Jiacheng Song

Background The optimal acceleration strategy for deep learning reconstruction (DLR) in cervical cancer T2-weighted imaging (T2WI) remains unestablished. Methods Forty-eight patients with primary cervical cancer prospectively underwent 3.0T pelvic MRI. Sagittal T2WI was acquired using conventional parallel imaging (Conv) and DLR at acceleration factors (AF) 2 and 3. Two radiologists evaluated quantitative metrics and six qualitative metrics. Results Compared to conventional imaging at the same acceleration, DLR significantly improved signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) (DL_AF2 vs. Conv_AF2, adjusted p = 0.005; DL_AF3 vs. Conv_AF3, adjusted p < 0.001). No significant quantitative differences were found between AF2 and AF3 for either reconstruction method. Despite statistically comparable objective metrics between DL_AF2 and DL_AF3, DL_AF2 received significantly higher subjective scores across all six qualitative dimensions (all adjusted p < 0.001), with median scores of 5 for all image quality metrics. DL_AF3 demonstrated significantly lower diagnostic confidence scores, suggesting perceptual over-smoothing at higher acceleration. Conclusion At equivalent scan times, DLR with moderate acceleration (AF 2) achieves the optimal balance between objective image quality and subjective anatomical fidelity for sagittal T2WI in cervical cancer evaluation.

Rationale and objectives Pregnancy-induced hypertension (PIH) is a common and serious pregnancy complication that threatens maternal safety and perinatal outcomes. We aimed to develop a nomogram model integrating intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) microstructure and microcirculation and multisequence MR texture features for predicting adverse outcomes in PIH. Materials and methods This retrospective study enrolled 63 PIH patients (37 with, 26 without adverse neonatal outcomes). Texture features were extracted from T1-, T2-weighted, and IVIM-DWI images. IVIM-DWI placental parameters included one DWI-based virtual magnetic resonance elastography (vMRE) parameter and six IVIM-based diffusion and perfusion parameters. Three predictive models were constructed: model 1 (texture score [T-score]) utilized multisequence texture features, model 2 incorporated IVIM-DWI parameters and model 3 (a nomogram model) combined T-score and IVIM-DWI parameters. Model performances were assessed via receiver operating characteristic (ROC) curves, area under the curve (AUC), decision curve analysis (DCA), and calibration. Results Texture analysis identified 12 features from three sequences for model 1, which demonstrated an AUC of 0.796. The placental shear modulus (μ_diff) from vMRE and perfusion fraction (f) from IVIM were identified as predictors of adverse neonatal outcomes in PIH patients; model 2 using these two parameters displayed an AUC of 0.785. Model 3, the nomogram model that included T-score, μ_diff, and f, exhibited enhanced predictive accuracy with an AUC of 0.844, sensitivity of 83.8%, and specificity of 73.1%. Conclusion The nomogram model constructed using placental IVIM-DWI-derived microstructural, microcirculatory, and multisequence texture features may facilitate predicting adverse neonatal outcomes in patients with PIH.

Objective: To establish a comprehensive model including MRI radiomics and clinicopathological features to predict post-operative disease-free survival (DFS) in early-stage (pre-operative FIGO Stage IB-IIA) cervical cancer. Methods: A total of 183 patients with early-stage cervical cancer admitted to our Jiangsu Province Hospital underwent radical hysterectomy were enrolled in this retrospective study from January 2013 to June 2018 and their clinicopathology and MRI information were collected. They were then divided into training cohort (n = 129) and internal validation cohort (n = 54). The radiomic features were extracted from the pre-operative T1 contrast-enhanced (T1CE) and T 2 weighted image of each patient. Least absolute shrinkage and selection operator regression and multivariate Cox proportional hazard model were used for feature selection, and the rad-score (RS) of each patient were evaluated individually. The clinicopathology model, T1CE_RS model, T1CE + T2_RS model, and clinicopathology combined with T1CE_RS model were established and compared. Patients were divided into high- and low-risk groups according to the optimum cut-off values of four models. Results: T1CE_RS model showed better performance on DFS prediction of early-stage cervical cancer than clinicopathological model (C-index: 0.724 vs 0.659). T1CE+T2_RS model did not improve predictive performance (C-index: 0.671). The combination of T1CE_RS and clinicopathology features showed more accurate predictive ability (C-index=0.773). Conclusion: The combination of T1CE_RS and clinicopathology features showed more accurate predictive performance for DFS of patients with early-stage (pre-operative IB-IIA) cervical cancer which can aid in the design of individualised treatment strategies and regular follow-up. Advances in knowledge: A radiomics signature composed of T1CE radiomic features combined with clinicopathology features allowed differentiating patients at high or low risk of recurrence.

Objectives: To determine the diagnostic ability of cervical mucosa radiomics signature of sagittal T₂WI and T₁ contrast-enhanced (CE) imaging in detecting early-stage cervical cancers with negative MRI. Methods: Preoperative images of postoperative pathology confirmed early-stage cervical cancer patients and normal cervix patients admitted to our hospital between January 2013 and December 2020 were retrospectively reviewed. Patients with cancer signals on T₂WI, T₁CE and DWI were deleted. Regions of interests (ROIs) were delineated on cervical mucosa (from cervical canal to cervical dome) with 5 mm width on sagittal T₂WI and T₁CE. The maximum-relevance and minimumredundancy (mRMR) and least absolute shrinkage and selection operator (LASSO) methods were used for the calculation of radiomics signature scores. Diagnostic performance was assessed and compared between radiomics prediction models (model 1: T₁CE; model 2: T₂WI; model 3: model one combined with model 2). Differential diagnostic ability of radiomics signature in detecting lymphatic vascular space invasion (LVSI) was further explored. Results: Diagnostic performance of model three was higher than model 1 and model 2 both in primary (model 3 0.874, model 1 0.857, model 2 0.816) and validation (model 3 0.853, model 1 0.847, model 2 0.634) cohorts. Model 3 showed statistical diagnostic difference compared with model 2 (primary p = 0.008, validation p = 0.000). However, the diagnostic improvement ability of model 3 showed no statistical difference compared with model 1 (primary p = 0.351, validation p = 0.739). Diagnostic efficiency of model 3 in detecting LVSI was not apparent (AUC 0.64). Conclusions: Radiomics analysis of cervical mucosa combining T₁CE and T₂WI is promising for predicting MRI invisible early-stage cervical cancers, however further ability in detecting LVSI was not apparent. Advances in knowledge: Conventional MRI was originally defined as meaningless in very early-stage cervical cancers. However, whether MRI radiomics analysis of cervical mucosa can detecting tiny changes of invisible early stage cervical cancers has not been researched yet.