Back
For best output, select "Paper Size" as "A4" and "Margin" as "0" or "None".
To save or print to PDF, please select Print Destination > Save as PDF, enable Background Graphics under "More Settings", then click "Save".
During the congress, E-Posters will be accessible to all participants on the congress website 24/7, as well as in the E-poster stations in the congress center.
Preparing your E-Poster
Please review the E-Poster format requirements carefully when preparing your E-Poster. Should your E-Poster not meet the mentioned requirements, it may not be displayed as described above.
E-Poster Submission Deadline
Please prepare and upload your E-Poster no later than March 14, 2026 11.59PM CET. After this date, you will no longer be able to prepare and upload your E-poster and it will not be displayed and accessible on the congress website.
Please follow the instructions below to input your abstract title.
Abstract titles should be brief and reflect the content of the abstract.
Long-term kidney transplant survival in pediatrics remains challenging. Biopsy remains the gold standard for detecting kidney allograft injury, but it is invasive and by obtaining <0.1% of the kidney (~20 nephrons) are prone to sampling bias. This study proposes a radiomics-based quantitative ultrasound approach to detect parenchymal changes linked to biopsy-proven positive pathology (path+) and rejection (rejection+).
Under a prospective IRB-approved study pediatric kidney transplant recipients underwent allograft ultrasound within 24 hours before surveillance and for-cause biopsies (Dec 2020–Aug 2025). Renal parenchyma was manually segmented on sagittal images, excluding artifacts, by a radiologist blinded to biopsy results and clinical information. Radiomic features across multiple domains were extracted from grayscale ultrasound images. These included gray-level co-occurrence matrix (GLCM), gray-level run length matrix (GLRLM), gray-level size zone matrix (GLSZM), local texture estimation (LTE), fractal dimension texture analysis (FDTA), and local binary pattern (LBP). A Random Forest model was trained to predict (1) path+ (determined by Chronic Allograft Damage Index score ≥2 and/or C4d positivity and/or category 6 positivity by Banff 2017 criteria) and (2) rejection+ (acute or chronic), using the top 10 most informative features selected by their mean Gini importance. Model performance was evaluated using 5-fold stratified group cross-validation reporting accuracy, precision, sensitivity, specificity, F1 score, and area under the receiver operating curve (AUROC).
Thirty-three transplant recipients (median age 16 years) underwent 50 ultrasound studies. The Random Forest model achieved an accuracy of 0.66, precision 0.74, sensitivity 0.84, specificity 0.15, F1 score 0.78, and an AUROC 0.66 for predicting path+ allograft, and accuracy of 0.60, precision 0.44, sensitivity 0.39, specificity 0.72, F1 score 0.41, and an AUROC 0.59 for predicting rejection+. Texture-based metrics such as GLRLM, GLSZM, and GLCM were most predictive for path+, while FDTA and GLSZM metrics were the most predictive features for rejection+.
Despite moderate performance of the model, radiomic features successfully identified subtle textural and morphological patterns linked to relevant histopathologic changes. The model performed better at identifying non-rejection cases and path+ allografts. A Larger dataset is needed to enhance model robustness and promote image-based, noninvasive graft surveillance that could potentially spare invasive biopsies.
This abstract content has been submitted to International Pediatric Radiology Meeting, Boston 2026. Re-submitting the abstract is permitted by the organizers of the original meeting(s).
