The global rise in steatotic liver disease poses a significant public health challenge. While non-contrast computed tomography scans hold promise for opportunistic detection of steatotic liver disease, their potential for staging and risk assessment remains underexplored. Here we present a multimodal AI model trained on a large dataset, comprising of (n=968) histopathologically and (n=1103) radiologically confirmed cases, validated against both histology (n=660) and MRI-PDFF (n=375) gold standards, demonstrating high accuracy in detecting mild to severe steatosis (AUC: 0.904-0.929) and clinically significant fibrosis (AUC: 0.824-0.888). Furthermore, integrating the model into the standard clinical pathway improves primary risk screening in a retrospective patient cohort (n=1192), identifying 36% more patients at risk of fibrosis progression. Using Cox proportional hazard model, we observe that the intermediate-high risk patients identified by the optimized clinical pathway exhibits a significantly higher incidence of cirrhosis (hazard ratio: 5.54: 2.69-11.42), showcasing the model's potential for early detection and management of steatotic liver disease.

Accurate tumor analysis is central to clinical radiology and precision oncology, where early detection, reliable lesion characterization, and pathology-level risk assessment directly guide diagnosis, staging, and treatment planning. Chain-of-Thought (CoT) reasoning is particularly critical in this setting, as it enables stepwise interpretation from imaging findings to clinical impressions and pathology-level conclusions, ensuring traceability and reducing diagnostic errors. Here, we target the clinical tumor analysis task and build a large-scale benchmark that operationalizes a multimodal reasoning pipeline, spanning findings, impressions, and pathology predictions. We curate TumorCoT, a large-scale dataset of 1.5M CoT-labeled VQA instructions paired with 3D CT scans, with step-aligned rationales and cross-modal alignments along the “findings → impression → pathology” trajectory, enabling standardized evaluation of both final accuracy and reasoning consistency. We further propose TumorChain, a multimodal interleaved reasoning framework that tightly couples 3D imaging encoders, clinical text understanding, and organ-level vision-language alignment. Through cross-modal alignment and iterative interleaved causal reasoning, TumorChain grounds visual evidence, aggregates conclusions, and issues pathology predictions after multiple rounds of self-refinement, improving traceability and reducing hallucination risk. TumorChain demonstrates consistent gains over strong unimodal and pipeline baselines in lesion detection, impression quality, and pathology classification, and successfully generalizes to the public DeepTumorVQA benchmark. Ablations validate the key contributions of interleaved reasoning and clinical CoT. Clinically, these advances lay the groundwork for reliable, interpretable tumor assessment to support real-world decision-making. We release the task, benchmark, and evaluation protocol to advance safe, explainable, and reproducible multimodal reasoning for high-stakes tumor analysis. Our project is available at https://anonymous.4open.science/r/TumorChain-D6E6.

Liver malignancies are frequently evaluated on contrast-enhanced computed tomography (CE-CT), but missed or delayed diagnoses remain a clinically important challenge in high-volume, real-world radiology workflows, highlighting the need for scalable diagnostic safety net approaches. To address this, we developed the Liver DiagnOsis Network (LiON), a CE-CT-based artificial intelligence (AI) system that supports flexible multiphase processing, clinical data integration and workflow-compatible liver malignancy diagnosis. LiON was trained on 6,443 patients and retrospectively validated across 22,251 patients from multicenter and real-world cohorts. LiON achieved high performance for malignancy diagnosis, with an area under the receiver operating characteristic curve (AUC) of 0.975 (95% confidence interval (CI): 0.971-0.979), and maintained robust performance in real-world cohorts and among patients with hepatic steatosis (AUC 0.971, 95% CI: 0.952-0.985) and cirrhosis (AUC 0.924, 95% CI: 0.901-0.946). We then conducted a single-arm trial in 10,333 patients in routine clinical practice, in which LiON functioned as an additional AI reader within the existing clinical workflow. The trial met its primary endpoint, defined as an AUC for malignancy diagnosis with the lower bound of the 95% CI exceeding 0.900, achieving an AUC of 0.952 (95% CI: 0.942-0.961). Secondary outcomes demonstrated that AI-human collaboration identified 51 previously overlooked lesions (15 malignancies) and triggered 37 amended radiology reports, 22 multidisciplinary team escalations and clinical management changes in a subset of patients. These findings suggest that AI, when deployed as a workflow-compatible diagnostic support, may help reduce missed or delayed diagnoses and guide clinical interventions. Nevertheless, further evidence from prospective comparative studies across diverse healthcare systems is warranted to assess effects on clinical outcomes. ClinicalTrials.gov identifier: NCT07153783 .

Medical images require comprehensive and accurate interpretation to support the diagnosis of diverse clincial conditions. Recent vision-language generalist models offer broad task coverage and promising zero-shot capabilities, yet often lack fine-grained anatomical and lesion awareness for reliable diagnosis and spatial interpretability. In contrast, supervised specialist models achieve strong performance on specific tasks but typically lack generalization across diseases and anatomies. In this work, we present SuG, a Super-Generalist framework that unifies generalist vision-language learning with specialist objectives, enabling both broad generalization and specialist-level diagnostic capability. We perform specialist-enhanced vision-language alignment in SuG by incorporating spatial priors from multiple segmentation experts, including anatomy, class-specific lesion and class-agnostic lesion segmentors that captures lesions beyond anatomies annotated during training. To improve lesion grounding capability, we leverage lesion masks as spatial priors to calibrate text-conditioned visual attention, encouraging disease-related semantics to focus on clinically relevant regions. We evaluate SuG on extensive chest and abdominal CT benchmarks, including CT-RATE, Merlin, MedVL-CT69K, and several in-house tumor datasets. SuG achieves state-of-the-art performance across a wide range of disease diagnosis tasks and surpasses specialist models on several critical tumor diagnosis benchmarks. Furthermore, SuG demonstrates strong lesion grounding capability, including robust generalization to lesion types lacking class-specific supervision.

Background:Currently, there is no fully automated tool available for evaluating the degree of cervical spinal stenosis. The aim of this study was to develop and validate the use of artificial intelligence (AI) algorithms for the assessment of cervical spinal stenosis. Methods:In this retrospective multi-center study, cervical spine magnetic resonance imaging (MRI) scans obtained from July 2020 to June 2023 were included. Studies of patients with spinal instrumentation or studies with suboptimal image quality were excluded. Sagittal T2-weighted images were used. The training data from the Fourth People's Hospital of Shanghai (Hos. 1) and Shanghai Changzheng Hospital (Hos. 2) were annotated by two musculoskeletal (MSK) radiologists following Kang's system as the standard reference. First, a convolutional neural network (CNN) was trained to detect the region of interest (ROI), with a second Transformer for classification. The performance of the deep learning (DL) model was assessed on an internal test set from Hos. 2 and an external test set from Shanghai Changhai Hospital (Hos. 3), and compared among six readers. Metrics such as detection precision, interrater agreement, sensitivity (SEN), and specificity (SPE) were calculated. Results:Overall, 795 patients were analyzed (mean age ± standard deviation, 55±14 years; 346 female), with 589 in the training (75%) and validation (25%) sets, 206 in the internal test set, and 95 in the external test set. Four tasks with different clinical application scenarios were trained, and their accuracy (ACC) ranged from 0.8993 to 0.9532. When using a Kang system score of ≥2 as a threshold for diagnosing central cervical canal stenosis in the internal test set, both the algorithm and six readers achieved similar areas under the receiver operating characteristic curve (AUCs) of 0.936 [95% confidence interval (CI): 0.916-0.955], with a SEN of 90.3% and SPE of 93.8%; the AUC of the DL model was 0.931 (95% CI: 0.917-0.946), with a SEN in the external test set of 100%, and a SPE of 86.3%. Correlation analysis comparing the DL method, the six readers, and MRI reports between the reference standard showed a moderate correlation, with R values ranging from 0.589 to 0.668. The DL model produced approximately the same upgrades (9.2%) and downgrades (5.1%) as the six readers. Conclusions:The DL model could fully automatically and reliably assess cervical canal stenosis using MRI scans.

Objective:To explore the predictive value of the Neutrophil-to-Lymphocyte Ratio (NLR) in the prognosis of patients with Lumbar Disc Herniation (LDH) undergoing Percutaneous Intradiscal Radiofrequency Thermocoagulation (PIRFT). Methods:A total of 121 patients with LDH undergoing PIRFT treatment were selected, ranging in age from 35 to 65 years old, with no gender restrictions. Blood samples were collected in the morning after admission while fasting, and the absolute neutrophil and lymphocyte counts in the blood were detected using the enzyme-linked immunosorbent assay (ELISA) method to calculate the Neutrophil-to-Lymphocyte Ratio (NLR). Patients were divided into two groups according to the modified Macnab criteria: the Effective group (E group) and the Invalid group (I group). The Visual Analogue Scale (VAS), Oswestry Disability Index (ODI), and Japanese Orthopaedic Association (JOA) scores were used to assess the pain level and activity ability of the patients before treatment and at 90 days and 180 days post-treatment. The correlation between NLR and ODI, JOA scores was analyzed using rank correlation analysis, and the predictive value of NLR for the therapeutic effect of PIRFT was analyzed using the Receiver Operating Characteristic (ROC) curve. Results:A total of 121 patients were ultimately enrolled. Based on the follow-up results at 90 days post-surgery, there were 110 cases in the effective group (E group) and 11 in the ineffective group (I group). The results showed that before treatment, the NLR levels in the E group were significantly lower than those in the I group (P < 0.05), and there were no significant differences in ODI and JOA scores between the two groups (P > 0.05). Ninety days after treatment, the NLR levels in the I group remained significantly higher than those in the E group (P < 0.05), and the E group's ODI and JOA scores showed significant improvement compared to before treatment (P < 0.05). In contrast, the I group only showed improvement in ODI scores (P < 0.05), with no significant change in JOA scores (P > 0.05). Additionally, the I group's ODI scores were significantly higher than those of the E group (P < 0.05), and their JOA scores were significantly lower than those of the E group (P < 0.05). All patients in the I group underwent a second radiofrequency ablation treatment. A comparison was made again after 180 days of treatment, and there were no significant differences in NLR levels between the two groups (P > 0.05). Both groups showed improvement in ODI and JOA scores compared to before treatment (P < 0.05), with no significant differences between the groups (P > 0.05). Rank correlation analysis showed that preoperative NLR levels were positively correlated with ODI scores at 90 days after PIRFT treatment (r = 0.386, P < 0.01) and negatively correlated with JOA scores (r = -0.326, P = 0.003). The results of the ROC curve analysis showed that the area under the ROC curve (AUC) was 0.803, the optimal diagnostic cutoff point was 1.975, the correct diagnosis index was 0.518, 95% CI: 0.650-0.955, P < 0.001. Conclusion:Preoperative inflammatory levels are one of the factors affecting the treatment of lumbar disc herniation with percutaneous radiofrequency ablation, and the Neutrophil-to-Lymphocyte Ratio (NLR) helps to predict the effectiveness of percutaneous disc radiofrequency ablation in treating lumbar disc herniation.

In this retrospective study, whether [68Ga]Ga-DOTA-FAPI-04 PET/MR imaging biomarkers can predict the progression-free survival (PFS) and overall survival (OS) of patients with advanced pancreatic cancer was investigated. Fifty-one patients who underwent [68Ga]Ga-DOTA-FAPI-04 PET/MR scans before first-line chemotherapy were recruited. Imaging biomarkers, including the maximum tumor diameter, minimum apparent diffusion coefficient (ADC), maximum and mean standardized uptake values (SUVmax and SUVmean), fibroblast activation protein- (FAP-) positive tumor volume (FTV and W-FTV) and total lesion FAP expression (TLF and W-TLF), were recorded for primary and whole-body tumors. A subgroup analysis of 28 patients with obstructive inflammation was performed, and the maximum and mean standardized uptake values (D-SUVmax and D-SUVmean), FAP-positive uptake volume (D-FTV), and total FAP expression (D-TLF) in the distal pancreas were assessed. Kaplan–Meier analysis and the Cox proportional hazards model were used to assess the relationships between these imaging biomarkers and PFS/OS. SUVmax was negatively correlated with the ADC (r = -0.288, P = 0.041), whereas tumor length was positively correlated with the FTV, TLF, W-FTV, and W-TLF (r = 0.311–0.508, P < 0.05). The median PFS was not reached in patients with a W-TLF ≤ 516.09 but was 91 days in patients with a W-TLF > 516.09 (P < 0.001). Univariate and multivariate Cox regression analyses identified W-TLF as an independent predictor of PFS (P = 0.001, hazard ratio (HR) = 4.949). The TNM stage, tumor length, ADC value, SUVmax, TLF, W-FTV, and W-TLF were significantly associated with OS (P < 0.05). Multivariate analysis further confirmed that tumor length, ADC, and W-TLF were independent predictors of OS (P < 0.05). A subgroup analysis including 28 patients with obstructive pancreatitis revealed that TNM stage, tumor length, W-FTV, W-TLF, D-FTV, and D-TLF were significantly associated with OS (P < 0.05). The multivariate analysis further identified W-TLF and D-FTV as independent predictors of OS (P < 0.05). [68Ga]Ga-DOTA-FAPI-04 PET/MRI biomarkers are associated with the PFS and OS of pancreatic cancer patients. Both the ADC and W-TLF are independent risk factors for patients with advanced disease. Increased fibroblast activity, driven by cancer-induced inflammation, may influence long-term survival outcomes following adjuvant therapy. These biomarkers have potential for guiding future clinical trials, enabling personalized treatment strategies, and ultimately improving the management and prognosis of patients with pancreatic cancer.

Rationale and Objectives To improve the diagnostic recognition of papillary renal neoplasm with reverse polarity (PRNRP) through comprehensive analysis of computed tomography (CT) and magnetic resonance imaging (MRI) findings. Materials and Methods A retrospective multi-center study was conducted on patients with pathologically confirmed PRNRPs from 2019 to 2024, encompassing six institutions. Clinical and pathological data were meticulously documented. Preoperative CT (n=23) and MRI (n=9) features were independently evaluated in consensus by two genitourinary radiologists, focusing on tumor location, morphologic features, attenuation, signal intensity, and enhancement patterns. Postoperative outcomes were assessed through medical record review or telephone follow-up. Results The study cohort comprised 26 patients (mean age 62±12 years, 13 men) with 26 well-defined PRNRPs (mean diameter 2.2±1.0 cm) were included. 15 cases (58%) were situated in the right kidney, 18(69%) were exophytic, and 23(88%) were quasi-spherical. A pseudocapsule was identified in eight cases (89%) on MRI. All cases demonstrated iso- or slight hyperattenuation (43.1±13.6 HU) on non-contrast CT, hypointensity on T2-weighted imaging (T2WI), and mild diffusion restriction on diffusion-weighted imaging (DWI). All cases exhibited mild or moderate enhancement in the corticomedullary phase, followed by progressive enhancement in the nephrographic and excretory phases. Concomitant renal cysts were found in 17 cases (65%). All cases showed no evidence of recurrence or metastasis. Conclusion PRNRP typically presents as a small hypovascular renal mass, and characterized by a pseudocapsule, iso- or slight hyperattenuation on non-contrast CT, heterogeneous T2WI hypointensity, and mild diffusion restriction on DWI.

This paper aims to develop a nonrigid registration method of preoperative and intraoperative thoracoabdominal CT images in computer-assisted interventional surgeries for accurate tumor localization and tissue visualization enhancement. However, fine structure registration of complex thoracoabdominal organs and large deformation registration caused by respiratory motion is challenging. To deal with this problem, we propose a 3D multi-scale attention VoxelMorph (MA-VoxelMorph) registration network. To alleviate the large deformation problem, a multi-scale axial attention mechanism is utilized by using a residual dilated pyramid pooling for multi-scale feature extraction, and position-aware axial attention for long-distance dependencies between pixels capture. To further improve the large deformation and fine structure registration results, a multi-scale context channel attention mechanism is employed utilizing content information via adjacent encoding layers. Our method was evaluated on four public lung datasets (DIR-Lab dataset, Creatis dataset, Learn2Reg dataset, OASIS dataset) and a local dataset. Results proved that the proposed method achieved better registration performance than current state-of-the-art methods, especially in handling the registration of large deformations and fine structures. It also proved to be fast in 3D image registration, using about 1.5 s, and faster than most methods. Qualitative and quantitative assessments proved that the proposed MA-VoxelMorph has the potential to realize precise and fast tumor localization in clinical interventional surgeries.

Purpose: This study aimed to investigate the value of Ga-68-fibroblast activation protein inhibitor (FAPI) PET/MR semiquantitative parameters in the prediction of tumor response and resectability after neoadjuvant therapy in patients with pancreatic cancer. Patients and Methods: This study was performed retrospectively in patients with borderline resectable or locally advanced pancreatic cancer who underwent Ga-68-FAPI PET/MRI from June 2020 to June 2022. The SUVmax, SUVmean, SUVpeak, uptake tumor volume (UTV), and total lesion FAP expression (TLF) of the primary tumor were recorded. The target-to-background ratios (TBRs) of the primary tumor to normal tissue muscle (TBRmuscle) and blood (TBRblood) were also calculated. In addition, the minimum apparent diffusion coefficient value of the tumor was measured. After 3-4 cycles of gemcitabine + nab-paclitaxel chemotherapy, patients were divided into responders and nonresponders groups according to RECIST criteria (v.1.1). They were also divided into resectable and unresectable groups according to the surgical outcome. The variables were compared separately between groups. Results: A total of 18 patients who met the criteria were included in this study. The UTV and TLF were significantly higher in nonresponders than in responders (P < 0.05). The SUVmax, SUVmean, and TBRmuscle were significantly higher in unresectable patients than in resectable ones (P < 0.05). Receiver operating characteristic curve analysis identified UTV (area under the curve [AUC] = 0.840, P = 0.015) and TLF (AUC = 0.877, P = 0.007) as significant predictors for the response to gemcitabine + nab-paclitaxel chemotherapy, with cutoff values of 25.05 and 167.38, respectively. In addition, SUVmax (AUC = 0.838, P = 0.016), SUVmean (AUC = 0.812, P = 0.026), and TBRmuscle (AUC = 0.787, P = 0.041) were significant predictors of the resectability post-NCT, with cutoff values of 14.0, 6.0, and 13.9, respectively. According to logistic regression analysis, TLF was found to be significantly associated with tumor response (P = 0.032) and was an independent predictor of tumor response (P = 0.032). In addition, apparent diffusion coefficient value was an independent predictor of tumor resectability (P = 0.043). Conclusions: This pilot study demonstrates the value of Ga-68-FAPI PET/MR for the prediction of tumor response and resectability after neoadjuvant therapy. It may aid in individualized patient management by guiding the treatment regimens.

This study aimed to investigate the value of 68Ga-fibroblast activation protein inhibitor (FAPI) PET/MR semiquantitative parameters in the prediction of tumor response and resectability after neoadjuvant therapy in patients with pancreatic cancer. This study was performed retrospectively in patients with borderline resectable or locally advanced pancreatic cancer who underwent 68Ga-FAPI PET/MRI from June 2020 to June 2022. The SUVmax, SUVmean, SUVpeak, uptake tumor volume (UTV), and total lesion FAP expression (TLF) of the primary tumor were recorded. The target-to-background ratios (TBRs) of the primary tumor to normal tissue muscle (TBRmuscle) and blood (TBRblood) were also calculated. In addition, the minimum apparent diffusion coefficient value of the tumor was measured. After 3–4 cycles of gemcitabine + nab-paclitaxel chemotherapy, patients were divided into responders and nonresponders groups according to RECIST criteria (v.1.1). They were also divided into resectable and unresectable groups according to the surgical outcome. The variables were compared separately between groups. A total of 18 patients who met the criteria were included in this study. The UTV and TLF were significantly higher in nonresponders than in responders (P < 0.05). The SUVmax, SUVmean, and TBRmuscle were significantly higher in unresectable patients than in resectable ones (P < 0.05). Receiver operating characteristic curve analysis identified UTV (area under the curve [AUC] = 0.840, P = 0.015) and TLF (AUC = 0.877, P = 0.007) as significant predictors for the response to gemcitabine + nab-paclitaxel chemotherapy, with cutoff values of 25.05 and 167.38, respectively. In addition, SUVmax (AUC = 0.838, P = 0.016), SUVmean (AUC = 0.812, P = 0.026), and TBRmuscle (AUC = 0.787, P = 0.041) were significant predictors of the resectability post-NCT, with cutoff values of 14.0, 6.0, and 13.9, respectively. According to logistic regression analysis, TLF was found to be significantly associated with tumor response (P = 0.032) and was an independent predictor of tumor response (P = 0.032). In addition, apparent diffusion coefficient value was an independent predictor of tumor resectability (P = 0.043). This pilot study demonstrates the value of 68Ga-FAPI PET/MR for the prediction of tumor response and resectability after neoadjuvant therapy. It may aid in individualized patient management by guiding the treatment regimens.

To assess the diagnostic performance of [68Ga]Ga-DOTA-FAPI-04 PET/MR imaging in the preoperative evaluation of pancreatic cancer and compare it with that of [18F]-FDG PET/CT plus contrast-enhanced CT (CECT). Thirty-one patients with pancreatic cancer underwent preoperative [68Ga]Ga-DOTA-FAPI-04 PET/MR, [18F]-FDG PET/CT, and CECT imaging. Two nuclear medicine physicians independently reviewed two sets of images (set 1, [68Ga]Ga-DOTA-FAPI-04 PET/MR; set 2, [18F]-FDG PET/CT plus CECT) and reached a consensus on tumour resectability, N staging (N0 or N positive) and M staging (M0 or M1). Based on the above indices, the resectability of the tumour was determined according to a five-point scale. Clinical, operative, and pathological findings were used as a reference standard to compare the diagnostic performance of the two imaging sets via the McNemar test. The diagnostic performance of [68Ga]Ga-DOTA-FAPI-04 PET/MR imaging was not significantly different from that of [18F]-FDG PET/CT plus CECT imaging in the assessment of tumour resectability (area under the receiver operating characteristic curve: 0.854 vs. 0.775, p = 0.192), N staging [accuracy: 82.4

Motivation: Cystic fluid appears hyperintense via T2WI, the most sensitive detection method and T2WI is a conventional sequence. However, distinguishing pancreatic MCNs from SCNs using T2WI is difficult because both neoplasms appear as hyperintense lesions, especially when both are unilocular. Goal(s): We aimed to develop and validate a T2WI radiomics nomogram for the differentiation of SCNs from MCNs. Approach: A radiomics model that was included clinical characteristics, MRI characteristics, and T2WI rad-scores for differentiating MCNs from SCN. Results: We developed and validated a T2WI radiomics nomogram that functions as a non-invasive and convenient tool for preoperatively predicting the presence of SCNs and MCNs. Impact: The tool has the potential to help clinicians identify patients requiring surveillance or surgery.

Pancreatic ductal adenocarcinoma (PDAC), the most deadly solid malignancy, is typically detected late and at an inoperable stage. Early or incidental detection is associated with prolonged survival, but screening asymptomatic individuals for PDAC using a single test remains unfeasible due to the low prevalence and potential harms of false positives. Non-contrast computed tomography (CT), routinely performed for clinical indications, offers the potential for large-scale screening, however, identification of PDAC using non-contrast CT has long been considered impossible. Here, we develop a deep learning approach, pancreatic cancer detection with artificial intelligence (PANDA), that can detect and classify pancreatic lesions with high accuracy via non-contrast CT. PANDA is trained on a dataset of 3,208 patients from a single center. PANDA achieves an area under the receiver operating characteristic curve (AUC) of 0.986–0.996 for lesion detection in a multicenter validation involving 6,239 patients across 10 centers, outperforms the mean radiologist performance by 34.1% in sensitivity and 6.3% in specificity for PDAC identification, and achieves a sensitivity of 92.9% and specificity of 99.9% for lesion detection in a real-world multi-scenario validation consisting of 20,530 consecutive patients. Notably, PANDA utilized with non-contrast CT shows non-inferiority to radiology reports (using contrast-enhanced CT) in the differentiation of common pancreatic lesion subtypes. PANDA could potentially serve as a new tool for large-scale pancreatic cancer screening.

Objective:To develop and verify a predictive model based on CT characteristics for predicting infected walled-off necrosis (IWON) in MSAP and SAP patients.Methods:The clinical and CT data of 1 322 patients diagnosed as MSAP and SAP according to the 2012 Atlanta revised diagnostic criteria in the First Affiliated Hospital of Naval Medical University from January 2015 to December 2020 were continuously collected. Finally, 126 patients who underwent enhanced CT scans within 3 days after admission and percutaneous catheter drainage of WON during hospitalization were enrolled. Among them, there were 63 MSAP and 63 SAP patients. According to the results of the culture from drainage fluid, the patients were divided into sterile walled-off necrosis group (SWON group, n=31) and infected walled-off necrosis group (IWON group, n=95). Patients were divided into training set (18 patients with SWON and 74 patients with IWON from January 2015 to December 2018) and validation set (13 patients with SWON and 21 patients with IWON from January 2019 to December 2020). Univariate and multivariate logistic regression analysis were performed to establish a model for predicting IWON. The model was visualized as a nomogram. The receiver operating characteristic curve (ROC) was drawn. The predictive efficacy of the model was evaluated by the area under the curve (AUC), sensitivity, specificity and accuracy, and the clinical application value was judged by decision curve analysis (DCA). Results:Univariate regression analysis showed that age, etiology, WON with bubble sign and the lowest CT value of WON were significantly associated with IWON. Multivariate logistic regression analysis showed that older age, biliary acute pancreatitis, WON with bubble sign, and the greater minimum CT value of WON were independent predictors for IWON. The formula for the prediction model was 0.12+ 0.01 age-0.75 hyperlipidemia-1.62 alcoholic-2.62 other causes+ 19.18 WON bubble sign+ 0.10 minimum CT value of WON. The AUC, sensitivity, specificity, and accuracy of the model were 0.85 (95% CI 0.76-0.94), 67.57%, 88.89%, and 71.74% in the training set and 0.78(95% CI0.62-0.94), 66.67%, 84.62%, and 73.53% in the validation set, respectively. The decision analysis curve showed that when the nomogram differentiated IWON from SWON at a rate greater than 0.38, using the nomogram could benefit the patients. Conclusions:The prediction model established based on CT characteristics might non-invasively and accurately predict the presence or absence of IWON in MSAP and SAP patients, and provide a basis for guiding treatment and evaluating prognosis.

This study aims to evaluate the utility of calculated computed tomography (CT) attenuation value ratio (AVR) and enhancement pattern in distinguishing pancreatic solid serous cystadenomas (SCAs) from nonfunctional pancreatic neuroendocrine tumors (NF-pNETs). A total of 142 consecutive patients with 22 solid SCAs and 120 NF-pNETs confirmed by pathology were included in this retrospective study. All patients underwent preoperative contrast-enhanced CT and were categorized into 2 groups, solid SCA and NF-pNET groups. Patients with NF-pNETs were matched to patients with solid SCAs via propensity scores. AVR was measured and defined as: attenuation value of tumor/attenuation value of normal pancreas. AVR and enhancement pattern performance were assessed according to the discriminative abilities of patients. After matching, 29 patients were allocated to the NF-pNET group. Before matching, sex, age, and the peak enhanced value phase were significantly different between solid SCA and NF-pNET patients (P < .05). After matching, no significant difference was observed between both groups (P > .05). Solid SCAs AVRs were significantly smaller than NF-pNETs AVRs in all unenhanced, arterial, portal venous, and delayed phases (P < .05). Solid SCAs showed significantly more wash-in and wash-out enhancement patterns than NF-pNETs (P < .05). For unenhanced, arterial, portal venous, and delayed phases, and enhancement pattern, the area under the curve (AUC) values were 0.96, 0.72, 0.80, 0.85, and 0.86, respectively. Low AVR on unenhanced CT and wash-in and wash-out enhancement patterns were useful for differentiating solid SCAs from NF-pNETs and may be useful for clinical decisions, a clearer opinion will be formed with further studies to be conducted with larger patient numbers.

Objectives: We aimed to develop and validate a multimodality radiomics model for the preoperative prediction of nonfunctional pancreatic neuroendocrine tumor (NF-pNET) grade (G). Methods: This retrospective study assessed 123 patients with surgically resected, pathologically confirmed NF-pNETs who underwent multidetector computed tomography and MRI scans between December 2012 and May 2020. Radiomic features were extracted from multidetector computed tomography and MRI. Wilcoxon rank-sum test and Max-Relevance and Min-Redundancy tests were used to select the features. The linear discriminative analysis (LDA) was used to construct the four models including a clinical model, MRI radiomics model, computed tomography radiomics model, and mixed radiomics model. The performance of the models was assessed using a training cohort (82 patients) and a validation cohort (41 patients), and decision curve analysis was applied for clinical use. Results: We successfully constructed 4 models to predict the tumor grade of NF- pNETs. Model 4 combined 6 features of T2-weighted imaging radiomics features and 1 arterial-phase computed tomography radiomics feature, and showed better discrimination in the training cohort (AUC = 0.92) and validation cohort (AUC = 0.85) relative to the other models. In the decision curves, if the threshold probability was 0.07-0.87, the use of the radiomics score to distinguish NF-pNET G1 and G2/3 offered more benefit than did the use of a "treat all patients" or a "treat none" scheme in the training cohort of the MRI radiomics model. Conclusion: The LDA classifier combining multimodality images may be a valuable noninvasive tool for distinguishing NF-pNET grades and avoid unnecessary surgery.

We sought to assess the performance of 68 Ga-FAPI-04 PET/MR for the diagnosis of primary tumours as well as metastatic lesions in patients with pancreatic cancer and to compare the results with those of 18F-FDG PET/CT. Prospectively, we evaluated 33 patients suspected to have pancreatic adenocarcinoma, of whom thirty-two were confirmed by histopathology, and one had autoimmune pancreatitis confirmed by needle biopsy and glucocorticoid treatment. Within 1 week, each patient underwent both 68 Ga-FAPI-04 PET/MR and 18F-FDG PET/CT. Comparisons of the detection abilities for primary tumours, lymph nodes, and metastases were conducted for the two imaging approaches. The original maximum standard uptake values (SUVmax) and normalised SUVmax (SUVmax/SUVbkgd) of paired lesions on 68 Ga-FAPI-04 PET/MR and 18F-FDG PET/CT were measured and compared. Thirty pancreatic cancer patients and three pancreatitis patients were enrolled. 68 Ga-FAPI-04 PET/MR and 18F-FDG PET/CT exhibited equivalent (100%) detection rates for primary tumours. The original/normalised SUVmax of primary tumours on 68 Ga-FAPI-04 PET was markedly higher than that on 18F-FDG (p < 0.05). Sixteen pancreatic cancer patients had pancreatic parenchymal uptake, whereas 18F-FDG PET images showed parenchymal uptake in only four patients (53.33% vs. 13.33%, p < 0.001). 68 Ga-FAPI-04 PET detected more positive lymph nodes than 18F-FDG PET (42 vs. 30, p < 0.001), while 18F-FDG PET was able to detect more liver metastases than 68 Ga-FAPI-04 (181 vs. 104, p < 0.001). In addition, multisequence MR imaging helped explain ten pancreatic cancers that could not be definitively revealed due to 68 Ga-FAPI-04 inflammatory uptake and identified more liver metastases than 18F-FDG (256 vs. 181, p < 0.001). 68 Ga-FAPI-04 PET might be better than 18F-FDG PET in the detection of suspicious lymph node metastases. MR multiple sequence imaging of 68 Ga-FAPI-04 PET/MR was helpful for explaining pancreatic lesions in patients with obstructive inflammation and detecting tiny liver metastases.

Purpose: To evaluate the diagnostic performance of the radiomics score (rad-score) for differentiating focal-type autoimmune pancreatitis (fAIP) from pancreatic ductal adenocarcinoma (PDAC). Methods: This retrospective review included 42 consecutive patients with fAIP diagnosed according to the International Consensus Diagnostic Criteria between January 2011 and December 2018. Furthermore, 334 consecutive patients with PDAC confirmed by pathology were also reviewed during the same period. Patients with PDAC and fAIP were matched via propensity score matching (PSM). All patients underwent multidetector computed tomography (MDCT). For each patient, 1409 radiomics features of the portal phase were extracted and reduced using the least absolute shrinkage and selection operator (LASSO) logistic regression algorithm. The portal rad-score performance was assessed based on its discriminative ability. Results: After PSM, we matched 55 patients with PDAC to 42 patients with fAIP, based on clinical and CT characteristics (e.g., patient age, sex, body mass index, location, size, enhanced mode). A rad-score for discriminating fAIP from PDAC, which contained four CT derived radiomic features, was developed (area under the curve = 0.97). The sensitivity, specificity, and accuracy of the radiomics model were 95.24%, 92.73% and 0.94, respectively. Conclusion: The portal rad-score can accurately and noninvasively differentiate fAIP from PDAC.

Objective: To develop an imaging-derived biomarker for prediction of overall survival (OS) of pancreatic cancer by analyzing preoperative multiphase contrast-enhanced computed topography (CECT) using deep learning. Background: Exploiting prognostic biomarkers for guiding neoadjuvant and adjuvant treatment decisions may potentially improve outcomes in patients with resectable pancreatic cancer. Methods: This multicenter, retrospective study included 1516 patients with resected pancreatic ductal adenocarcinoma (PDAC) from 5 centers located in China. The discovery cohort (n=763), which included preoperative multiphase CECT scans and OS data from 2 centers, was used to construct a fully automated imaging-derived prognostic biomarker—DeepCT-PDAC—by training scalable deep segmentation and prognostic models (via self-learning) to comprehensively model the tumor-anatomy spatial relations and their appearance dynamics in multiphase CECT for OS prediction. The marker was independently tested using internal (n=574) and external validation cohorts (n=179, 3 centers) to evaluate its performance, robustness, and clinical usefulness. Results: Preoperatively, DeepCT-PDAC was the strongest predictor of OS in both internal and external validation cohorts [hazard ratio (HR) for high versus low risk 2.03, 95% confidence interval (CI): 1.50–2.75; HR: 2.47, CI: 1.35–4.53] in a multivariable analysis. Postoperatively, DeepCT-PDAC remained significant in both cohorts (HR: 2.49, CI: 1.89–3.28; HR: 2.15, CI: 1.14–4.05) after adjustment for potential confounders. For margin-negative patients, adjuvant chemoradiotherapy was associated with improved OS in the subgroup with DeepCT-PDAC low risk (HR: 0.35, CI: 0.19–0.64), but did not affect OS in the subgroup with high risk. Conclusions: Deep learning-based CT imaging-derived biomarker enabled the objective and unbiased OS prediction for patients with resectable PDAC. This marker is applicable across hospitals, imaging protocols, and treatments, and has the potential to tailor neoadjuvant and adjuvant treatments at the individual level.