Background and purpose:Diffuse large B-cell lymphoma(DLBCL)is subclassified into germinal center B-cell-like(GCB)and non-GCB subtypes,which differ in prognosis and treatment response.However,current distinction still relies on invasive pathological assays.This study developed radiomics and deep-learning models based on multiparametric magnetic resonance imaging(MRI)to non-invasively differentiate the two subtypes preoperatively,thereby reducing dependence on histopathological examination.Methods:This study retrospectively included patients with pathologically confirmed DLBCL diagnosed at Huashan Hospital,Fudan University,and other institutions between March 2013 and December 2024.Using multiparametric MRI data,we developed DLBCL-subtype classification models that combined 4 radiomics-based machine-learning algorithms:support vector machine(SVM),logistic regression(LR),Gaussian process(GP)and Naive Bayes(NB),with 3 deep-learning architectures[densely-connected convolutional networks 121(DenseNet121),residual network 101(ResNet101)and EfficientNet-b5].Additionally,two radiologists with different experience levels independently classified DLBCL on MRI in a blinded fashion.Model and radiologist performance were quantified using the area under the receiver operating characteristic curve(AUC),accuracy(ACC),and F1-score to evaluate their ability to distinguish GCB from non-GCB subtypes.This study was approved by the Ethics Committee of Huashan Hospital of Fudan University(No.KY2024-663),and all patients signed informed consents.Results:A total of 173 patients were enrolled(55 with GCB subtype and 118 with non-GCB subtype).Radiomics and deep learning methods effectively distinguished DLBCL subtypes.Among these,the GP radiomics model(based on T1-CE+T2-FLAIR+ADC sequences)and DenseNet121 deep learning model(based on T1-CE+T2-FLAIR+ADC sequences)demonstrated optimal performance.Both achieved excellent results on the internal validation set(GP:AUC=0.900,ACC=0.896,F1=0.840;DenseNet121:AUC=0.846,ACC=0.854,F1=0.774)and maintained robustness on the external validation set.Furthermore,the classification efficacy of the optimal AI model surpassed that of experienced radiologists(highest physician AUC=0.678).Conclusion:Radiomics and deep-learning models based on multiparametric MRI features can effectively differentiate GCB from non-GCB subtypes of DLBCL.Among them,GP and DenseNet121 exhibit outstanding performance,especially when integrating multi-sequence feature sets for classifying DLBCL subtypes on complex imaging data.

Objective:To investigate the dynamic imaging characteristics of hepatocellular carcinoma (HCC) following Yttrium-90 selective internal radiation therapy ( 90Y-SIRT) and to compare these with imaging findings after transarterial chemoembolization (TACE). Methods:This retrospective case-control study included 24 HCC patients who received 90Y-SIRT at Zhongda Hospital, Southeast University, and West China Hospital, Sichuan University, between September 2021 and June 2023, establishing the 90Y-SIRT group. Additionally, 45 HCC patients who underwent their first TACE treatment at Zhongda Hospital, Southeast University during the same period were included as the TACE group. Patients underwent MRI and/or CT follow-ups at 1-3 months (first follow-up) and 3-6 months (second follow-up) after treatment. The analyzed imaging features included tumor characteristics, peritumoral features, and measurements of tumor and liver volumes, with postoperative change rates calculated. Imaging differences between the 90Y-SIRT and TACE groups were statistically compared using the Mann-Whitney U test or χ2 test. Results:At the first follow-up, compared to baseline, a higher proportion of lesions in the 90Y-SIRT group exhibited a reduction in arterial phase enhancement in the viable region (10/13) than in the TACE group (10/29), with a statistically significant difference ( P=0.040). The necrotic region of the tumor on T 1WI showed significantly lower signal intensity in the 90Y-SIRT group than in the TACE group ( Z=2.98, P=0.006). The change in the apparent diffusion coefficient value in the viable region compared to baseline was 157.0×10 -3(-62.0×10 -3, 311.5×10 -3) mm2/s in the 90Y-SIRT group and -56.0×10 -3 (-216.8×10 -3, 110.0×10 -3) mm2/s in the TACE group, with a statistically significant difference ( Z=-2.71, P=0.008). At the first and second follow-up, the contralateral liver lobe volume increased significantly in the 90Y-SIRT group, with a statistically significant difference from the TACE group ( Z=-3.21, -3.78, both P=0.001). Regarding peritumoral imaging characteristics, a statistically significant difference was observed between the two groups in the low signal intensity of the liver lobe or segment where the tumor waslocated during the hepatobiliary phase ( P=0.020, 0.040). Both HCC groups exhibited progressive tumor volume reduction after treatment. In the 90Y-SIRT group, the change rates of lesion volume relative to baseline at the two follow-ups were -23.0% (-45.6%, 7.9%) and -68.7% (-82.7%, -28.5%), respectively. In the TACE group, the values were -29.8% (-53.6%, -2.7%) and -38.0% (-65.3%, -10.7%). The differences between the two groups were not statistically significant ( Z=-0.52, P=0.605; Z=-1.79, P=0.073). Conclusion:There is a statistically significant difference in the tumor imaging features and peritumoral imaging characteristics between 90Y-SIRT and TACE. 90Y-SIRT demonstrates a notable advantage in promoting contralateral liver lobe regeneration while also contributing to tumor size reduction.

Background: s/Aims: Non-invasive models stratifying clinically significant portal hypertension (CSPH) are limited. Herein, we developed a new non-invasive model for predicting CSPH in patients with compensated cirrhosis and investigated whether carvedilol can prevent hepatic decompensation in patients with high-risk CSPH stratified using the new model. Methods: Non-invasive risk factors of CSPH were identified via systematic review and meta-analysis of studies involving patients with hepatic venous pressure gradient (HVPG). A new non-invasive model was validated for various performance aspects in three cohorts, i.e., a multicenter HVPG cohort, a follow-up cohort, and a carvediloltreating cohort. Results: In the meta-analysis with six studies (n=819), liver stiffness measurement and platelet count were identified as independent risk factors for CSPH and were used to develop the new “CSPH risk” model. In the HVPG cohort (n=151), the new model accurately predicted CSPH with cutoff values of 0 and –0.68 for ruling in and out CSPH, respectively. In the follow-up cohort (n=1,102), the cumulative incidences of decompensation events significantly differed using the cutoff values of <–0.68 (low-risk), –0.68 to 0 (medium-risk), and >0 (high-risk). In the carvediloltreated cohort, patients with high-risk CSPH treated with carvedilol (n=81) had lower rates of decompensation events than non-selective beta-blockers untreated patients with high-risk CSPH (n=613 before propensity score matching [PSM], n=162 after PSM). Conclusions Treatment with carvedilol significantly reduces the risk of hepatic decompensation in patients with high-risk CSPH stratified by the new model.

Background: s/Aims: Non-invasive models stratifying clinically significant portal hypertension (CSPH) are limited. Herein, we developed a new non-invasive model for predicting CSPH in patients with compensated cirrhosis and investigated whether carvedilol can prevent hepatic decompensation in patients with high-risk CSPH stratified using the new model. Methods: Non-invasive risk factors of CSPH were identified via systematic review and meta-analysis of studies involving patients with hepatic venous pressure gradient (HVPG). A new non-invasive model was validated for various performance aspects in three cohorts, i.e., a multicenter HVPG cohort, a follow-up cohort, and a carvediloltreating cohort. Results: In the meta-analysis with six studies (n=819), liver stiffness measurement and platelet count were identified as independent risk factors for CSPH and were used to develop the new “CSPH risk” model. In the HVPG cohort (n=151), the new model accurately predicted CSPH with cutoff values of 0 and –0.68 for ruling in and out CSPH, respectively. In the follow-up cohort (n=1,102), the cumulative incidences of decompensation events significantly differed using the cutoff values of <–0.68 (low-risk), –0.68 to 0 (medium-risk), and >0 (high-risk). In the carvediloltreated cohort, patients with high-risk CSPH treated with carvedilol (n=81) had lower rates of decompensation events than non-selective beta-blockers untreated patients with high-risk CSPH (n=613 before propensity score matching [PSM], n=162 after PSM). Conclusions Treatment with carvedilol significantly reduces the risk of hepatic decompensation in patients with high-risk CSPH stratified by the new model.

Background: s/Aims: Non-invasive models stratifying clinically significant portal hypertension (CSPH) are limited. Herein, we developed a new non-invasive model for predicting CSPH in patients with compensated cirrhosis and investigated whether carvedilol can prevent hepatic decompensation in patients with high-risk CSPH stratified using the new model. Methods: Non-invasive risk factors of CSPH were identified via systematic review and meta-analysis of studies involving patients with hepatic venous pressure gradient (HVPG). A new non-invasive model was validated for various performance aspects in three cohorts, i.e., a multicenter HVPG cohort, a follow-up cohort, and a carvediloltreating cohort. Results: In the meta-analysis with six studies (n=819), liver stiffness measurement and platelet count were identified as independent risk factors for CSPH and were used to develop the new “CSPH risk” model. In the HVPG cohort (n=151), the new model accurately predicted CSPH with cutoff values of 0 and –0.68 for ruling in and out CSPH, respectively. In the follow-up cohort (n=1,102), the cumulative incidences of decompensation events significantly differed using the cutoff values of <–0.68 (low-risk), –0.68 to 0 (medium-risk), and >0 (high-risk). In the carvediloltreated cohort, patients with high-risk CSPH treated with carvedilol (n=81) had lower rates of decompensation events than non-selective beta-blockers untreated patients with high-risk CSPH (n=613 before propensity score matching [PSM], n=162 after PSM). Conclusions Treatment with carvedilol significantly reduces the risk of hepatic decompensation in patients with high-risk CSPH stratified by the new model.

Objective:To investigate the dynamic imaging characteristics of hepatocellular carcinoma (HCC) following Yttrium-90 selective internal radiation therapy ( 90Y-SIRT) and to compare these with imaging findings after transarterial chemoembolization (TACE). Methods:This retrospective case-control study included 24 HCC patients who received 90Y-SIRT at Zhongda Hospital, Southeast University, and West China Hospital, Sichuan University, between September 2021 and June 2023, establishing the 90Y-SIRT group. Additionally, 45 HCC patients who underwent their first TACE treatment at Zhongda Hospital, Southeast University during the same period were included as the TACE group. Patients underwent MRI and/or CT follow-ups at 1-3 months (first follow-up) and 3-6 months (second follow-up) after treatment. The analyzed imaging features included tumor characteristics, peritumoral features, and measurements of tumor and liver volumes, with postoperative change rates calculated. Imaging differences between the 90Y-SIRT and TACE groups were statistically compared using the Mann-Whitney U test or χ2 test. Results:At the first follow-up, compared to baseline, a higher proportion of lesions in the 90Y-SIRT group exhibited a reduction in arterial phase enhancement in the viable region (10/13) than in the TACE group (10/29), with a statistically significant difference ( P=0.040). The necrotic region of the tumor on T 1WI showed significantly lower signal intensity in the 90Y-SIRT group than in the TACE group ( Z=2.98, P=0.006). The change in the apparent diffusion coefficient value in the viable region compared to baseline was 157.0×10 -3(-62.0×10 -3, 311.5×10 -3) mm2/s in the 90Y-SIRT group and -56.0×10 -3 (-216.8×10 -3, 110.0×10 -3) mm2/s in the TACE group, with a statistically significant difference ( Z=-2.71, P=0.008). At the first and second follow-up, the contralateral liver lobe volume increased significantly in the 90Y-SIRT group, with a statistically significant difference from the TACE group ( Z=-3.21, -3.78, both P=0.001). Regarding peritumoral imaging characteristics, a statistically significant difference was observed between the two groups in the low signal intensity of the liver lobe or segment where the tumor waslocated during the hepatobiliary phase ( P=0.020, 0.040). Both HCC groups exhibited progressive tumor volume reduction after treatment. In the 90Y-SIRT group, the change rates of lesion volume relative to baseline at the two follow-ups were -23.0% (-45.6%, 7.9%) and -68.7% (-82.7%, -28.5%), respectively. In the TACE group, the values were -29.8% (-53.6%, -2.7%) and -38.0% (-65.3%, -10.7%). The differences between the two groups were not statistically significant ( Z=-0.52, P=0.605; Z=-1.79, P=0.073). Conclusion:There is a statistically significant difference in the tumor imaging features and peritumoral imaging characteristics between 90Y-SIRT and TACE. 90Y-SIRT demonstrates a notable advantage in promoting contralateral liver lobe regeneration while also contributing to tumor size reduction.

Background and purpose:Diffuse large B-cell lymphoma(DLBCL)is subclassified into germinal center B-cell-like(GCB)and non-GCB subtypes,which differ in prognosis and treatment response.However,current distinction still relies on invasive pathological assays.This study developed radiomics and deep-learning models based on multiparametric magnetic resonance imaging(MRI)to non-invasively differentiate the two subtypes preoperatively,thereby reducing dependence on histopathological examination.Methods:This study retrospectively included patients with pathologically confirmed DLBCL diagnosed at Huashan Hospital,Fudan University,and other institutions between March 2013 and December 2024.Using multiparametric MRI data,we developed DLBCL-subtype classification models that combined 4 radiomics-based machine-learning algorithms:support vector machine(SVM),logistic regression(LR),Gaussian process(GP)and Naive Bayes(NB),with 3 deep-learning architectures[densely-connected convolutional networks 121(DenseNet121),residual network 101(ResNet101)and EfficientNet-b5].Additionally,two radiologists with different experience levels independently classified DLBCL on MRI in a blinded fashion.Model and radiologist performance were quantified using the area under the receiver operating characteristic curve(AUC),accuracy(ACC),and F1-score to evaluate their ability to distinguish GCB from non-GCB subtypes.This study was approved by the Ethics Committee of Huashan Hospital of Fudan University(No.KY2024-663),and all patients signed informed consents.Results:A total of 173 patients were enrolled(55 with GCB subtype and 118 with non-GCB subtype).Radiomics and deep learning methods effectively distinguished DLBCL subtypes.Among these,the GP radiomics model(based on T1-CE+T2-FLAIR+ADC sequences)and DenseNet121 deep learning model(based on T1-CE+T2-FLAIR+ADC sequences)demonstrated optimal performance.Both achieved excellent results on the internal validation set(GP:AUC=0.900,ACC=0.896,F1=0.840;DenseNet121:AUC=0.846,ACC=0.854,F1=0.774)and maintained robustness on the external validation set.Furthermore,the classification efficacy of the optimal AI model surpassed that of experienced radiologists(highest physician AUC=0.678).Conclusion:Radiomics and deep-learning models based on multiparametric MRI features can effectively differentiate GCB from non-GCB subtypes of DLBCL.Among them,GP and DenseNet121 exhibit outstanding performance,especially when integrating multi-sequence feature sets for classifying DLBCL subtypes on complex imaging data.

Early-stage diagnosis of liver cancer is challenging, with an overall poor prognosis. The tumor microenvironment of primary liver cancer is complex, exhibiting significant heterogeneity both interpersonally and intratumorally. Therefore, it is of paramount importance to dynamically analyze biological markers in the tumor microenvironment of primary liver cancer in vivo. In recent years, significant progress has been made in the imaging diagnosis and treatment of liver cancer with the development of molecular imaging. Molecular imaging techniques utilize specific nano-imaging probes to evaluate pathological changes of liver cancer at the molecular and cellular levels in real-time. These techniques enable precise imaging to reveal key molecular biomarkers involved in the occurrence and progression of liver cancer, exploring their associations with cancer progression and outcomes. This article focuses on molecular imaging, emphasizing the current research status and latest advancements in the field of liver cancer diagnosis and therapy using techniques such as CT, MRI, optical imaging, PET imaging, and multimodal imaging. It also identifies important future directions and significant challenges for further development.

Objective:To clarify the evidences of hippocampal injury after radiotherapy avoiding hippocampus and explore its relationships with cognition.Methods:A prospective design was adopted in this study.A total of 183 patients with nasopharyngeal carcinoma treated by intensity modulated radiation therapy (IMRT group) and 30 matched healthy control (HC group)were collected in the Affiliated Hospital of Jiangsu University and Southeast University Affiliated Zhongda Hospital from January 2017 to December 2019. All subjects were assessed by Montreal Cognitive Assessment (MoCA-B) at baseline and 6 months after radiotherapy, then the patients with nasopharyngeal carcinoma were divided into cognitive impairment group and non-cognitive impairment group. Subjects were scanned with Siemens 3.0 T MR, and T 1WI was used as analysis sequence.The individual standardized hippocampus ROIs were extracted based on Montreal Neurological Institute(MNI) brain template.All texture features were calculated using the Radiomics developed by C++and Delphi, and the intra group correlation coefficients (ICC), average direction, machine learning (random forest) and autocorrelation matrix were used for reducing the features dimension. One-way ANOVA and generalized linear models were used to compare the differences among different groups. Pearson correlations analyses were used to evaluate the relationships between important texture features and clinical data. Logistic regressions were used to calculate the abilities of texture features to predict cognitive impairment. Results:After 9 patients who lost follow-up were excluded, a total of 164 patients with nasopharyngeal carcinoma were included as IMRT group.Texture features of ROIs were extracted and dimensionally reduced successfully. Five differences features (Variance, Entropy, GlevNonU, RLNonUni and Contrast)were found among HC group, cognitive impairment group and non-cognitive impairment group, and the last three further showed significant differences within IMRT group (GlevNonU, P=0.011;RLNonUni, P<0.001;Contrast, P<0.001). Hippocampal doses were positively correlated with Variance ( r=0.448, P<0.05), and negatively correlated with Entropy ( r=-0.461, P<0.05). There was a positive correlation between MoCA-B scores with GlevNonU, RLNonUniand Contrast ( r=0.503, P<0.05; r=0.587, P<0.05; r=0.531, P<0.05). GlevNonU and Contrast were independent predictors of cognitive impairment in hippocampal avoidance of radiotherapy (OR=0.731, 95%CI 0.610-0.857; OR=0.651, 95%CI 0.496-0.853). Conclusion:Results of texture analysis could be used as micro imaging evidences of hippocampal injury in radiotherapy avoiding hippocampus, and could also effectively predict the occurrences of cognitive impairment.

Objective:To explore the teaching effect of organ-system-based curriculum (OSBC) on cultivating the post competency of radiologists.Methods:Based on the teaching design of OSBC, our study has completed the teaching practice for imaging diagnosis of prostate diseases, focal liver lesions, small pulmonary nodules and intestinal obstruction. The imaging diagnosis of prostate diseases was taken as teaching point. Fifty-two trainees were divided into four groups: junior standardized residents and clinical-type postgraduates (JSRCP) group, senior grade of standardized residents and clinical-type postgraduates (SG-SRCP), advanced training radiologist (ATR) group, intern doctors (ID) group. The teaching framework of pre-training assessment, training and post-training test was designed, and the teaching effect and the operability evaluation of OSBC was compared in terms of test scores and subjective evaluation before and after the training. SPSS 18.0 was used for t test. Results:The test scores after training of four groups were significantly improved compared to the test scores before training. The test scores of SG-SRCP group and ATR group were significantly higher than those of ID group ( F=16.609, P<0.001). The results of subjective evaluation showed that the SG-SRCP and ATR group had the highest degree of satisfaction. Conclusion:OSBC education mode has a good training effectiveness of middle and advanced stages course of medical imaging. In the future teaching, OSBC teaching should be explored among different levels of students.