Quantitative Imaging in Medicine and Surgery最新文献

Background: The heterogeneous quality of medical ultrasound (US) imaging across different geographical regions presents a significant challenge for developing robust artificial intelligence (AI) systems in healthcare. While high-income regions benefit from standardized imaging protocols and advanced equipment, resource-constrained environments often produce data with pronounced quality variations, limiting the generalization capabilities of conventional deep learning models. We introduce USHydraNet, a novel multi-decoder framework designed to manage the heterogeneity of multi-regional US images in medical image analysis.

Methods: USHydraNet integrates a vision transformer (ViT) encoder or UNet encoder with multiple decoders, optimizing feature extraction across diverse data distributions. The framework employs a dynamic routing paradigm that selects the optimal decoder output by analyzing image-level and feature-level statistical moments. This adaptive mechanism enables robust performance across varying data quality levels without compromising accuracy.

Results: Experimental validation across four public and one private medical datasets demonstrates that the USHydraNet model achieves superior performance over baseline architectures. In classification tasks, Ablation studies revealed that all metrics of the USHydraNet model showed improvements ranging from 10% to 20%. Comparative evaluations indicated that the USHydraNet model outperformed the four other models in terms of metric gains. For segmentation tasks, ablation experiments revealed that USHydraNet improved Dice scores by 12.23% and 19.52%, and intersection over union (IoU) by 0.59% and 1.05% across two datasets. Comparative experiments demonstrated that the USHydraNet model outperformed the four other models in all metric improvements. Even on unfamiliar datasets, it maintained robust performance with 89.26% Dice and 84.14% IoU.

Conclusions: USHydraNet is a promising framework for reducing performance disparities in medical image analysis across regions with varying healthcare infrastructures. thereby promoting equitable access to AI-assisted diagnosis in resource-limited settings.

Background: Chronic total occlusion (CTO) percutaneous coronary intervention (PCI) with the Guidezilla catheter (Boston Scientific) via invisible septal collateral channels (CCs) has not been extensively investigated. This study aimed to evaluate the incremental value of invasive coronary angiography (ICA) characteristics combined with the Guidezilla catheter for predicting technical success during retrograde CTO PCI through invisible septal CCs.

Methods: We analyzed 166 consecutive retrograde CTO PCI procedures performed via invisible septal CCs (with or without the Guidezilla catheter) across 40 centers from January 2019 to December 2022. Clinical, ICA, and procedural characteristics, along with outcomes, were assessed. Logistic regression analysis was performed to identify technical success predictors, while the discriminatory power of the multivariate model was evaluated via receiver operator characteristic (ROC) curves.

Results: The overall technical success rate for CTO PCI via invisible septal CCs was 81.33%. Multivariate analysis identified the following characteristics as independent and significant predictors of technical success via invisible septal CCs: a ratio of the right coronary artery-posterior descending artery length (RCA-PDA length) to the length of the PDA origin to the apex (LOPA) greater than 0.64 [hazard ratio (HR) =6.19, 95% confidence interval (CI): 2.55-14.99; P<0.001] and the use of the Guidezilla catheter (HR =2.42; 95% CI: 1.01-5.79; P=0.048). The combination of RCA-PDA length:LOPA ratio and Guidezilla status demonstrated superior predictive capability [areas under the ROC curve (AUC) =0.74; 95% CI: 0.67-0.81] compared to the Guidezilla-only model (AUC =0.62; 95% CI: 0.54-0.69; P=0.017).

Conclusions: Incorporating the ICA characteristic of the RCA-PDA length:LOPA ratio significantly improved predictive accuracy when compared to the use of Guidezilla catheter status alone. Thus, ICA characteristics provide significant incremental value in predicting technical success when the Guidezilla catheter is applied for retrograde CTO PCI via invisible septal CCs.

Background: Kidney transplantation (KT) improves left ventricular (LV) systolic function, but few studies have examined its impact on left atrial (LA) function. This study aimed to assess LA structural and functional changes in patients with end-stage renal disease (ESRD) after KT through use of two-dimensional speckle-tracking echocardiography (2D-STE).

Methods: A prospective cohort of 163 patients with ESRD (85 in the KT group; 78 in the dialysis group) who underwent echocardiography at baseline and 12-month follow-up was included, with LA strain and stiffness being measured via 2D-STE. Inter- and intragroup changes were compared, with multivariate linear regression and adjustment for confounders.

Results: KT significantly improved LA function at 12 months: LA reservoir strain (LASr) increased from 39.0%±11.12% to 42.9%±13.74% (P=0.043), and LA conduit strain (LAScd) increased from 22.4%±7.52% to 25.3%±7.99% (P=0.016). LA volume index (LAVI), LA stiffness (LASt), and LA contractile strain (LASct) showed no significant changes. No improvements were observed in the dialysis group. At follow-up, the KT group had significantly better LA function than did the dialysis group, as indicated by LASr (42.9%±13.74% vs. 36.1%±13.92%; P=0.004), LAScd (25.3%±7.99% vs. 20.1%±10.16%; P<0.001), and LASt (0.25±0.14 vs. 0.40±0.29; P<0.001), with significance persisting after adjustment. Structural parameters (LAVI and LASct) did not differ between the groups.

Conclusions: Our study provides insights into the cardiovascular protective mechanisms of KT and indicates that it can improve LA reservoir and conduit function, maintain lower stiffness in patients with ESRD, and thus potentially delay atrial remodeling and stiffness progression.

Background: Trabecular microarchitecture is a primary contributor to bone strength. The aim of this study was to characterize the trabecular microarchitecture of the normally ageing population by sex using multi-detector computed tomography (MDCT) scans.

Methods: We retrospectively searched the medical records of patients receiving routine MDCT scans with the spine protocol between 2014 and 2021. The final random cohort without systemic diseases, consisting of 360 participants (180 women) aged 50 to 80 years and matching sex, was divided into six groups (each group containing a 5-year interval). The L1 vertebra was used for trabecular microarchitecture analysis with European Conformity (CE)-marked software.

Results: We observed differences by sex in trabecular thickness (Tb.Th) mean, trabecular separation (Tb.Sp) mean, and trabecular number (Tb.N) in two age groups (55-59 and 75-80 years). We found that, in both sexes of older adults, Tb.N was significantly negatively correlated with age. From age 60 to 69 years, bone volume to total volume (BV/TV) percentage, Tb.N, fractal dimensions [two-dimensional fractal dimension (D2D) and three-dimensional fractal dimension (D3D)] gradually decreased, whereas Tb.Th mean, and Tb.Sp mean increased. Remarkably, the rates of decline in Tb.N and D2D were greater compared to the others, and the rates of increase in Tb.Sp mean were greater compared to the others. D3D was the most sensitive trabecular microarchitecture parameter, consistently detecting sex-related differences across age group of 50 to 69 years.

Conclusions: This study demonstrates that vertebral trabecular microarchitecture undergoes significant age- and sex-specific changes that are detectable using clinical high-resolution computed tomography (CT). The identification of critical remodelling phases-particularly in men between 55-59 and 75-80 years-and the consistent sensitivity of D3D to sex differences highlight the value of CT-based architectural assessment. These findings suggest that routine abdominal CT, when enhanced with advanced reconstruction techniques, may serve as a useful tool for early detection of bone fragility and support individualized strategies for fracture risk assessment.

Background: Patients in the surgical intensive care unit (SICU) frequently present with coagulation disorders and cardiopulmonary interaction disturbances. This study aimed to examine alterations in coagulation status and pulmonary vascular resistance (PVR), and compare their predictive efficacy for in-hospital mortality in this population.

Methods: In this prospective observational study (January 2025 to April 2025), 61 consecutive SICU patients were enrolled and stratified into survivors (n=43) and non-survivors (n=18). Comprehensive echocardiographic assessments were performed to quantify PVR. Coagulation test and thromboelastography (TEG) were used to monitor coagulation status. Multivariable logistic regression identified independent predictors of in-hospital mortality.

Results: Non-survivors exhibited marked prolonged activated partial thromboplastin time (APTT) [33.20 (30.50-39.30) vs. 28.70 (24.60-30.40) s, P<0.0001], elevated fibrin degradation products (FDP) level [36.00 (15.00-54.30) vs. 16.70 (7.50-33.60) μg/L, P<0.0001], elevated PVR [3.08 (2.61-3.71) vs. 1.83 (1.42-2.35) Wood units, P<0.0001] and pulmonary artery systolic pressure (PASP) [42.00 (34.00-49.00) vs. 31.00 (24.00-44.00) mmHg, P<0.0001] compared to survivors, alongside significant reductions in cardiac chamber dimensions, systolic function and reduced cardiac output (all P<0.05). Multivariate analysis identified PVR [odds ratio (OR) =284.280; 95% confidence interval (CI): 23.887-3,383.278; P<0.0001], APTT (OR =1.799; 95% CI: 1.367-2.369; P<0.0001), and FDP (OR =1.155; 95% CI: 1.071-1.244; P=0.0002) as independent in-hospital mortality determinants. Receiver operating characteristic (ROC) analysis demonstrated superior predictive performance of PVR [area under the curve (AUC) =0.865; cutoff, 2.175 Wood units; sensitivity, 97.94%; specificity, 73.36%] and the combined model incorporating PVR, APTT, and FDP (AUC =0.978; cutoff, 0.593; sensitivity, 92.59%; specificity, 97.40%) compared to isolated APTT (AUC =0.749; cutoff, 29.500 s; sensitivity, 83.52%; specificity, 71.18%) or FDP measurements (AUC =0.645; cutoff, 26.000 μg/L; sensitivity, 70.21%; specificity, 69.87%) (all pairwise comparisons, P<0.05).

Conclusions: Among SICU patients, echocardiography-derived PVR demonstrates superior clinical utility over coagulation tests and TEG in predicting in-hospital mortality among SICU patients, highlighting the prognostic significance of PVR and the combined model incorporating PVR, APTT, and FDP in critical illness outcomes.

Background and objective: Epimuscular fat (EF) is a distinct adipose tissue located outside the epimysial border of muscles, with growing interest in its biomechanical and metabolic functions. Although initial studies have investigated EF in the rotator cuff, its role in the lumbar spine and its potential relationship with low back pain (LBP) remain poorly understood. This narrative review examines the structural, functional, and molecular characteristics of EF, focusing on its presence in the paraspinal muscles and its potential impact on spinal stability and pathology.

Methods: A narrative literature search was conducted to identify key studies investigating EF through imaging, histological, and molecular analyses. The databases searched included PubMed, Cochrane Library, and Scopus, with additional manual screening via Google Scholar and reference lists.

Key content and findings: Seven studies were included. Findings indicate that, at the spinal level, EF accumulates predominantly in the lower lumbar region (L4-S1), where it shows a strong association with body mass index (BMI) and age. EF may disrupt muscle-fascia interactions, altering spinal biomechanics and potentially contributing to LBP. Furthermore, EF exhibits characteristics of beige adipose tissue, distinguishing it from intramuscular fat (IF) and suggesting a potential metabolic function. Advances in magnetic resonance imaging (MRI) have improved EF detection, particularly through DIXON and IDEAL sequences, although a lack of standardized segmentation protocols limits its diagnostic and research applications.

Conclusions: This review highlights the emerging role of EF in spinal health, emphasizing the need for further studies to clarify its contribution to LBP pathophysiology. Understanding EF's properties and its interaction with spinal biomechanics could open new avenues for diagnostic and therapeutic interventions in musculoskeletal disorders.

Background: There is currently no individualized prognostic tool to predict outcomes in patients with brain tumors following status epilepticus (SE), despite its clinical importance for counseling and therapeutic decision-making. This proof-of-principle, retrospective monocentric study investigated whether electroencephalography (EEG)-derived functional connectivity patterns differ in brain tumor patients after SE with respect to survival and tumor type.

Methods: EEG data from 37 brain tumor patients with SE were analyzed. Thirty epochs per frequency band (delta-gamma) were selected using spectral power. Source-space connectivity was measured via weighted Phase Lag Index (wPLI). Permutation tests compared connectivity between survival (>1 vs. <1 year) and tumor subgroups (glioma, meningioma, metastases).

Results: The cohort had a mean age of 68.1 years; 67.6% were female. One-year survival was 32.4%, with a mortality rate of 46.0% and 21.6% lost to follow-up. Tumor diagnoses included 11 adult-type diffuse gliomas, 12 meningiomas, 11 brain metastases, and single cases of primary cerebral lymphoma, schwannoma, and pineal tumor. Higher source connectivity in the alpha and beta bands was observed in patients with longer survival, and higher delta band connectivity in patients with brain metastases compared to those with glioblastoma or meningioma.

Conclusions: Functional connectivity analysis in source space did not reliably differentiate patients by survival or tumor type following SE. These findings suggest that current connectivity metrics are insufficient as standalone prognostic tools in this context. Larger, more homogeneous cohorts and stratification by clinical and tumor characteristics are needed to clarify the potential prognostic value of EEG connectivity in this population.

Background: Four dimensional (4D) flow magnetic resonance imaging (MRI) enables both visualisation and quantification of hemodynamics in abdominal aortic aneurysms (AAAs). However, its clinical implementation remains limited due to time-consuming and cumbersome post-processing, as well as a lack of standardisation. While previous work has proposed automated solutions for thoracic and intracranial cases, applications in the abdominal region, particularly for AAAs, remain underexplored. The aim of this study was to develop an automated post-processing pipeline for 4D flow MRI in AAA patients, incorporating aneurysm identification, via automated segmentation, blood flow visualisation and hemodynamic parameter quantification.

Methods: 4D flow MRI and three dimensional (3D) cine balanced steady state free precession (bSSFP) scans were acquired from 16 patients with an AAA in the same field of view and with an isotropic resolution of 1 mm. Using 5-fold cross-validation, an nnU-Net was trained to segment the aorta based on 3D cine bSSFP MRI. The automated post-processing pipeline was built using MATLAB. Wall shear stress (WSS) values were assessed within the extracted aneurysm. nnU-Net performance was assessed with the Dice-similarity coefficient (DSC), 95% percentile Hausdorff distance (HD95) and by calculating the Pearson correlation coefficients between WSS values obtained in the aneurysm based on segmentation with both methods.

Results: The resulting post-processing pipeline demonstrated robustness to anatomical variability, including tortuosity and intraluminal thrombus (ILT), both characteristic of AAA, and consistently produced output without user-interference. The nnU-Net performance was excellent for lumen segmentation and good for ILT segmentation, with DSCs of 0.93 (0.05) and 0.80 (0.26) respectively. The HD95 was 4.87 (8.59) mm and 6.04 (3.71) mm for respectively lumen and thrombus segmentation. Significant correlations (ρ =0.92 or higher) were found between WSS values derived using the manual and nnU-Net segmentations.

Conclusions: An automated post-processing pipeline was developed specifically for 4D flow MRI in patients with AAA. The pipeline is robust for varying anatomies and may facilitate 4D flow MRI implementation in the clinical workflow.

Background: Stereotactic body radiotherapy (SBRT) is used for inoperable hepatocellular carcinoma (HCC) patients, yet radiation-induced liver disease (RILD) poses risks, particularly in fibrotic livers. T1rho imaging, which allows high scan-rescan stability and high sensitivity to changes from baseline measure, has emerged as a promising noninvasive biomarker for liver injury but its utility in RILD remains underexplored. This study characterizes liver T1rho change in rat models of RILD with/without pre-existing fibrosis.

Methods: Male Sprague-Dawley (SD) rats were divided into radiation therapy (RT) (n=41; 25 Gy right liver irradiation) and thioacetamide (TAA) + RT (n=46; 6-week TAA-induced fibrosis + 20 Gy RT) groups. Nine control rats, and each experimental rat at the endpoint had in vivo T1rho MRI and followed by histology. Pathological assessments [hematoxylin-eosin (HE), Masson's trichrome (MT), Picro-Sirius Red (PSR), transforming growth factor-beta (TGF-β), and alpha-smooth muscle actin (α-SMA)] were performed at 2, 4, 8, and 12 weeks post-RT to quantify fibrosis, collagen, inflammation, and ballooning degeneration.

Results: The control rats had normal liver T1rho value of 37.2±0.93 ms. An RT of 25 Gy led to an elevated liver T1rho of 40.14±1.06 ms 2 weeks post-RT (P<0.001). With TAA pre-treatment, 20 Gy RT led to an elevated liver T1rho of 45.24±1.24 ms 2 weeks post-RT, which is statistically significantly higher than the liver T1rho 2 weeks post-RT with the dosage of 25 Gy (P<0.001), suggesting pre-existing liver injury by TAA sensitizing liver to RT injury. For both the RT and TAA + RT model, there was a treatment-duration dependent elongation of liver T1rho value. Liver T1rho values were strongly correlated with collagen-based histological markers (mostly, Pearson r>0.9). Tentatively Inflammation contributed to T1rho elongation beyond collagen deposition, while ballooning degeneration did not show a notable effect on liver T1rho.

Conclusions: This study demonstrates that T1rho MRI is a sensitive and noninvasive biomarker of RILD severity in rat models, reflecting both fibrosis and inflammation. These findings support its potential utility for early detection, individualized risk assessment, and mechanistic studies of RILD, particularly in clinical settings where histological validation is not feasible.