Tumor heterogeneity hinders accurate prediction of response to immune checkpoint inhibitors (ICIs) in hepatocellular carcinoma (HCC). We developed a habitat model using routine in-phase (IP) and opposed-phase (OP) MRI sequences, which provide complementary structure- and lipid-related information, to predict response to ICI therapy in HCC. This study included 883 HCC patients from six institutions and public databases. We developed habitat, radiomics, peritumoral, clinical, and combined models to predict partial response (PR) in five institutions. Kaplan–Meier analysis was performed to evaluate survival differences based on prediction scores. Habitat model interpretability was assessed using Shapley additive explanations (SHAP) along with multi-omics and pathological analyses in the public databases and one external institution. The habitat model showed numerically higher AUCs among the independent models, with AUCs of 0.923, 0.918, and 0.897 in the training set and two external validation sets, respectively. The combined model also showed favorable performance with AUCs of 0.959, 0.965, and 0.955, respectively. Calibration curves showed generally close agreement, while decision curve analyses suggested potential clinical net benefit of the habitat and combined models. Patients with higher habitat scores had significantly longer progression‑free and overall survival (all P < 0.05). Exploratory biological analyses showed that the higher habitat score group was associated with enhanced immune activity, downregulated fatty acid-related pathways, and shorter tumor–inflammatory spatial distances. The IP and OP MRI-based habitat model showed potential for treatment-response prediction and prognostic stratification in patients with HCC receiving ICI therapy.
To establish the thermal preservation state as a critical pre-analytical variable in quantitative post-mortem CT of donated bodies, with the aim of enabling reliable multi-donor comparisons, improving the performance of automated segmentation tools, and providing a translational experimental model applicable to CT-guided thermal interventions. Eight body donors underwent whole-body CT in up to three preservation states: fresh, frozen (− 20 °C), and thawed. CT datasets were automatically segmented using TotalSegmentator in 3D Slicer. Mean Hounsfield unit (HU) values were extracted for major organs and tissue groups. The primary analysis was within-donor: Friedman tests with Wilcoxon signed-rank post-hoc comparisons in donors with complete data (n = 7), complemented by linear mixed-effects models using all available scans. Freezing produced a consistent reduction in CT attenuation across all soft tissues and muscle, with near-complete recovery after thawing. In the primary repeated-measures analysis, preservation state had a significant effect after FDR correction in six of the nine tissue groups (liver, spleen, brain, aorta, pancreas and muscle; q = 0.023–0.042), but not in kidney, heart or bone. Brain and pancreas showed the strongest effects (Kendall’s W = 0.80 and 0.76). In the liver, the median within-donor change was − 59.7 HU at freezing and + 60.3 HU at thawing. Mixed-effects models confirmed a significant reduction at freezing in liver, spleen, brain, kidney, pancreas and muscle (β = −35.1 to −52.7 HU) and no residual difference after thawing in any tissue group. Bone attenuation remained stable across all preservation states. Freeze–thaw preservation systematically alters CT attenuation in soft tissues while minimally affecting bone. Preservation state should therefore be considered a critical pre-analytical variable in quantitative post-mortem CT, particularly for studies relying on automated segmentation, radiomics, and translational applications involving thermal tissue imaging.
The petrous portion of the temporal bone is a complex pyramidal-shaped structure of the central skull base, which contains osseous, air cell, vascular, and neural structures in close proximity, making imaging interpretation challenging. A broad spectrum of lesions may involve this region, including inflammatory, neoplastic (benign or malignant), vascular, and osseous lesions. In the pediatric population, ongoing skull base maturation and normal anatomical variability further complicate imaging assessment and may mimic pathology. Cross-sectional imaging plays a central role, with CT providing detailed assessment of osseous structures and MRI offering superior soft tissue characterization. Familiarity with the normal developmental anatomy, common anatomical variants, imaging pitfalls, and characteristic imaging features of pediatric bone lesions is essential for accurate diagnosis and appropriate patient management. This review provides a structured overview of pediatric petrous apex lesions, emphasizing imaging protocols, normal developmental variants, and key imaging features across inflammatory, neoplastic, vascular, and miscellaneous entities.
To evaluate the predictive efficacy and risk stratification potential of coronary computed tomography angiography (CCTA)-derived pericoronary adipose tissue (PCAT) radiomics for major adverse cardiovascular events (MACEs). We retrospectively enrolled patients with suspected coronary artery disease (CAD) who underwent CCTA at our hospital between June 2017 and June 2020, with 5-year follow-up for MACE occurrence. Patients were grouped by MACE status. Machine learning was applied to construct and compare three models: PCAT radiomics model, fat attenuation index (FAI) model, and clinical model. Prospective validation was conducted in 578 patients with suspected CAD undergoing CCTA, who were stratified into high- and low-risk groups by the optimal cutoff of the PCAT radiomics model. The predictive and risk stratification performance of all models was assessed. Fourteen PCAT radiomics features (5 first-order and 9 texture features) were finally screened. The PCAT radiomics model showed significantly superior MACE predictive performance over the FAI and clinical models (P < 0.001). Only the PCAT radiomics model identified significantly higher MACE risk in the high-risk versus low-risk group (HR=2.611, 95
BACKGROUND We investigated whether the cerebral blood volume (CBV) index and pretreatment systemic inflammatory indices (SIIs) provide complementary prognostic information for 90-day functional outcomes after successful endovascular therapy (EVT) for anterior circulation large vessel occlusion and evaluated the advantages of machine learning models in terms of prognostic outcomes and quantified adverse outcomes using the population attributable fraction (PAF). METHODS This retrospective analysis of a prospective registry included 474 patients who were treated between November 2020 and June 2024. An independent 53-patient multicenter cohort was used for external validation. SIIs were calculated from the first venous sample obtained at presentation, before intravenous thrombolysis, and before arterial puncture for EVT. Multivariate logistic regression and eight predictive algorithms were evaluated. PAF estimates were considered exploratory. RESULTS Among 474 patients, 41.6% and 28.5% achieved favorable and excellent outcomes, respectively. Higher CBV indices were more frequent among patients with favorable outcomes, whereas higher inflammatory indices were associated with worse outcomes. Logistic regression showed the highest AUC for mRS 0-2 (0.868, 95% CI 0.806-0.929), while SVM had the highest AUC for mRS 0-1 (0.839, 95% CI 0.764-0.915). The external validation AUCs were 0.816 and 0.802. PAF analysis indicated that improving the CBV index increased favorable and excellent outcomes by 97.2% and 74.8%, respectively, whereas reducing inflammatory overactivation improved them by 22.4%-39.7% and 51.2%-59.5%, respectively. CONCLUSION Baseline CBV index and pretreatment SII values provided complementary prognostic information after successful EVT. However, prospective validation in a broader population is required before clinical implementation.
