients exhibited CRPV enlargement. Multivariable analysis identified postoperative acute kidney injury (AKI) as a significant independent predictor associated with CRPV increase (OR = 2.990, 95% CI: 1.050–8.513, p = 0.040), whereas preexisting hypertension was inversely associated with CRPV growth (OR = 0.446, 95% CI: 0.206–0.966, p = 0.041). Conclusions Contralateral renal compensation following PN is relatively modest compared to radical nephrectomy. Postoperative AKI is associated with compensatory renal hypertrophy, whereas hypertension is negatively associated with this process. These hypothesis‐generating findings underscore the importance of perioperative renal function management and blood pressure control. Prospective studies are needed to validate these associations.

Recent advancements in radiological imaging have raised the possibility of diagnosing prostate cancer (PCa) without biopsy; however, the safety, feasibility, and diagnostic accuracy of this approach require comprehensive evaluation. This study proposes and evaluates an initial decision-making algorithm using PSMA PET/CT and mpMRI for selecting candidates suitable for radical prostatectomy without prior biopsy (RP-WPB). Patient enrollment was conducted strictly according to the prospectively established decision-making algorithm. Candidates for RP-WPB were required to fulfill four essential criteria: PSA > 4 ng/mL, PI-RADS score≥4, miPSMA score≥2, and co-positive lesions identified on mpMRI and PSMA PET/CT. Patients staged as cT3-4, cN1, or cM1 (solitary metastasis) underwent RP-WPB directly. For patients with stage cT2N0M0, PSA levels were further stratified: those with PSA ranging from 4 to 30 ng/mL were invited to participate in the prospective study, whereas individuals with PSA ≥ 30 ng/mL qualified for RP-WPB only if they satisfied additional conditions, including age≥75 years, PSA density (PSAD) ≥ 0.2 ng/mL/cm3, and willingness to undergo non-neurovascular-bundle-sparing surgery. From January 2022 to February 2024, 150 patients were prospectively enrolled following the algorithm; 30 patients withdrew, and 120 underwent RP-WPB. Among the latter, 84 patients were classified as cT2N0M0, 27 as cT3-4, 10 as cN1, and 9 as cM1. The detection rate of clinically significant PCa (csPCa) (ISUP grade ≥2) patients was 100

OBJECTIVE:This study aims to develop a cascaded deep learning (DL) system based on multiparametric MRI to establish an automated pipeline for the segmentation and classification of small renal masses (SRMs). MATERIALS AND METHODS:A retrospective collection of SRM patients with pathologically confirmed from three institutions was conducted. MRI data from Institution 1 were randomly divided into a training set and an internal test set. Data from other institutions served as the external test set. A cascaded DL system was developed, incorporating automated segmentation and benign-malignant classification. Diagnostic performance was evaluated using receiver operating characteristic analysis and compared against three radiologists of varying experience. RESULTS:A total of 965 patients with SRM were included. Institution 1 contributed 888 cases, with 712 used for training and 176 as an internal test set; Institutions 2 and 3 provided 77 cases as an external test set. The optimal classification model using automated segmentation labels achieved AUCs of 0.936 and 0.788 on internal and external test sets, respectively. Performance was comparable to models using manual segmentation (internal: 0.936 vs. 0.944, P = 0.671; external: 0.788 vs. 0.832, P = 0.629). On the external test set, the model performed comparably to the senior radiologist, while it significantly outperformed the senior radiologist on the internal test set. The model significantly outperformed the junior radiologist on both test sets. This finding remained consistent in the subgroup of tumors smaller than 3 cm. CONCLUSION:The cascaded DL system demonstrated robust performance across multiple centers, enabling non-invasive and efficient discrimination of SRM malignancy, showing promise as a clinical support tool.

To explore the role of MRI-based habitat radiomics in assessing the metastatic status of renal cell carcinoma (RCC). This study retrospectively collected 241 patients with RCC who underwent nephrectomy and lymphadenectomy at four centers. MRI data from the first center were split into a training set (n = 150) and an internal test set (n = 38); data from the other centers (n = 53) were used for external testing. Based on corticomedullary-phase enhancement and T2WI signal intensity, primary lesions were segmented into 15 habitat subregions. Radiomic features were extracted from the whole-tumor and habitat subregions, respectively. Machine learning algorithms were employed to construct models. Clinical indicators were then integrated to establish a combined model, with performance comparisons and subgroup analyses conducted based on both the internal and external test sets. Among the 241 patients (mean age 53 ± 13 years; 169 males), 36.1

To develop and validate a machine learning (ML)-based pipeline for automated segmentation and classification of complicated cystic renal masses (cCRMs) on MRI. This multicenter retrospective study enrolled 275 patients (median age, 48 years; 85 females) with pathologically confirmed 275 cCRMs (203 malignant) who underwent renal MRI from January 2013 to December 2023. cCRMs from one institution were used as a training set (n = 215), while those from the other three institutions served as a test set (n = 60). 3D V-Net and random forest algorithms were employed for segmentation and classification, respectively. Segmentation and classification performance was evaluated using the Dice similarity coefficient (DSC) and the area under the curve (AUC), respectively. Two junior and two senior radiologists independently classified cCRMs in the test set into Bosniak categories II–IV based on the Bosniak classification, version 2019. In the test set, the ML pipeline achieved DSC of 0.718 for cCRMs (n = 60) on excretory phase images. Additionally, classification performance of the ML pipeline (AUC = 0.835, 95

OBJECTIVES:This study aims to develop an artificial intelligence (AI)-based automated segmentation method for small renal masses (SRMs) using multi-center, multi-scanner, multi-sequence MRI data. METHODS:MR images from 988 pathologically confirmed SRM patients from three different centers were retrospectively included. Segmentation networks were independently developed for each MRI sequence using deep learning techniques. A GE dataset of 733 patients from Center 1 was used for training and validation. A GE test set, consisting of internal (99 from Center 1) and external test sets (81 from Center 2 and 3), was created for evaluation. Furthermore, a non-GE generalization set, consisting of 75 patients from Center 2 and 3, was used to assess the generalization ability. The method's performance was evaluated in terms of detection rate and segmentation accuracy (Dice similarity coefficient [DSC]). Subgroup analysis and multiple linear regression were used for further exploration. RESULTS:Our method demonstrated promising results in the detection and segmentation of SRMs. All patients in the GE test set were correctly detected in at least one sequence. Our model achieved a median DSC of 0.769-0.855 across five MRI sequences and demonstrated reasonable generalization to non-GE scanners (median DSC range: 0.523-0.785). CONCLUSIONS:The implementation of automated segmentation achieved encouraging outcomes in both correct-detection rates and segmentation accuracy across a diverse cohort spanning multiple centers and scanners, suggesting its potential as a key component of future diagnostic pipelines for SRMs.

Purpose: The purpose of this study was to assess the capabilities of MRI-based Node Reporting and Data System (Node-RADS) in diagnosing regional lymph node metastasis (RLNM) and to estimate its prognostic significance in patients with renal cell carcinomas (RCCs). Materials and methods: Patients with RCC who underwent nephrectomy and regional lymph node dissection between January 2010 and August 2023 were retrospectively included. Two senior radiologists scored lymph nodes in consensus using MRI-based Node-RADS. The performance of MRI-based Node-RADS for the diagnosis of RLNM was estimated using area under receiver operating characteristic (AUC) curves and compared against size criteria. Three additional readers scored all lesions to assess interobserver agreement. Progression-free survival and overall survival were estimated and compared between patients with low (1-3) and high (4-5) scores. Results: Overall, 216 patients with RCC were enrolled, including 58 with RLNM. There were 157 men and 59 women with a median age of 54 years (range: 8-83 years). Node-RADS showed larger AUC (0.93 [95 % confidence interval (CI): 0.87-0.97]) and higher specificity (96.8 % [95 % CI: 92.8-99.0]) compared to size criteria (0.88 [95 % CI: 0.83-0.94] and 87.3 % [95 % CI: 81.1-92.1], respectively) for the diagnosis of RLNM (P = 0.039 and P < 0.001, respectively). Substantial interobserver agreement in Node-RADS scoring was obtained between the three readers (weighted kappa, 0.75 [95 % CI: 0.69-0.80]). During a median follow-up of 56 months, patients with high Node-RADS score experienced poorer progression-free survival (P < 0.001) and overall survival (P < 0.001) than those with low Node-RADS score. At multivariable Cox regression analysis, Node-RADS was an independent variable associated with RCC prognosis after adjustment for confounders. Conclusions: The MRI-based Node-RADS demonstrates notable performance in detecting RLNM and showed potential prognostic significance for RCCs.