http://med-sci.cn/cgzcx/en/contents_list.asp?issue=202511 zh-cn An RSS feed for Chinese Journal of Magnetic Resonance Imaging http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.001 Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.002 ObjectiveTo develop and validate a prognostic model for heart failure with reduced ejection fraction (HFrEF) patients through the integration of cardiac magnetic resonance (CMR) cine-based radiomics with clinical and imaging characteristics. Materials and MethodsThis study retrospectively enrolled 503HFrEF patients diagnosed according to guidelines and undergoing CMR between January 2018 and April 2023. Clinical baseline data, laboratory results, electrocardiograms, and echocardiographic parameters were collected as part of electronic health records (EHR), with follow-up for adverse cardiovascular events, including cardiac death, heart failure rehospitalization, and cardiac transplantation. All patients went through standardized CMR examination. The unsupervised nnU-Netv2 algorithm was employed to extract functional parameters from the CMR cine sequences as imaging features. Additionally, radiomic features were obtained from the same sequences with an open-source software package. After intra- and inter-group consistency testing, features were reduced via minimum redundancy maximum relevance analysis. Classifier with the best performance was selected to build the model. Models combining radiomics with imagingclinical data and standalone radiomics models were developed. The predictive power of the model was assessed by area under the curve (AUC), precision, recall, and F1- score. ResultsAfter applying stringent inclusion and exclusion criteria, a total of 389 patients with HFrEF were enrolled for model development. Of the patients followed for a median of 1041 days (IQR: 212, 1238), 87 (22.4%) experienced the endpoint. The median survival time was 495 days (IQR: 8, 1900). Twelve clinical features were identified via univariable Cox regression, which included NYHA class Ⅲ/Ⅳ and BNP. Subsequently, feature selection and dimensionality reduction yielded a final set of four imaging and nine radiomic features. Ensemble learning (EL) demonstrated optimal performance across the models. Superior prognostic performance was attained by the combined radiomics and imaging features model generated by EL classifier, which yielded an AUC of 0.789, an accuracy of 81.6%, a precision of 72.5%, a recall of 71.6%, and an F1-score of 72.0%. ConclusionsThis study leveraged non-contrast CMR cine to innovatively develop a radiomics-based prognostic models with relatively good predictive performance. Model^,s predictive efficiency was further enhanced by integrating clinical and cardiac functional imaging features. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.003 ObjectiveTo verify the feasibility of in vivo cardiac diffusion tensor imaging (cDTI) and further explore its potential clinical value in identifying myocardial microstructural remodeling in hypertrophic cardiomyopathy (HCM). Materials and MethodsThe prospective study enrolled 37 patients with HCM and 19 healthy controls. The clinical data, cine imaging, late gadolinium enhancement imaging, T1 mapping imaging and cDTI imaging parameters of the two groups were analyzed. The differences in parameters between the two groups were compared using independent sample t-test, chi-square test and non-parametric test. Correlation analysis was conducted between cDTI parameters and native T1, as well as left ventricular wall thickness (LVWT). Covariance analysis was used to evaluate the cDTI parameters after adjusting for native T1 and LVWT. ResultsLVWT, left ventricular mass index and native T1 of HCM patients were higher than healthy controls and differences were statistically significant. cDTI parameters like mean diffusivity (MD) (P < 0.001) and fractional anisotropy (FA) (P < 0.001) were lower than healthy controls. Secondary eigenvector angle (E2A) was higher than controls (P < 0.001). These differences were statistically significant. Both FA and E2A have high diagnostic efficacy in differentiating HCM patients from healthy controls, with the area under the curve (AUC) being 0.98 (95% confidence intervals: 0.94 to 1.00) and 0.88 (95% confidence intervals: 0.80 to 0.97), respectively. Correlation analysis shows LVWT was moderately negatively correlated with FA (r = -0.754, P < 0.001), and moderately positively correlated with E2A (r = 0.636, P < 0.001). Native T1 was moderately negatively correlated with FA (r = -0.504, P < 0.001), and weakly positively correlated with E2A (r = 0.330, P = 0.013). Analysis of covariance shows that, after adjusting for LVWT and native T1, FA of HCM patients was still significantly decreased. ConclusionsAs a non-invasive quantitative assessment method for myocardial tissue characteristics that does not rely on contrast agents, in vivo cDTI has the ability to identify myocardial disarray in HCM patients. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.004 ObjectiveTo explore the correlation between left ventricular entropy (LV entropy) based on cardiac magnetic resonance (CMR) late gadolinium enhancement (LGE) and myocardial fibrosis imaging markers including Native T1 mapping and extracellular volume fraction (ECV) in patients with heart failure with preserved ejection fraction (HFpEF), as well as its diagnostic value for these patients. Materials and MethodsA total of 150 patients diagnosed with HFpEF at Shandong Provincial Hospital between May 2019 and April 2023 were retrospectively included in the study. They were divided into two groups based on the presence of late gadolinium enhancement regions[the LGE (+) group (110 cases) and the LGE (-) group (40 cases)]. Additionally, 59 healthy individuals were selected as the control group.They were randomly split into the training set and validation set at an 8∶2 ratio. The Spearman correlation coefficient was used to examine the relationship between Native T1 mapping, ECV and LV entropy. The diagnostic efficacy of different parameters was assessed by receiver operating characteristic (ROC) curves. ResultsLV entropy was moderately positively correlated with both Native T1 mapping and ECV (r = 0.48, r = 0.68, respectively; P < 0.001) in all patients. In subgroups of patients with positive LGE, LV entropy was moderately positively correlated with Native T1 mapping, and strongly correlated with ECV (r = 0.57, r = 0.74, respectively; P < 0.001). In subgroups of patients with negative LGE, LV entropy moderately positively correlated with ECV (r = 0.36; P = 0.024), but not significantly correlated with Native T1 mapping. ROC curve analysis showed that in both the training set and the validation set, the diagnostic efficacy of left ventricular entropy was higher than that of Native T1 mapping and ECV. The AUC values were 0.895, 0.732, 0.748 (for the training set) and 0.893, 0.731, 0.747 (for the validation set), respectively. The AUC values of the combined model of the three parameters were 0.916 (training set) and 0.914 (validation set), respectively. The DeLong test showed that the diagnostic performance was improved after the combined of CMR parameters (P < 0.05). ConclusionsLeft ventricular entropy derived from CMR-LGE can characterize extracellular space heterogeneity. It provides supplementary value to traditional imaging markers of myocardial fibrosis in detecting pathological changes in patients with HFpEF, and also exhibits certain diagnostic value for these patients. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.005 ObjectiveTo investigate the correlation between left ventricular fast long-axis strain (FLAS) measured from cardiac magnetic resonance (CMR) cine sequences and extensive late gadolinium enhancement (LGE) in patients with hypertrophic cardiomyopathy (HCM), and to compare its predictive value for extensive myocardial fibrosis against global longitudinal strain (GLS). Materials and MethodsA retrospective analysis was conducted on clinical baseline data and CMR imaging from 131 patients diagnosed with HCM and 75 age- and sex-matched normal controls at the First Hospital of Lanzhou University between January 2017 and January 2021. Patients were divided into extensive LGE and non-extensive LGE groups based on LGE extent. FLAS parameter was measured from the two- and four-chamber left ventricular cine sequences. Pearson correlation analysis was used to explore the relationship between FLAS and LGE extent. Univariate and multivariate logistic regression analyses were employed to assess the association between FLAS and LGE. The diagnostic performance of FLAS for identifying extensive LGE was evaluated using receiver operating characteristic (ROC) curve analysis. ResultsPatients in the extensive LGE group were younger [(52 ± 11) years vs. (57 ± 13) years, P = 0.025)], had higher diastolic blood pressure [(79 ± 10) mmHg vs. (74 ± 12) mmHg, P = 0.016], and showed a statistically significant lower incidence of hypertension (27.3% vs. 51.7%, P = 0.008) compared to the non-extensive LGE group. FLAS was significantly lower in the extensive LGE group [( -9.24% ± 2.73%) vs. (-12.41% ± 2.84%), P < 0.001)] and a moderate negative correlation was observed between LGE extent and FLAS (r = -0.497, P < 0.001). ROC curve analysis showed that the area under the curve (AUC) for FLAS in identifying extensive LGE was 0.802, which was significantly superior to GLS (AUC = 0.709) and traditional CMR parameters (LVMi, LVESVi, LVEF, and MLVT, with AUCs of 0.626, 0.703, 0.725, and 0.702, respectively). Multivariate regression analysis further confirmed FLAS as an independent predictor of extensive LGE (OR = 1.497, 95% CI: 1.550 to 2.663, P < 0.001). ConclusionsAs a novel contrast-agent-free CMR functional parameter, FLAS demonstrates superior performance to GLS and traditional parameters in identifying extensive myocardial fibrosis in HCM patients. It holds promise as a non-invasive tool for myocardial fibrosis assessment, serving as an alternative to LGE, especially in patients with contraindications to gadolinium contrast or renal insufficiency. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.006 ObjectiveFour-dimensional flow cardiac magnetic resonance imaging (4D Flow CMR) was used to analyze the left ventricular blood flow and kinetic energy characteristics in children with premature ventricular complexes (PVCs). This study aimed to evaluate the relationship between left ventricular hemodynamic changes, left ventricular function, and PVCs burden, as well as to explore the predictive value of 4D Flow indices for the degree of premature ventricular complex burden in children with PVCs. Materials and MethodsChildren with PVCs who underwent 4D Flow CMR scanning at Shengjing Hospital Affiliated to China Medical University from January 2024 to January 2025 were enrolled. The case group was divided into two subgroups based on the degree of premature ventricular complex burden detected by 24-hour Holter monitoring: the low-burden group (PVCs-M group, premature ventricular complex burden ≤1‰) and the high-burden group (PVCs-S group, premature ventricular complex burden >1‰). Additionally, children without cardiopulmonary diseases who underwent CMR examination were included as the control group. All children underwent CMR scanning after 24-hour Holter monitoring. The balanced steady-state free precession (bSSFP) sequence was used to acquire cardiac cine sequences, and functional parameters such as left/right ventricular stroke volume index (L/RVSVi) and left ventricular ejection fraction (LVEF) were measured. A three-dimensional retrospective ECG-triggered navigator-gated 4D Flow sequence was employed for whole-heart blood flow scanning, and left ventricular blood flow components and kinetic energy (KE) were determined using post-processing software. Differences in routine CMR indicators and hemodynamic indicators were compared between the case group and the control group, as well as between the subgroups within the case group. Binary logistic regression was used to screen and identify predictive indicators for evaluating the burden degree, and receiver operating characteristic curve analysis was applied to assess the predictive efficacy of various CMR indicators. ResultsA total of 49 subjects were enrolled in this study, including 14 in the control group and 35 in the case group. The case group was further divided into the PVCs-M group (16 subjects) and the PVCs-S group (19 subjects). There were no statistically significant differences in age, gender, or heart rate between the control group and the case group. In terms of conventional CMR indicators, no statistically significant differences were observed between the case group and the control group. However, within the case group, the RVEDVi of the PVCs-S group was lower than that of the PVCs-M group [(77.82 ± 17.73) mL/m² vs. (65.97 ± 13.23) mL/m², P = 0.030]. For 4D Flow indicators, the direct flow in the case group was lower than that in the control group (31.70% ± 11.69% vs. 38.49% ± 6.13%, P = 0.045), while the time deviation was increased [25.80 (0,34.40) ms vs. 0 (0,0) ms, P = 0.001]. Meanwhile, compared with the PVCs-M group, the PVCs-S group showed an increase in residual volume [20.31% (19.08%, 30.20%) vs. 16.65% (13.71%, 23.21%), P = 0.016], and decreases in direct flow and systolic KEi_EDV [28.16% ± 8.60% vs. 35.90% ± 13.66%, P = 0.049; 5.59 (4.47,6.41) μJ/mL vs. 7.76 (7.09,8.33) μJ/mL, P = 0.003]. Binary multivariate Logistic regression analysis was performed on the indicators with P < 0.05 between subgroups in the case group. The results showed that RVEDVi, residual volume, and systolic KEi_EDV were independently associated with the burden of premature ventricular contractions (PVCs) in children. A combined prediction model for PVC burden was constructed using these indicators individually and in combination. The combined prediction model was found to have higher predictive power than individual indicators (AUC = 0.924, P < 0.001). Conclusions4D Flow CMR can measure abnormal changes in left ventricular blood flow components and kinetic energy characteristics in children with PVCs under free breathing, and these changes are closely related to left cardiac function. Among 4D Flow CMR indicators, systolic KEi_EDV, RVEDVi, and residual volume are independently associated with the burden of premature ventricular contractions (PVCs) in children. The prediction model using the three indicators in combination has high predictive value for PVC burden, and is expected to be used for clinical risk stratification and prognosis prediction in children with PVCs. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.007 ObjectiveThis study employed three-dimensional pseudo-continuous arterial spin labeling (3D-pCASL) to evaluate the impact of hypoxic-ischemic brain injury (HIBD) on cerebral blood flow (CBF) in extremely preterm infants, and to explore the clinical value of 3D-pCASL in assessing cerebral perfusion and early prognosis of HIBD in these infants. Materials and MethodsA total of 110 extremely preterm infants clinically diagnosed with HIBD and born at the Third Affiliated Hospital of Zhengzhou University between January 2022 and September 2024 were retrospectively enrolled as the study group. Additionally, 83 extremely preterm infants without HIBD born during the same period were selected as the control group. All infants underwent 3D-pCASL sequences and conventional MRI scans at different corrected gestational ages (CGA). Participants were stratified by CGA at MRI into Subgroup 1 [CGA 32 to 36⁺⁶ weeks: HIBD (n = 58), control (n = 60)] and Subgroup 2 [CGA 37 to 41⁺⁶ weeks: HIBD (n = 52), control (n = 23)]. CBF values were compared between HIBD and control infants within each subgroup, and across subgroups at different CGAs; specifically, correlations between CBF values in differential brain regions at CGA 32 to 36⁺⁶ weeks and Apgar scores at 1 min and 5 min after birth, as well as Neonatal Behavioral Neurological Assessment (NBNA) scores at 40 weeks CGA were analyzed. Results(1) At CGA 32 to 36⁺⁶ weeks, CBF values in the HIBD group were significantly higher than controls in bilateral temporal lobes, parietal lobes, occipital lobes, basal ganglia regions, thalami, and the right central sulcus cortex (P < 0.05); however, no statistically significant differences in regional CBF values were observed between the HIBD and control groups at CGA 37 to 41⁺⁶ weeks. (2) Within the HIBD group, CBF values in bilateral central sulcus cortices were significantly higher at CGA 37 to 41⁺⁶ weeks compared to CGA 32 to 36⁺⁶ weeks (P < 0.05), with no significant differences in other regions of interest (P > 0.05). In controls, CBF values in bilateral temporal lobes, occipital lobes, basal ganglia regions, and central sulcus cortices were significantly elevated at CGA 37 to 41⁺⁶ weeks versus CGA 32 to 36⁺⁶ weeks (P < 0.05), while other regions showed no significant changes (P > 0.05). (3) The CBF values in various regions of interest showed no correlation with 1-min and 5-min Apgar scores at birth. Additionally, bilateral thalamic and left basal ganglia region CBF at CGA 32 to 36⁺⁶ weeks negatively correlated with NBNA scores at 40 weeks of CGA (r = -0.284, -0.292, -0.272; P < 0.05). ConclusionsThe occurrence of HIBD may affect early cerebral perfusion in extremely preterm infants. Altered perfusion in specific brain regions could influence early prognosis, and 3D-pCASL holds potential value in assessing cerebral perfusion changes and early prognosis in these infants following HIBD. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.008 ObjectiveTo systematically explore the structural remodeling characteristics of the brain in patients with self-limited epilepsy with centrotemporal spikes (SeLECTS) based on individual cortical thickness and surface area structural covariance networks. Materials and MethodsA prospective study was conducted from October 2021 to October 2023, collecting data from 46 children diagnosed with SeLECTS based on clinical presentation and electroencephalogram at the Third Affiliated Hospital of Zhengzhou University, along with 46 healthy controls matched for age and gender. All participants underwent 3D-T1-weighted structural MRI scans, and morphological measures (cortical thickness and brain surface area) were obtained through data preprocessing using Freesurfer software. Individual structural covariance network differences were analyzed using the network template perturbation method. ResultsSeLECTS patients showed significant covariance enhancement in both cortical thickness and surface area structural covariance networks (P_corrected < 0.05, FWE correction). The thickness covariance network was characterized by widespread interhemispheric synchrony enhancement (P_corrected < 0.001) and local modular enhancement in the parietal-central region (P_corrected = 0.005). The surface area covariance network highlighted coordinated enhancement along the limbic-motor-sensory integration pathway (P_corrected = 0.002) and short-range covariance enhancement in the cingulate and the pars opercularis (P_corrected = 0.033). ConclusionsSeLECTS patients exhibited covariance enhancement across multiple brain regions in both cortical thickness and surface area-based structural covariance networks. These findings provide new neuroimaging evidence for understanding the network-level pathological mechanisms of SeLECTS and reveal potential intervention targets, laying the foundation for the development of personalized precision treatment strategies. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.009 ObjectiveTo explore cerebral blood flow (CBF) and CBF connectivity characteristics between healthy control (HC) and major depressive disorder (MDD) in adolescents accompanied by non-suicidal self-injury (NSSI). Materials and MethodsWe enrolled 35 adolescent patients with MDD accompanied by NSSI, along with 30 gender-and-age-matched healthy controls (HC). All participants underwent comprehensive emotional and cognitive assessments, including: Self-Rating Depression Scale (SDS), Self-Rating Anxiety Scale (SAS), Snaith-Hamilton Pleasure Scale (SHAPS), 24-item Hamilton Depression Scale (HAMD-24), Hamilton Anxiety Scale (HAMA), Montreal Cognitive Assessment (MoCA), Mini-Mental State Examination (MMSE), Ottawa Self-Injury Inventory. All subjects underwent 3.0 Tesla MRI scanning, including 3D T1-weighted imaging (T1WI) and arterial spin labeling (ASL) sequences. We compared cerebral blood flow (CBF) differences and CBF connectivity patterns between the two groups, and analyzed correlations between CBF alterations in specific brain regions and clinical symptom severity. ResultsCompared with HCs, the MDD with NSSI group exhibited significantly higher anxiety and depression scores, along with reduced cognitive performance (P < 0.001). In the MDD with NSSI group, cerebral blood flow (CBF) values showed negative correlations with depression, anxiety, and anhedonia scores, but positive correlation with MMSE scores. Relative to HCs, MDD with NSSI patients demonstrated decreased CBF in the right putamen and right insula (P < 0.05, FWE-corrected, t = 4.9). Using the right putamen (the primary region showing CBF differences) as a seed region for whole-brain CBF connectivity analysis in the MDD with NSSI group revealed distinct connectivity patterns. Specifically, we found positive CBF connectivity between the right putamen and right insula, right frontal operculumand right inferior frontal gyrus (pars opercularis) (P < 0.05, FWE-corrected, t = 12.4). No brain regions showed negative CBF connectivity with the right putamen in MDD with NSSI patients. ConclusionsThis study provides novel insights to explore the underlying neural mechanisms of MDD with NSSI behavior, demonstrating that CBF analysis serves as an effective approach for investigating MDD with NSSI, which offers new neuroimaging evidence of prevention strategies and therapeutic interventions for self-injurious behaviors. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.010 ObjectiveTo investigate alterations in voxel-wise degree centrality (DC) and functional connectivity (FC) and their relationship with disease severity in Parkinson^,s disease (PD) patients using resting-state functional magnetic resonance imaging (rs-fMRI). Materials and MethodsForty-six PD patients admitted to the Department of Neurology, Affiliated Hospital 6 of Nantong University and 24 health controls (HC) between January 2024 and April 2025 were enrolled. Twenty-four age-, sex-, and education-matched HC were recruited concurrently. Data of rs-fMRI data were acquired from both groups. Whole-brain voxel-wise DC was calculated and compared between groups. Brain regions showing significant DC differences were used as seeds for whole-brain FC analysis, which was also compared between groups. Identify varying brain DC and FC values in PD patients, investigate their association with disease duration, Hoehn-Yahr (H-Y) staging, Unified Parkinson^,s Disease Rating Scale (UPDRS) Ⅰ/Ⅱ/Ⅲ/Ⅳ scores, and cognitive and psychiatric assessments. ResultsPD patients had significantly lower Montreal cognitive assessment scores and higher Hamilton anxiety rating scale and Hamilton depression rating scale scores compared to the HC group (t = -5.559, 8.218, 7.249, P < 0.001). Compared to HCs, PD patients exhibited significantly lower DC values in the right lenticular nucleus and pallidum, right superior temporal gyrus, and right precentral gyrus, and higher DC values in the left gyrus rectus, left angular gyrus, and left medial superior frontal gyrus (t = -5.677, -3.833, -4.752, 5.827, 4.218, 5.063, P < 0.001). Significant FC reductions were found in PD patients for the following connections: right lenticular nucleus and pallidum to left insula, right putamen, left inferior parietal but supramarginal and angular gyri; right superior temporal gyrus to left precuneus; right precentral gyrus to right calcarine fissure and surrounding cortex; left gyrus rectus to right fusiform gyrus and right superior temporal gyrus; and left medial superior frontal gyrus to bilateral lenticular nucleus and putamen (t = -4.884, -4.341, -3.961, -4.945, -4.809, -4.518, -4.541, -5.004, -4.535, P < 0.001). In PD patients, the DC value in the right precentral gyrus, and the FC values for right lenticular nucleus and pallidum - left insula, right superior temporal gyrus - left precuneus, and left medial superior frontal gyrus - left lenticular nucleus and putamen, were negatively correlated with UPDRS Ⅱ scores (r = -0.332, -0.342, -0.319, -0.406, P < 0.05). The DC value in the left medial superior frontal gyrus was positively correlated with H-Y stage and UPDRS Ⅱ/Ⅲ scores (r = 0.371, 0.300, 0.454, P < 0.05). The FC value of right lenticular nucleus and pallidum - right lenticular nucleus and putamen was negatively correlated with disease duration (r = -0.299, P = 0.044). ConclusionsIn PD patients, there are complex changes in brain DC and FC at rest, with some abnormalities correlating with disease severity. This provides a basis for finding neuroimaging biomarkers to objectively evaluate the condition of PD patients. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.011 ObjectiveTo investigate the development of bilateral hippocampi in fetuses with intrauterine growth restriction (IUGR) using MRI in the second and third trimester of pregnancy. Materials and MethodsA retrospective analysis was conducted on the prenatal MR images of 51 fetuses with growth restriction and 50 normal fetuses who visited the prenatal diagnosis department of our hospital from January 2023 to Februry 2025. Linear regression was used to evaluate the relationship between gestational age and the long diameter of the fetal sagittal hippocampus and the hippocampal infolding angle (HIA). The t-test was used to compare the long diameter of the sagittal hippocampus and HIA between the two groups. ResultsGestational age was significantly positively correlated with the bilateral sagittal hippocampal long diameter and HIA in the normal group and IUGR group in (the normal group, the right sagittal hippocampal long diameter r = 0.936, P < 0.001; the left sagittal hippocampal long diameter r = 0.901, P < 0.001; the right HIA r = 0.867, P < 0.001; the left HIA r = 0.856, P < 0.001. In the IUGR group, the right sagittal hippocampal long diameter r = 0.807, P < 0.001; the left sagittal hippocampal long diameter r = 0.778, P < 0.001; the right HIA r = 0.786, P < 0.001; the left HIA r = 0.763, P < 0.001). In the normal group, the regression equations between hippocampal longitudinal diameter (Y) and gestational age (X) were: left side, Y = 1.11 + 0.64X (β = 0.901, SE = 0.045, t = 14.388, R² = 0.812, P < 0.001); right side, Y = 1.96 + 0.69X (β = 0.936, SE = 0.038, t = 18.399, R² = 0.876, P < 0.001). The regression equations between HIA (Y) and gestational age (X) were: left side, Y = 53.13 + 0.54X (β = 0.856, SE = 0.047, t = 11.447, R² = 0.812, P < 0.001); right side, Y = 52.57 + 0.59X (β = 0.867, SE = 0.049, t = 12.074, R² = 0.752, P < 0.001). In the IUGR group, the regression equations between hippocampal longitudinal diameter (Y) and gestational age (X) were: left side, Y = 4.72 + 0.39X (β = 0.778, SE = 0.045, t = 8.660, R² = 0.605, P < 0.001); right side, Y = 6.14 + 0.37X (β = 0.807, SE = 0.038, t = 9.579, R² = 0.652, P < 0.001). The regression equations between HIA (Y) and gestational age (X) were: left side, Y = 55.88 + 0.40X (β = 0.763, SE = 0.049, t = 8.273, R² = 0.605, P < 0.001); right side, Y = 56.71 + 0.41X (β = 0.786, SE = 0.046, t = 8.911, R² = 0.618, P < 0.001). The mean hippocampal longitudinal diameters in the normal group were (19.698 ± 2.075) mm (right) and (19.006 ± 2.002) mm (left), while those in the IUGR group were (17.941 ± 1.284) mm (right) and (17.186 ± 1.408) mm (left). The differences between groups were statistically significant (right: t = 5.104, P < 0.001; left: t = 5.273, P < 0.001). The mean HIA values in the normal group were (71.018 ± 1.907)° (right) and (70.008 ± 1.769)° (left), while in the IUGR group they were (69.958 ± 1.480)° (right) and (68.911 ± 1.499)° (left). These intergroup differences were also statistically significant (right: t = 3.113, P = 0.002; left: t = 3.362, P = 0.001). ConclusionsThe size of the hippocampus and HIA in IUGR fetuses change with gestational age, there are differences in the hippocampal development of IUGR fetuses and that of normal fetuses. MRI can provide imaging diagnostic support for hippocampal development abnormalities in IUGR. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.012 ObjectiveTo study the degree of daytime sleepiness, insomnia, and the main causes of insomnia in pilots using sleep-related scales, and to establish an imaging prediction model for the degree of daytime sleepiness in pilots using resting-state functional magnetic resonance imaging (fMRI) and connectome-based predictive modeling (CPM). Materials and MethodsFrom May 2023 to October 2024, 96 pilots were recruited from Lintong Sanatorium. The Epworth Sleeping Scale (ESS) was used to assess the daytime mental status of pilots, and pilots were divided into a sleepiness group and a normal group based on their scores. The Insomnia Severity Index (ISI) was used to assess the severity of insomnia, the Pittsburgh Sleep Quality Index (PSQI) was used to assess the quality of sleep, the Self-rating Anxiety Scale (SAS) was used to assess the level of anxiety, and the Self-rating Depression Scale (SDS) was used to assess the level of depression. The fMRI data of pilots was collected and functional connectivity matrices were constructed. The CPM was used to construct a prediction model for the degree of daytime sleepiness in pilots. ResultsThe proportion of pilots experiencing daytime sleepiness was approximately 15.6%. The ISI score of pilots in the sleepiness group was higher than that of pilots in the normal group (FDR-corrected P = 0.042), while the PSQI score of sleep quality was lower than that of pilots in the normal group (FDR-corrected P = 0.047). There was no significant difference in anxiety and depression levels between the two groups. There was a weak correlation between the ISI score and anxiety level (r = 0.236, P = 0.020), as well as the depression level (r = 0.212, P = 0.040). There was also a weak correlation between the flight hours of pilots and the sleep disorder (r = 0.216, P = 0.035), as well as the total PSQI score (r = 0.202, P = 0.048). There was a correlation between the predicted ESS score and its true value (r = 0.296), and permutation testing indicated that this correlation was significant (P = 0.004). The functional connectivity that contributed most to the predictive efficacy of the model primarily existed between the limbic system and the default mode network, ventral attention network, and frontal and parietal networks. ConclusionsThe proportion of pilots experiencing daytime sleepiness is approximately 15.6%, and it is associated with insomnia and poor sleep quality. CPM effectively predicted the degree of daytime sleepiness in pilots, and the functional connectivity that significantly contributed to this prediction were between the limbic system and the default mode network, ventral attention network, and frontal and parietal networks. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.013 ObjectiveTo investigate glymphatic pathway impairment following traumatic brain injury (TBI) and its association with consciousness states and clinical outcomes in patients with post-traumatic disorders of consciousness (DoC). Materials and MethodsThis study recruited 22 patients with post-traumatic DoC and 30 healthy controls for magnetic resonance imaging scans. Glymphatic function was evaluated via choroid plexus volume (CPV) and the diffusion tensor imaging along the perivascular space (DTI-ALPS) index. Consciousness levels and functional recovery were assessed using the Coma Recovery Scale-Revised (CRS-R) and Glasgow Outcome Scale-Extended (GOSE). Intergroup differences in DTI-ALPS and CPV were compared. Pearson correlation analysis examined relationships among DTI-ALPS, CPV, CRS-R, and GOSE scores in the DoC group. ResultsThe DoC group exhibited significantly lower DTI-ALPS indices and larger choroid plexus volume/total intracranial volume (CPV/TIV) compared to HC (P < 0.05). In DoC patients, DTI-ALPS positively correlated with CRS-R scores (r = 0.43, P = 0.04) but negatively with CPV (r = -0.46, P = 0.04). CPV/TIV showed negative correlation with CRS-R (r = -0.64, P = 0.01). At 3-month MRI follow-up, GOSE scores positively correlated with DTI-ALPS (r = 0.45, P = 0.04) and negatively with CPV/TIV (r = -0.59, P = 0.01). ConclusionsTBI induces glymphatic pathway dysfunction, which influences consciousness states and clinical outcomes. These findings may reveal potential neural mechanisms of post-TBI DoC, clarify the relationship between glymphatic impairment and residual consciousness, and offer novel neuroimaging biomarkers for clinical diagnosis and prognosis in post-traumatic DoC. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.014 ObjectiveTo investigate the clinical value of conventional MRI features and T1WI enhancement histogram analysis in the preoperative non-invasive differentiation of isocitrate dehydrogenase (IDH) wild-type glioblastoma (IDH-wt GBM) and IDH mutant grade 4 astrocytoma (IDHmut-Astro-4). Materials and MethodsA retrospective analysis was conducted on clinical, imaging, and pathological data from IDH-wt GBM (n = 44) and IDHmut-Astro-4 (n = 40) cases confirmed by histopathological diagnosis. Using FireVoxel software, the entire tumor was contoured layer by layer on axial T1WI enhanced images to obtain histogram parameters of the tumor region. Categorical variables were analyzed using chi-square tests or Fisher^,s exact tests, while continuous variables were analyzed using independent samples t-tests or Mann-Whitney U tests. Diagnostic performance between the two groups was assessed using receiver operating characteristic (ROC) curves. ResultsThe IDH-wt GBM group, age (P < 0.001), tumor necrosis (P < 0.001), tumor enhancement degree (P = 0.021), and maximum diameter of peritumoral edema (P < 0.001) were all greater than those in the IDHmut-Astro-4 group. In the IDH-wt GBM group, the variance coefficient (P = 0.009), skewness (P = 0.002), kurtosis (P < 0.001), and entropy (P < 0.001) of the T1WI enhancement histogram parameters were all greater than those in the IDHmut-Astro-4 group, with statistically significant differences. ROC curve analysis showed that age + fusion of conventional MRI features + fusion of histogram parameters had the best diagnostic performance for distinguishing between the two groups, with an area under the curve, sensitivity, specificity, accuracy, positive predictive value, and negative predictive value of 0.962 (0.896 to 0.992), 86.36%, 92.50%, 89.26%, 92.68%, and 86.05%, respectively. ConclusionsConventional MRI features and T1WI-enhanced histogram analysis aid in the preoperative non-invasive differentiation of IDH-wt GBM and IDHmut-Astro-4, with the highest diagnostic efficacy achieved by the combination of age + fused conventional MRI features + fused histogram parameters. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.015 ObjectiveTo analyze preoperative visually accessible rembrandt images (VASARI) features in patients with intracranial extraventricular ependymoma (IEE) and evaluate their predictive value for world health organization (WHO) grading and Ki-67 proliferation index. Materials and MethodsClinical and preoperative cranial MRI data of 30 pathologically confirmed IEE patients (18 WHO grade 2, 12 grade 3) who underwent surgical resection at Second Hospital of Lanzhou University (January 2012 to September 2024) were retrospectively analyzed. Two experienced neuroradiologists independently evaluated MRI characteristics according to VASARI criteria. SPSS 27.0 was used to analyze correlations between VASARI features and WHO grade/Ki-67 index. Diagnostic efficacy was assessed using receiver operating characteristic (ROC) curves. ResultsThe VASARI total score was significantly higher in WHO grade 3 group (92.00 ± 18.75) versus grade 2 (76.22 ± 18.89, P < 0.05). ROC analysis showed AUC of 0.736 (95% CI: 0.541 to 0.931) for differentiating grade 3 from grade 2 tumors. At optimal cut-off ≥ 59.5, sensitivity was 94.1% and specificity 30.8%. Significant intergroup differences (P < 0.05) existed in cystic change rate (F8), enhancement rim thickness (F11), and peritumoral edema percentage (F14), with grade 3 tumors exhibiting higher cystic rates, thicker enhancement rims, and more extensive edema. VASARI total score positively correlated with WHO grade (r = 0.391, P = 0.032) and Ki-67 index (r = 0.370, P = 0.044). For predicting high Ki-67 expression, AUC was 0.633 (95% CI: 0.421 to 0.845) with 69.2% sensitivity and 82.4% specificity at cut-off ≥ 76.5. ConclusionsThe VASARI MRI features (F8, F11, F14, and total score) have certain value in non-invasively distinguishing WHO grade 2 from grade 3 IEE preoperatively and in predicting the Ki-67 proliferation index. They can serve as an auxiliary assessment tool to provide reference for clinical diagnosis and treatment. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.016 ObjectiveTo explore the value of multiphase real-time free-breathing cine imaging in cardiac function assessment compared with standard breath-hold cardiac cine MRI, thereby expanding the clinical applicability of magnetic resonance imaging (MRI) cardiac function evaluation. Materials and MethodsCardiac cine images were acquired from 42 healthy volunteers using a 3.0 T MRI system with both a multiphase real-time free-breathing cine sequence and a standard breath-hold cine sequence. A self-developed algorithm based on the signal intensity ratio between myocardium and blood pool was applied to rearrange the multiphase myocardial images into diastolic and systolic phases. Left ventricular functional parameters, including end-diastolic volume (EDV), end-systolic volume (ESV), ejection fraction (EF), stroke volume (SV), and myocardial mass (MASS), were quantified. Consistency and differences between the two methods were statistically compared. ResultsReal-time free-breathing cine images exhibited lower signal contrast between myocardium and blood pool compared to standard breath-hold imaging, yet remained sufficient for ventricular functional analysis. Multiphase cine data (120 phases) were successfully sorted, classified, and processed automatically to calculate cardiac functional parameters by tracking signal variations throughout the cardiac cycle. There was no statistically significant difference between the two methods in terms of cardiac function indicators such as EDV, ESV, EF, and MASS (P > 0.05), while the difference in SV was statistically significant (P < 0.05). All parameters demonstrated high consistency via correlation analysis with correlation coefficients 0.968 (EDV), 0.927 (ESV), 0.954 (SV), 0.942 (EF), and 0.953 (MASS) respectively (P < 0.001). ConclusionsReal-time cine imaging under free-breathing conditions, performed without the need for breath-holding or electrocardiographic gating, significantly improves the applicability of cardiac magnetic resonance function assessment in patients unable to comply with breathing instructions. The newly developed data processing method ensures accurate evaluation of left ventricular function, offering a reliable alternative for clinical scenarios such as heart failure or cases where breath-hold compliance is unfeasible. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.017 ObjectiveTo explore the value of a SHapley Additive exPlanations (SHAP) machine learning model based on dynamic contrast-enhanced (DCE) and high-resolution delayed phase images for the preoperative prediction of luminal and non-luminal subtypes of invasive breast cancer. Materials and MethodsClinical, pathological, and imaging data of 182 patients with pathologically confirmed invasive breast carcinoma of no special type were retrospectively collected and divided into a luminal group (121 cases) and a non-luminal group (61 cases) based on pathological results. Using 3D Slicer software, lesion margins were delineated on DCE and high-resolution delayed phase breast MRI images of invasive breast cancer patients, and radiomic features were extracted. Patients were randomly split into training and test sets in a 7∶3 ratio. Univariate t-test or Mann-Whitney U test and least absolute shrinkage and selection operator (LASSO) were used for feature selection. Clinical models, radiomics models, and combined models were built using logistic regression, support vector machine (SVM), and AdaBoost algorithms, respectively. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity. Model performance comparisons were conducted using DeLong^,s test. SHAP analysis was used to visualize feature contributions in the models. ResultsThere were statistically significant differences in histological grade and carbohydrate antigen-125 between the two groups, with P < 0.05. After dimensionality reduction, 2 and 4 optimal radiomics features were respectively retained for DCE and high-resolution delayed-phase images. The combined models of logistic, SVM, and AdaBoost based on DCE features, high-resolution delayed-phase features, and clinical features had better performance. The AUCs in the training set were 0.854, 0.853, and 0.962, respectively, with accuracies of 71.8%, 75.1%, and 89.4%, sensitivities of 74.0%, 77.3%, and 85.1%, and specificities of 69.7%, 72.9%, and 93.6%, respectively. The AUCs in the test set were 0.828, 0.836, and 0.802, respectively, with accuracies of 72.5%, 76.3%, and 72.5%, sensitivities of 74.1%, 77.0%, and 71.8%, and specificities of 67.5%, 74.5%, and 73.5%, respectively. The combined models of logistic and AdaBoost had statistically significant differences between the training set and the test set (P = 0.044, P < 0.001). The combined model of SVM had no statistically significant difference between the training set and the test set (P = 0.277). In the test set, the combined model of SVM was superior to the clinical model of SVM, and the difference was statistically significant (P < 0.001). ConclusionsInterpretable machine learning models can preoperatively predict luminal and non-luminal subtypes of invasive breast cancer, holding significant clinical value for formulating personalized treatment plans and prognostic assessments for patients. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.018 ObjectiveTo investigate the efficacy of Delta radiomics model based on gadoxetic acid disodium (Gd-EOB-DTPA) enhanced MRI for preoperative prediction of microvascular invasion (MVI) in hepatocellular carcinoma (HCC). Materials and MethodsA total of 189 pathologically confirmed HCC patients were retrospectively enrolled (91 MVI-positive, 98 MVI-negative). Regions of interest (ROI) of the tumor were delineated on preoperative axial non-contrast T1-weighted imaging (T1WI) and Gd-EOB-DTPA-enhanced hepatobiliary phase (HBP) images. Radiomics feature extraction was performed to calculate Delta radiomics feature values. Feature selection was conducted through paired t-tests, Spearman correlation analysis, and least absolute shrinkage and selection operator (LASSO) algorithm. Logistic regression (LR) classifier was used for constructing three models (T1WI, HBP, and Delta), with receiver operating characteristic (ROC) curves generated to evaluate predictive performance. ResultsThe Delta radiomics model based on LR algorithm demonstrated optimal performance, achieving area under the curve (AUC) of 0.888 (95% CI: 0.834 to 0.942) (training set) and 0.800 (95% CI: 0.687 to 0.913) (validation set). The combined model integrating Delta radiomics features with clinical baseline data showed superior predictive efficacy, with AUC of 0.898 (95% CI: 0.846 to 0.950) (training set) and 0.811 (95% CI: 0.702 to 0.921) (validation set). ConclusionsThe Gd-EOB-DTPA-enhanced MRI-based Delta radiomics model shows potential clinical value in preoperative MVI prediction for HCC patients. The combined model incorporating both Delta radiomics and clinical baseline parameters exhibits enhanced predictive performance. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.019 ObjectiveThis study aimed to evaluate an explainable machine learning model for predicting lymphovascular invasion (LVI) in rectal cancer. The model was built using diffusion kurtosis imaging (DKI), intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI), and clinical data. Materials and MethodsThis retrospective study included 91 patients with pathologically confirmed rectal cancer. Patients were stratified into LVI-positive (+) and LVI-negative (-) groups according to histopathological findings. All patients underwent magnetic resonance imaging (MRI) examinations, including diffusion kurtosis imaging (DKI) and intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI). Six quantitative parameters were extracted from the scans: mean kurtosis (MK), mean diffusivity (MD), true diffusion coefficient (D), pseudo-diffusion coefficient (D^*), perfusion fraction (f), and apparent diffusion coefficient (ADC). In addition, clinical variables such as serum tumor marker levels and pathological lymph node status were recorded. Finally, univariate followed by multivariate logistic regression analyses were performed to identify independent predictors of LVI. A combined predictive model was developed using logistic regression (LR). Model performance was evaluated using receiver operating characteristic (ROC) curves and the area under the curve (AUC), with five-fold cross-validation for internal validation. Feature contributions were visualized using SHapley Additive exPlanations (SHAP), and an online risk calculator was developed for individualized prediction. ResultsUnivariate and multivariate regression analyses identified MK, D, f, and carcinoembryonic antigen (CEA) as independent predictors of LVI in rectal cancer. The LR-based combined model achieved an AUC of 0.887, with a mean AUC of 0.893 in five-fold cross-validation. SHAP analysis clearly illustrated the contribution of each feature to the prediction. The web-based risk calculator enabled real-time visualization of individualized risk estimates. ConclusionsThe explainable machine learning model based on DKI and IVIM-DWI quantitative parameters combined with clinical features effectively predicts LVI status in rectal cancer. This model not only demonstrates excellent predictive performance but also enhances clinical applicability and generalizability through transparent feature contribution and individualized risk assessment. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.020 ObjectiveTo evaluate the efficacy of rectal cancer cell microstructural parameters fitted by time-dependent diffusion magnetic resonance imaging (td-dMRI) in preoperative prediction of lymph node metastasis (LNM) and tumor deposit (TD). Materials and MethodsA retrospective analysis was conducted on the imaging and clinical data of 88 patients with rectal cancer who underwent surgery in our hospital from December 2023 to March 2025. All patients received preoperative td-dMRI examinations. Using the IMPULSED model, cellular microstructural parameters and apparent diffusion coefficient (ADC) were extracted, including cell diameter (d), intracellular volume fraction (V_in), extracellular diffusion coefficient (D_ex), cellularity, ADC_{OGSE25Hz/PGSE} value, and ADC_{OGSE40Hz/PGSE} value. The correlation between td-dMRI parameters and pathological results was verified. Based on postoperative pathological data, patients were divided into LNM-negative group (Neg-LNM, n = 40) and LNM-positive group (Pos-LNM, n = 48), as well as TD-negative group (Neg-TD, n = 70) and TD-positive group (Pos-TD, n = 18). Differences in microstructural parameters and ADC ratios between the groups were compared. Receiver operating characteristic (ROC) curves and logistic regression analysis were used to assess the diagnostic efficacy of single and combined models in predicting TD and LNM. ResultsThere was a good correlation between td-dMRI parameters and pathological measurements (n = 16; all r > 0.70; all P < 0.05). In the Pos-LNM group, the d value of cancer cells and ADC_{OGSE25Hz/PGSE} values were significantly higher than those in the Neg-LNM group (both P < 0.05), while cellularity was lower than that in the Neg-LNM group (P < 0.05). In the Pos-TD group, the d value and ADC_{OGSE25Hz/PGSE} value were significantly higher than those in the Neg-TD group (both P < 0.05), and cellularity was lower than that in the Neg-TD group (P < 0.05). Univariate logistic regression analysis showed that d value and ADC_{OGSE25Hz/PGSE} value could predict LNM and TD (all P < 0.05). Multivariate logistic regression analysis combining d and ADC_{OGSE25Hz/PGSE} indicated that the combined model could predict LNM and TD (P < 0.05). The AUC values of the d value, ADC_{OGSE25Hz/PGSE} value, and the combined indicator for predicting LNM and TD were all greater than 0.70. ConclusionsThe microstructural parameter d and ADC ratio based on td-dMRI have favorable clinical application potential in predicting LNM and TD. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.021 ObjectiveTo explore the feasibility of habitat imaging (HI) for non-invasive quantitative visualization of zonal heterogeneity and risk prediction in prostate cancer (PCa). Materials and MethodsThis retrospective study involved 147 patients who underwent multi-parametric magnetic resonance imaging (mpMRI) and confirmed PCa by radical prostatectomy (RP) at Xijing Hospital from January 2018 to August 2024. Patients were divided into training and test sets in a 7∶3 ratio. According to RP results, PCa was categorized into transition zone (TZ) and peripheral zone (PZ). The apparent diffusion coefficient (ADC), perfusion fraction (f) and mean kurtosis (MK) values of each voxel were integrated to delineated habitat subregions and generate habitat maps. The differences between PZ and TZ PCa were compared from multiple perspectives including clinical, pathological and imaging. According to the 2019 International Society of Urological Pathology (ISUP) guidelines, the habitat maps were matched with RP specimens to assess the ISUP grade of each subregion, and the patients were classified into low-risk (ISUP ≤ 2) and high-risk (ISUP ≥ 3) groups. Logistic regression analysis was applied to identify factors associated with high-risk PCa and to construct a predictive model called zone-based habitat imaging (zHI)-clinial imaging. Then the model^,s efficacy was evaluated. ResultsHabitat 1 had lower ADC, f and higher MK values compared to habitats 2 and 3. Compared with TZ, PZ PCa exhibited worse clinical and pathological features, with a higher proportion of habitat 1. Logistic regression analysis indicated that anatomical zone (OR = 3.50, 95% CI: 1.01 to 12.09) and the proportion of Habitat 1 (OR = 3.63, 95% CI: 1.37 to 9.62) were independent risk factors for high-risk PCa (P < 0.05). The area under the curve (AUC) of the zHI-clinical imaging model for risk assessment in the training and test sets were 0.889 (95% CI: 0.822 to 0.955) and 0.883 (95% CI: 0.740 to 0.925), respectively. ConclusionsThis study comprehensively verified the zonal heterogeneity of PCa and constructed a model based on anatomical zone and HI features, which demonstrated enhanced efficacy in non-invasive quantitative visualization and prediction of PCa risk. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.022 ObjectiveTo explore the value and risk stratification of based on biparametric magnetic resonance imaging (bp-MRI) of prostate imaging report and data system version 2.1 (PI-RADS v2.1) combined with prostate specific antigen density (PSAD) in the differential diagnosis of clinically significant prostate cancer (csPCa) with tPSA 4-20 ng/mL. Materials and MethodsRetrospectively analyzed the data of 304 patients undergoing bp-MRI examination with pathological results between October 2017 and June 2023 in the General Hospital of Ningxia Medical University. The patients were divided into csPCa group (Gleason ≥ 7, n = 66) and non-csPCa (Gleason < 7 and benign diseases, n = 238) according to the pathological results. The independent risk factors were screened by univariate and multivariate logistic regression analysis, then the clinical model was constructed, and the clinical net benefit was analyzed by decision curve (DCA). Diagnostic performance was evaluated by using the area under the receiver operating characteristic (ROC) curve, and the independent risk factors were graded and combined. ResultsThe diagnostic efficacy of clinical model (PI-RADS v2.1 + PSAD) is the best (AUC = 0.901, 95% CI: 0.858 to 0.944). Classify and combine PI-RADS v2.1 and PSAD grades, when PI-RADS v2.1 ≤ 2 and PSAD ≤ 0.15 ng/mL², the csPCa positive rate is 0%; when PI-RADS v2.1 = 3 and PSAD < 0.30 ng/mL², the csPCa positive rate is less than 15%; when PI-RADS v2.1 is 4 to 5 and PSAD is 0.15 to 0.29 ng/mL², the csPCa positive rate is 46.5%; when PI-RADS v2.1 is 4 to 5 and PSAD ≥ 0.30 ng/mL², the csPCa positive rate is as high as 81.3%. ConclusionsThe patients with PI-RADS v2.1 ≤ 2 or PI-RADS v2.1 = 3 and PSAD < 0.30 ng/mL² can avoid unnecessary biopsies. PI-RADS v2.1 combined with PSAD can significantly improve the diagnostic efficiency of tPSA 4-20 ng/mL csPCa. The combination of PI-RADS v2.1 and PSAD is helpful for risk assessment of patients with csPCa before puncture, so as to can reduce unnecessary puncture of some patients and provide certain decision-making guidance for clinic. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.023 ObjectiveThe primary aim of this study was to construct a predictive model utilizing Delta radiomics features derived from magnetic resonance imaging (MRI) scans taken before and after neoadjuvant chemotherapy (NACT), in order to stratify the risk of postoperative recurrence based on pathological risk factors in patients diagnosed with locally advanced cervical cancer (LACC). Materials and MethodsThis retrospective study enrolled 221 cervical cancer patients who underwent surgery after NACT. Based on the presence of intermediate- and high-risk pathological factors, patients were classified into a low-risk group (n = 128) and an intermediate-to-high-risk group (n = 93). Delta radiomics features were extracted and calculated from sagittal T2-weighted imaging (Sag_T2WI) and axial contrast-enhanced T1-weighted imaging (Ax_T1CE) acquired before and after NACT, followed by feature dimensionality reduction and selection. Clinical features were screened using univariate analysis. Using the random forest algorithm, we constructed separate models: a Delta-T1CE radiomics model, a Delta-T2WI radiomics model, and a clinical model. Subsequently, a dual-sequence fusion model and a clinical-dual-sequence fusion model were built. Model performance was evaluated using receiver operating characteristic (ROC) curves, and the clinical utility of the models was assessed via decision curve analysis (DCA), followed by performance evaluation and comparison. ResultsThe areas under the curve (AUC) for the Delta-T1CE model, Delta-T2WI model, and clinical model in the test set were 0.844 [95% confidence interval (CI): 0.739 to 0.926], 0.938 (95% CI: 0.880 to 0.981), and 0.675 (95% CI: 0.543 to 0.800), respectively. Among the non-fusion models, the Delta-T2WI model demonstrated the best predictive performance. The dual-sequence fusion model and the clinical-dual-sequence fusion model showed no significant difference in performance compared to the Delta-T2WI model, with test set AUCs of 0.945 (95% CI: 0.888 to 0.986) and 0.944 (95% CI: 0.890 to 0.985), respectively. Decision curve analysis revealed that the Delta-T2WI model and the two fusion models provided higher clinical net benefit than the Delta-T1CE radiomics model and the clinical model. ConclusionsDelta radiomics models, particularly those based on T2WI, can effectively predict recurrence risk stratification in cervical cancer patients after NACT, offering significant reference value for formulating treatment strategies for patients stratified into the intermediate-to-high-risk group. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.024 ObjectiveThis study integrates MRI radiomics with deep learning techniques to construct a classification model for accurate prediction of osteoarthritis (OA) and rheumatoid arthritis (RA) with synovitis. Materials and MethodsThrough a retrospective analysis of knee MRI data from patients diagnosed with OA or RA between January 2018 and December 2024, eligible scans were selected based on inclusion/exclusion criteria to establish the dataset. To address the challenge of small-sample learning in medical imaging, an improved ResNet3D-18 model based on model-agnostic meta-learning (MAML) was developed, enabling rapid task adaptation in small-sample scenarios to enhance classification performance. Three-dimensional gradient-weighted class activation mapping (Grad-CAM) was employed to interpret prediction results. After anonymizing clinical information, two radiologists independently annotated the MRI scans, with discrepancies resolved by consensus. Model performance was assessed using five-fold cross-validation. ResultsThe dataset comprised 56 RA patients (60 knees) and 56 OA patients (61 knees). Under limited sample conditions, the model demonstrated superior performance in single-modality classification using T1WI (accuracy: 86.2%, AUC: 0.914) compared to T2WI (accuracy: 82.8%, AUC: 0.875). The multimodal model integrating T1WI and T2WI achieved optimal classification (accuracy: 97.5%, AUC: 0.975), outperforming manual classification (accuracy: 94.2%, AUC: 0.932). Grad-CAM heatmaps revealed that the model^,s attention patterns were highly consistent with the clinical-pathological characteristics of both diseases. ConclusionsBy integrating MRI radiomics with deep learning, the proposed classification model effectively overcomes the limitation of insufficient training data through the MAML strategy, enabling accurate and reliable prediction of OA and RA with synovitis in the knee joint. This study provides new technical support and a theoretical foundation for early clinical diagnosis, personalized treatment, and prognostic assessment. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.025 ObjectiveTo investigate the diagnostic efficacy of multi-b-value diffusion-weighted imaging (DWI) based on six diffusion models for metabolic dysfunction-associated steatohepatitis (MASH). Materials and MethodsThirty sprague-dawley (SD) rats were randomly divided into three groups (10 rats each) using a random number table: normal control group, metabolic-associated fatty liver (MAFL) group, and MASH group. The MAFL and MASH groups were modeled by feeding a high-fat diet for 10 weeks and 14 weeks, respectively. After modeling, all rats underwent liver multi-b-value DWI. Six models were used to process the data and obtain quantitative parameters of liver parenchyma: mono-exponential model, intravoxel incoherent motion (IVIM) model, diffusion kurtosis imaging (DKI) model, stretched-exponential model (SEM), fractional order calculus (FROC) model, and continuous-time random walk (CTRW) model. The mono-exponential model parameters included apparent diffusion coefficient (ADC), the IVIM model parameters included pure diffusion coefficient (IVIM_D), pseudo-diffusion coefficient (IVIM_D^*), perfusion fraction (IVIM_f), the DKI model parameters included mean diffusion coefficient (DKI_MD), mean kurtosis coefficient (DKI_MK), the SEM model parameters included distributed diffusion coefficient (SEM_DDC), heterogeneity index (SEM_α), the FROC model parameters included diffusion coefficient (FROC_D), spatial parameter (FROC_μ), the CTRW model parameters included anomalous diffusion coefficient (CTRW_D), spatial diffusion heterogeneity index (CTRW_β) and temporal diffusion heterogeneity index (CTRW_α). For the single-exponential model, ADC₁ was obtained using conventional two b-values, and ADC₂ was obtained using multi-b-values. Immediately after MRI examination, the rats were euthanized, and liver specimens were collected for pathological analysis to obtain the nonalcoholic fatty liver disease (NAFLD) activity score (NAS). One-way analysis of variance (ANOVA) or Kruskal-Wallis test was used to compare parameter differences among groups. Spearman rank correlation analysis was used to explore the correlation between MRI quantitative parameters and NAS. The diagnostic efficacy of each parameter for MASH was analyzed using the receiver operating characteristic (ROC) curve. ResultsThe quantitative parameters of liver parenchyma, including ADC₂, IVIM_D, DKI_MD, DKI_MK, SEM_DDC, FROC_D, CTRW_D, and CTRW_α, showed statistically significant differences between any two groups (P<0.05). ADC₁, SEM_α and FROC_β only differed between the normal group and MASH group (P < 0.05). ADC₁, ADC₂, IVIM_D, DKI_MD, SEM_DDC, SEM_α, FROC_D, FROC_β, CTRW_D, and CTRW_α were negatively correlated with NAS (r = -0.479 to -0.886), while IVIM_f and DKI_MK were positively correlated with NAS (r = 0.460, 0.860). ROC curve analysis showed that ADC₂, IVIM_D, DKI_MD, DKI_MK, SEM_DDC, SEM_α, FROC_D, FROC_β, CTRW_D, and CTRW_α had moderate to high diagnostic efficacy for MASH (area under the curve: 0.780 to 0.960). Among them, ADC₂, DKI_MK, and FROC_D were significantly superior to SEM_α and FROC_β (P < 0.05). ConclusionsMultiple diffusion models can be used for MASH diagnosis, with the ADC value from the multi-b-value mono-exponential model, MK value from the DKI model, and D value from the FROC model demonstrating the best efficacy and are expected to become the best parameters for non-invasive diagnosis of MASH as an alternative to liver biopsy. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.026 ObjectiveTo assess the impact of intelligent quick magnetic resonance (IQMR) reconstruction technology on the image quality and quantitative parameters of diffusion-weighted imaging (DWI) in prostate cancer. Materials and MethodsAxial T2-weighted imaging with fat-saturation (T2WI-FS) and field of view optimized and constrained undistorted single-shot DWI (FOCUS-DWI) sequences, along with clinical data, were retrospectively collected from 31 patients with prostate cancer. The FOCUS-DWI images were processed using the IQMR post-processing system to automatically generate IQMR-FOCUS-DWI images. Two radiologists independently evaluated two sets of images FOCUS-DWI and IQMR-FOCUS-DWI scoring them for noise level, geometric distortion, artifacts, and overall image quality. The signal-to-noise ratio (SNR) of prostate cancer lesions and the contrast-to-noise ratio (CNR) of prostate cancer lesions to the internal obturator muscles were measured and compared between the two image sets. Additionally, the apparent diffusion coefficient (ADC) values of prostate cancer lesions were measured and compared between the two sequences. ResultsQualitative analysis showed that IQMR-FOCUS-DWI images received higher scores than FOCUS-DWI images in noise level, geometric distortion, and overall image quality (P ≤ 0.005). Although IQMR-FOCUS-DWI images also received higher artifact scores, the difference was not statistically significant (P = 0.313). Quantitative analysis revealed that SNR and CNR were significantly higher in IQMR-FOCUS-DWI images compared to FOCUS-DWI images (P < 0.001). There was no statistically significant difference in the lesion ADC values between the IQMR-FOCUS-DWI sequence and the FOCUS-DWI sequence (P = 0.061). ConclusionsCompared to the FOCUS-DWI sequence, IQMR technology significantly improves the image quality of prostate DWI, resulting in higher SNR, CNR, and subjective image scores. ]]> Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.029 Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.030 Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.031 Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.032 Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.033 Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.034 Thu,20 Nov 2025 00:00:00 GMT http://med-sci.cn/cgzcx/en/en_articlexml.asp?doi=10.12015/issn.1674-8034.2025.11.035 Thu,20 Nov 2025 00:00:00 GMT