Pediatric Radiology最新文献

Background: Gestational age (GA) is essential for assessing fetal development, but conventional methods such as last menstrual period and ultrasound are often inaccurate, particularly in late pregnancy. Recent advances in deep learning (DL) and MRI offer more reliable and consistent GA estimation by capturing detailed fetal brain development.

Objective: This study aimed to develop deep learning models for GA prediction using multi-view fetal brain MRI and to compare their performance with conventional biometric regression techniques.

Materials and methods: A total of 1,321 fetal MRI scans were used to train and evaluate various DL models, while an additional 80 publicly available MRI scans served as an external test set. Two training strategies were explored: transfer learning versus training from scratch, and single-view versus multi-modality input.

Results: The pre-trained ResNet-101 model achieved a mean absolute error (MAE) of 4.47 days and a coefficient of determination (R²) of 0.96 on the internal test set. On the external test set, the model yielded an MAE of 6.57 days, outperforming the biometric regression method, which achieved an MAE of 9.42 days. Explainability analysis revealed that the model predominantly focused on the lateral ventricles, cerebellum, and surrounding brain regions for GA prediction.

Conclusions: The integration of multi-view MRI and transfer learning significantly enhanced the predictive accuracy of DL models for GA estimation. The proposed approach outperformed conventional biometric regression and highlighted clinically relevant anatomical regions, demonstrating its potential for use in prenatal diagnostic applications.

Diffusion tensor imaging (DTI) offers a non-invasive window into kidney microstructure by measuring directional water diffusion. In pediatric populations, where early detection of kidney dysfunction is crucial, DTI shows promise for evaluating structural integrity, diagnosing conditions, and monitoring chronic diseases such as autosomal recessive polycystic kidney disease (ARPKD). This review briefly presents the principles of renal DTI, key acquisition techniques, and important nuances in applying this modality to kidney evaluation. We provide an overview of representative post-acquisition processing pipelines for diffusion tensor generation, tractography, and quantitative analysis. We then summarize current applications of DTI in assessing kidney structure, including its use in select diseases, with focused emphasis on pediatric conditions such as ureteropelvic junction obstruction (UPJO), polycystic kidney disease, and pediatric kidney transplantation. Applications for other renal disorders are also reviewed. Finally, we outline current challenges related to standardization and highlight future research directions needed to refine methodology and further establish the clinical utility of renal DTI.

Restrictive cerebral venopathy was recently described in a young patient with cerebral venous ischemia, elevated intracranial pressure, and intracranial calcifications. It was anatomically characterized by extensive formation of tortuous small to medium-sized cortical veins and angiographic absence of the deep venous system. We report similar, albeit not identical, angiographic features in an 11-year-old girl with infantile autism, attention deficit disorder, dyslexia, and camptodactyly. Angiography revealed a venous anomaly characterized by diffuse marked tortuousities involving mainly pial and small cortical veins, partial maldevelopment of the deep venous system, and aplasia of the transverse sinuses. Magnetic resonance imaging showed no signs of venous ischemia. Genetic analyses identified a complex rearrangement involving three chromosomal segments. In conclusion, a unique case of non-restrictive cerebral venous dysgenesis associated with chromotripsis is presented.

Background: Fluoroscopically guided jejunal tube placement via percutaneous endoscopic gastrostomy (PEG-J) provides minimally invasive post-pyloric access in children. Limited data exist regarding routine application and procedural risks.

Objective: To evaluate the safety and technical success of PEG-J in pediatric patients, performed without general anesthesia or sedation.

Materials and methods: All pediatric cases of fluoroscopically guided PEG-J procedures performed between 2011 and 2025 were included. Fluoroscopic images were reviewed to determine the final position of the tube tip. Technical success, complications, anatomical variants, and tube patency were assessed. Fluoroscopy time and dose area product (DAP) were documented.

Results: A total of 126 PEG-J procedures in 60 children (36 males) were analyzed. The technical success rate was 85% (107/126) with final tube tip placement in the jejunum in 88 cases (82%) and in the duodenum in 19 cases (18%). Nineteen procedures (15%) were unsuccessful, including six with documented anatomical causes (steep vertical duodenal entry, n=2; malrotation, hiatus hernia, hooked stomach in superior mesenteric artery syndrome, steep take-off of the jejunum with kinking of the tube at the ligament of Treitz, n=1 each) and 13 without documented reasons. The median fluoroscopy time was 5 min 24 s (range, 2 s-37 min), at a frame rate of 0.5 frames per second. The median DAP was 6.1 cGy·cm² (range, 0.08-343 cGy·cm²).

Conclusion: Fluoroscopically guided PEG-J placement is a safe and effective procedure in pediatric patients, with high technical success and low radiation exposure.

Artificial intelligence (AI) is increasingly shaping radiology, though its integration into paediatric radiology has progressed more slowly due to challenges specific to the paediatric population. This is especially true in the field of paediatric abdominal imaging. Key barriers include regulatory and ethical issues, the scarcity of large paediatric datasets necessary for algorithm training, reduced vendor interest linked to limited economic incentives, and the inherent differences in children throughout the developmental stages including organ size, signal/sonographic characteristics, and pathologies. Despite these obstacles, AI has the potential to enhance clinical care by augmenting radiologists' workflow across both interpretive and non-interpretive tasks. Currently, most published research focuses on AI's role in musculoskeletal imaging. Although AI is expanding its reach in other imaging domains, paediatric imaging lags behind, as does its potential in abdominal imaging. The use of AI in paediatric abdominal imaging has received limited attention in the existing literature. Emerging research applications cover multiple tasks: detection, classification, functional analysis, severity prediction, automated segmentation, image quality optimization, and acceleration of image acquisition. This review aims to provide practicing radiologists with a concise, simple, and clinically oriented overview of the potential applications and limitations of these new AI tools in paediatric abdominal imaging, categorized by organ. For the time being, most applications described in the literature remain confined to the research setting. To advance these approaches towards clinical utility, validation on larger and more heterogeneous datasets is required. Moving forward, it will be essential to integrate human expertise with AI systems to strengthen diagnostic capacity in paediatric abdominal radiology and to promote paediatric-specific regulatory standards, clear governance structures, and human-centred oversight.