RPS 1512 - Body imaging in children: from head to toe

Author Block: J. Kou, Y. Tang; Chongqing/CN
Purpose: Glomerulonephritis (GN) includes a diverse range of kidney diseases that often exhibit subclinical manifestations in children. While renal biopsy is the gold standard, its invasiveness, susceptibility to sampling errors, and time requirements impede rapid diagnosis. This study aimed to create a noninvasive diagnostic model for childhood GN by integrating deep learning and radiomics techniques using renal ultrasound images.
Methods or Background: A total of 469 renal ultrasound images were selected from children undergoing ultrasound-guided biopsy and split into training and validation sets at an 8:2 ratio to train a U-Net model for kidney segmentation. Radiomic features were extracted from the segmented regions and categorized by GN types: IgA nephropathy (127 cases), minimal change disease (83 cases), and Henoch-Schönlein purpura nephritis (103 cases). These categories were also split into training and validation sets at an 8:2 ratio. ANOVA was used for feature selection in the training set, followed by LASSO regression for dimensionality reduction, yielding 37 features. A random forest algorithm was then used to develop a GN classification model, which was evaluated using the validation set.
Results or Findings: The segmentation model demonstrated excellent performance, achieving 95.19% accuracy on the validation set. Thirty-seven selected features were used to build a strong classification model, which showed high accuracy and predictive power across GN categories, with AUC values between 0.91 and 0.98.
Conclusion: The combination of deep learning and radiomics using renal ultrasound images shows great potential for classifying childhood GN subtypes, offering a noninvasive method to enhance diagnostic efficiency and patient outcomes.
Limitations: Firstly, the relatively limited data sources may have introduced some regional bias to our findings. Furthermore, we focused only on the pathological subtypes of three GN, which inevitably limited the scope of the model.
Funding for this study: 0
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: the Ethics Committee of Children's Hospital at Chongqing Medical University.

Author Block: F. Maccioni, V. Cardinale, E. Damato, L. Busato, S. Veraldi, A. Valenti, C. Catalano; Rome/IT
Purpose: Primary sclerosing cholangitis (PSC) is a severe liver disease frequently associated with inflammatory bowel disease (IBD) with a late diagnosis, mostly based on biliary changes at MRCP.
To identify specific biomarkers for PSC using delayed phases of gadolinium enhancement to detect ductal fibrosis, such as intestinal fibrosis in Crohn's disease.
Methods or Background: A prospective study based on the association of MRCP and gadolinium-enhanced MRI, including delayed (7 minutes) phases, was performed in 3 groups of pediatric patients, one with PSC and IBD, one with IBD only, and one of controls. Three radiologists blindly and independently analyzed: a) intra and extrahepatic bile ducts strictures (IHBDs, EHBDs) at MRCP; b) gallbladder volume; c) gallbladder wall gadolinium-enhancement; d) IHBDs-EHBDs gadolinium-enhancement.
Results or Findings: We included 39 patients, 12 with PSC and IBD (31%) , 16 with IBD only (41%) and 11 controls (28%). At MRCP, IHBDs strictures were detected in 82% PSC-IBD patients (p<0.001). Delayed enhancement of gallbladder wall was observed in 100% of PSC-IBD patients (100% sensitivity, 90% specificity, (p<0.001); delayed enhancement of the extrahepatic biliary ducts in 89% (89% sensitivity, 100% specificity, (p<0.001); and delayed enhancement of the intrahepatic biliary duct in 55,6% (56% sensitivity, 100% specificity).
Conclusion: Delayed gadolinium-enhancement of the gallbladder wall and extrahepatic bile duct showed remarkable sensitivity and specificity for PSC. These biomarkers may potentially increase MRI diagnostic accuracy in high-risk IBD patients.
Limitations: The main limitation is the small number of patients.
Funding for this study: No funding
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the ethics commitee of our hospital.

Author Block: J. Chen, F. Xu, Y. Gao, M. Yu, Y. Tang; Chongqing/CN
Purpose: This study constructs a nomogram prediction model based on combi-elastography indexes and laboratory indicators for the differential diagnosis of biliary atresia (BA) and other cholestatic liver diseases (non-BA), with a view to recognizing BA at an early stage and carrying out treatment in time.
Methods or Background: A total of 111 children aged < 180 d with cholestatic hepatitis are included in the study, 75 in the BA group and 36 in the non-BA group. Conventional ultrasound, combi-elastography, and laboratory tests are performed on each patient before pathologic biopsy. The variables are selected through logistics regression to construct a nomogram model, and the effectiveness of the model is evaluated.
Results or Findings: Multifactorial logistic regression analysis shows that gamma-glutamyl transferase (GGT) , total bilirubin (TBIL) , and liver fibrosis-related F index (FI) of the combi-elastography index could be used as independent predictors to differentiate BA from other causes of cholestasis. A nomogram model of these three indexes is constructed which shows better performance, with an area under the operating characteristic curve (AUC) of 0.887 (p < 0.001), sensitivity of 83.3%, and specificity of 81.3%. The internal validation of the model is performed using 1,000 bootstrap resamples and Decision curve analysis indicates that this model had a better diagnostic efficacy and accuracy.
Conclusion: The nomogram model based on combi-elastography indexes and laboratory indicators has certain value in differentiating BA from other cholestatic liver diseases.
Limitations: First, our sample size was small and it was a single-center study.Futher, we can expand the sample size to verify the validity of the nomogram model and provide more valuable insights for clinicians. Second, additional indicators can be included in the future, such as MMP-7 to improve the diagnostic efficacy of the nomogram model.
Funding for this study: Chongging Municipal Science and Health Joint Medical Research Project(2024MSXM050)
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by the Institutional Review Board of the Children's Hospital of Chongqing Medical University (ethical approval number 2024-216), and all examinations and surveys were conducted after obtaining consent from the parents.

Author Block: D. C. Missud, S. Le Cam, I. Mannes, M. Duché, S. Franchi-Abella; Le Kremlin-Bicêtre/FR
Purpose: Spleen stiffness measurement (SSM) is known as a biomarker of significant portal hypertension in chronic liver disease, but has been poorly studied in the context of liver transplantation (LT). SSM may be particularly interesting to evaluate portal vein anastomotic stenosis, a common and serious complication of LT.
Methods or Background: Retrospective study including all portal vein stenoses assessed with transhepatic portal venography (TPV) among paediatric liver recipients between 2015 and 2024. Diagnostic performance of SSM for the diagnosis of portal vein stenosis prior to TPV was evaluated.
Results or Findings: 36 children who underwent 58 TPV (stenosis group) were included, with a median age at TPV of 3.0 years. Findings were compared to those of 58 randomly selected paediatric liver recipients without portal vein complication (control group).
In the stenosis group, there were 17 mild, 20 moderate, and 21 severe stenoses.
SSM was significantly increased in the moderate and severe groups vs no or mild stenosis groups (p < 0.005).
The Area Under the Curve was 0.96 for significant vs no to mild stenosis.
A SSM cut-off of 33.7 kPa led to a sensitivity of 0.976 and a specificity of 0.840 for the diagnosis of significant stenosis.
Conclusion: SSM correlates very well with transhepatic portal venography when there is a suspicion of significant portal vein anastomotic stenosis following LT in paediatric patients. Further studies may demonstrate that SSM is a good biomarker of portal hypertension following LT in paediatric patients, regardless of the aetiology.
Limitations: This is a retrospective study which doesn't have the strength of a prospective clilnical trial.
This is a series of only 58 TPVs, but it is nonetheless one the largest series ever published regarding paediatric liver recipients with portal vein complications.
Funding for this study: No external funding
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: This is a usual care clinical retrospective study, under review by the local ethics committee.

Author Block: Z. C. Beyoğlu, N. G. Akyel, E. Arslantaş, T. Banaz, M. Söyleyici, E. Ayaz, S. Akpınar Tekgündüz; Istanbul/TR
Purpose: The presence of image-defined risk factors (IDRFs) in neuroblastoma plays a crucial role in clinical decision-making, particularly when choosing between primary tumor resection and neoadjuvant chemotherapy. This study aims to evaluate how the presence of IDRFs influences the surgical outcomes of patients with abdominal neuroblastoma, specifically focusing on complete tumor resection and the likelihood of recurrence within one year.
Methods or Background: This retrospective study involved 60 patients diagnosed with abdominal neuroblastoma. Of these, 30 patients had tumors with IDRFs present, while the remaining 30 patients had no IDRFs identified. The recurrence rates and surgical outcomes were analyzed over a one-year follow-up period. Factors such as tumor residuals and overall recurrence rates were examined to assess the impact of IDRFs on surgical success and prognosis.
Results or Findings: A total of 60 patients were included in the analysis. Among the 30 patients with IDRFs, 9 patients showed no signs of recurrence within one year, while 12 patients experienced tumor recurrence, and 9 had residual tumor tissue post-surgery. In the group of 30 patients without IDRFs, 22 patients did not experience recurrence, 5 patients had recurrences, and 3 had residual tumor tissue after surgery. These findings suggest that the presence of IDRFs significantly impacts the likelihood of recurrence and surgical success.
Conclusion: The study found that the presence of IDRFs in abdominal neuroblastoma patients is an important predictor of recurrence. Patients with IDRFs had a lower non-recurrence rate (30%) compared to those without IDRFs (73%). Therefore, IDRF presence can be a key factor in determining surgical approach and prognosis.
Limitations: The study is limited by the small sample size and variability in imaging techniques used during follow-up, which could affect the consistency of the results.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by the Ethics Committee of Başakşehir Çam and Sakura City Hospital, Istanbul/Turkey

Author Block: J. F. Schäfer¹, J. Spogis¹, J. Fuchs¹, B. Hero², T. Simon², A. Eggert³, M. Müller⁴, S. Warmann³, B. Timmermann⁵; ¹Tübingen/DE, ²Cologne/DE, ³Berlin/DE, ⁴Heidelberg/DE, ⁵Essen/DE
Purpose: The SIOPEN HR-NBL2 protocol for the treatment of high-risk neuroblastoma (NB) is a multinational trial with radiation randomization for patients with macroscopic residual tumors after induction chemotherapy and tumor resection. However, precise definitions of residual tumor on cross-sectional imaging are not available yet. This study aims to validate a newly developed scoring system for assessing residual tumors proposed by the German Neuroblastoma registry.
Methods or Background: Patients treated according to the GPOH NB guidelines and irradiated at the West German Proton Therapy Center Essen were retrospectively included if pre-/post-operative and current MRI were available at the time of radiotherapy. The score is based on MRI (tumor size and diffusion restriction), mIBG uptake, and the surgical report, assigning a lesion score from 1 to 3 for each point in time. Anonymized imaging data were uploaded to a browser-based imaging platform (mRay, Germany) for multi-reader, multi-institutional evaluation.
Results or Findings: A total of 15 patients (mean age 5.1y; SD 2.1y) with 45 MRIs were assessed by two experienced readers (R1/R2). The mean preoperative tumor volume was 100 ml (range 3–344 ml). Image-defined risk factors were identified in all patients except one. Based on the surgical reports, complete macroscopic resection was achieved in 9 patients, while imaging analysis by R1 and R2 confirmed complete resection in 4 cases. On a lesion-based analysis, surgery identified 6 residual lesions in 6 patients, while R1 and R2 identified 25 lesions in 11 patients (mean size: 13 mm, range: 4–32 mm). The inter-reader agreement for MRI scoring was excellent (ICC 0.87; 95% CI: 072–0.94).
Conclusion: Regarding MRI findings, the newly developed scoring system for residual tumors in pediatric neuroblastoma is feasible. Further validation through multi-institutional, platform-based readings is planned.
Limitations: Preliminary data
Funding for this study: No Funding.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: University of Essen, Germany

Author Block: G. G. Koodaly, V. Smriti, A. D. Baheti, S. Kulkarni, N. Shetty, K. B. Gala, M. Ramadwar, S. Quereshi, G. Chinnaswamy; Navi Mumbai/IN
Purpose: Tumor rupture poses as a life-threatening complication in pediatric solid tumors, and requires a high-risk protocol management. This study aims to identify the potential imaging predictors for tumor rupture.
Methods or Background: The clinical data of children with pediatric solid tumor rupture at our institution from January 2021 to June 2024 were reviewed retrospectively.
Results or Findings: Total of 22 cases, which comprised of 14 hepatoblastoma, 6 Wilms’ tumor, 1 neuroblastoma, and 1 Ewing sarcoma were analyzed. Patients were aged between 2 to 10 years (median 4.5 years). 17 patients were treatment naïve and 5 on chemotherapy.

Patients commonly presented with abdominal pain, distension, nausea, vomiting, and signs of shock due to significant drop in hemoglobin. Imaging via ultrasound and CT scans revealed tumors, with a notable correlation between tumor rupture and factors such as high PRETEXT scores, tumor sizes over 10 cm, a greater percentage of necrotic component, intratumoral bleed and hyperdense ascitic fluid (>25 Hounsfield units).

Among the hepatoblastoma cases, 71% had PRETEXT III and above. At diagnosis, 64% of the patients had hemoglobin levels ≤ 8 g/L, with 18% at ≤ 6 g/L. Seven patients required angioembolization, 2 underwent surgery and rest were treated conservatively. Unfortunately, one patient developed tumor lysis syndrome and 3 succumbed to tumor rupture.
Conclusion: This study identifies the imaging predictors of tumor rupture in pediatric solid tumors and associated risk factors, such as chemotherapy, larger size of tumor and other high-risk factors.
Limitations: Retrospective study. Its prevalence in cohort is not known.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: Retrospective study

Author Block: R. Smith-Bindman¹, S. Albers², M. Kwan³, W. Bolch⁴, E. Bowles⁵, C. Stewart¹, R. Greenlee⁶, J. Pole⁷, D. L. Miglioretti²; ¹San Francisco, Ca/US, ²Davis, CA/US, ³Oakland, CA/US, ⁴Gainesville, FL/US, ⁵Seattle, WA/US, ⁶Marshfield, WI/US, ⁷Brisbane/AU
Purpose: Risks of hematologic malignancies associated with medical imaging ionizing radiation exposure have not been evaluated in the U.S. or Canada.
Methods or Background: This retrospective cohort study followed 3,724,622 children born at one of 6 U.S. healthcare systems or in Ontario, Canada from 1/1/1996 to 4/30/2016 from birth until the earliest of a cancer diagnosis, death, emigration from Ontario, 6 months after disenrollment from healthcare system, age 21, or 12/31/2017. Active bone marrow radiation doses from medical imaging examinations were estimated. Hazards ratios (HR) and relative risks (RR) of hematologic malignancies associated with cumulative radiation exposure were estimated.
Results or Findings: A total of 2,961 hematologic malignancies were diagnosed during 35,735,719 person-years of follow up, including lymphoid malignancies (n=2,349, 79.3% of malignancies), myeloid or acute leukemia (myeloid, n=460, 15.5%); and histiocytic and dendritic cell malignancies (H&D, n=129, 5.1%). Malignancy risk increased with cumulative dose (p<0.0001); e.g., risk was 1.70 times higher among children with a cumulative dose of 15 to < 20 mGy vs. <1mGy (95%CI=1.27-2.28). The risk of malignancy was 3.7 times higher (95%CI=2.82-4.71) for children with a cumulative exposure of 100 mGy vs. no exposure and was significantly elevated for cancer subtypes. RRs decreased with increasing time since exposure and increase with age at exposure and attained age. We estimate 27 excess hematologic malignancies by age 21 per 10,000 children with a cumulative exposure of 30 mGy or higher vs. <1 mGy, equivalent to average dose of approximately 2 head CTs.
Conclusion: Children and adolescents who undergo radiation-based medical imaging are at a small, but significant increased risk of hematologic malignancy.
Limitations: While reverse causation is a potential limitation, analyses of clinical indications confirmed symptoms related to hematologic malignancy were rare in included studies.
Funding for this study: US National Institutes of Health, National Cancer Institute R01CA185687, R50CA211115
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The requirement for individual informed consent was waived for the study

Author Block: A. A. Mohammed, M. F. Mcentee, A. England, N. Moore, E. K. Mahon, M. Maher, R. Young; Cork/IE
Purpose: The aim of this study is to compare two whole-body CT protocols for SPA and to assess whether the ULD protocol can provide diagnostic images of sufficient quality compared to the standard dose (STD).
Methods or Background: In this cross-sectional study, two sets of images of a newborn whole-body anthropomorphic phantom were acquired using different protocols, one with STD and the other with ULD protocol. The effective dose (ED) of both protocols was calculated using the Monte Carlo dose simulation approach.
The image quality arising from both protocols was then assessed at the ECR 2024 Congress using a four-section questionnaire. The questionnaire included demographic information, a comparison of the visualization of different bony anatomical structures, and confidence in diagnosis using either protocol. The Wilcoxon signed-rank test was used to evaluate the significant differences between STD and ULD image quality scores. VGC analyser was used for image quality rating and comparison.
Results or Findings: 46 participants were included in this study. For all body parts, STD showed significantly higher image quality than ULD (AUCVGC=0.75). 76% of the participants were confident to use the STD protocol for SPA diagnosis, whereas, 41% were confident to use the ULD protocol. The percentage effective dose difference between protocols was 93.5%(STD=0.56 mSv vs. ULD=0.04mSv) and most of the participants underestimated the dose reduction.
Conclusion: This study successfully compared the STD and ULD whole-body CT in phantom and shows ULD CT is a promising technique which may compete with digital radiography for SPA diagnosis.
Limitations: The limitations of the study are:
1-the use of phantom involves no patient movement experienced, and no pathologies.
2-sampling bias was due to the observers as only people attending the ECR 2024 could participate.
3-only compares two whole-body CT protocols.
Funding for this study: Taif University
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The ethical approval provided by the University College Cork

Author Block: S. Consalvo, A. Curto Vilalta, A. W. Marka, S. Breden, B. Schlossmacher, C. Eisfeld, D. Rückert, R. Von Eisenhart-Rothe, F. Hinterwimmer; Munich/DE
Purpose: To address the challenge in musculoskeletal radiology of early detection of bone tumours in children with x-ray imaging. While machine learning (ML) has shown proficiency in differentiating tumour entities, a critical gap remains in initial tumour detection, particularly for non-oncology-trained professionals and general practitioners. Paediatric tumours are often incidentally discovered, underscoring the need for more sophisticated tools.
Methods or Background: This retrospective study utilized X-ray data from a diverse cohort of paediatric patients from our local musculoskeletal tumour database. The dataset comprised 817 images (567 pathological and 250 healthy) from 511 patients, including ten benign, intermediate and malignant tumour entities. We employed the ResNet18 architecture for classification, supported by cross-validation techniques and excessive data augmentation strategies. Our methodology focused on enhancing the ML system's ability to generalize across various clinical scenarios.
Results or Findings: The ML model demonstrated high performance with an accuracy of 96.39%, a sensitivity of 96.0%, and a specificity of 96.0% into “tumour” and “no tumour”. The variance in cross-validation splits was 0.05, 0.10, and 0.13, respectively, indicating stable results across different test sets. These metrics reflect the model's reliability and potential effectiveness in clinical settings.
Conclusion: Current ML applications in orthopaedic oncology are progressing yet remain insufficiently performant for widespread clinical use. However, our findings underscore the potential of ML tools in aiding both young professionals and general practitioners. Additionally, future advancements should focus on multimodal approaches that incorporate not only X-ray data but also MRI and, crucially, clinical data. Integrating these diverse data sources will enhance the performance and applicability of ML in diagnosing and managing paediatric bone tumours, offering a more holistic and effective approach to patient care.
Limitations: 90% Monocentric Data.
Funding for this study: Nemetschek Innovation Foundation and Bavarian Ministry of Science and the Arts
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Klinikum rechts der Isar, Technical University of Munich.