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.
Fundus imaging has emerged as a valuable noninvasive tool for assessing systemic diseases. Retinal photographs can capture microvascular changes that reflect underlying biochemical and metabolic abnormalities. This study aimed to establish a computer-aided diagnostic framework integrating fundus images with clinical data using machine learning to predict key biochemical markers in patients with chronic kidney disease (CKD), focusing on vascular calcification, inflammation, and cardiac remodeling.
Clinical and imaging data were retrospectively collected from 396 CKD patients treated at a teaching hospital, over a three-year period beginning in January 2021. Each patient contributed 30 clinical variables and a corresponding fundus image acquired closest to the date of biochemical testing. After institutional review board approval (IRB-BM), variables were tested for normality using the Kolmogorov–Smirnov test. Pearson correlation coefficients and decision tree classifiers were used for feature selection and importance ranking. Fundus images were processed using convolutional neural networks (CNNs) or ResNet50 to extract image features. A multilayer perceptron (MLP) model integrated image and clinical features via early fusion to predict abnormal biochemical levels based on established cutoff values for iPTH and NT-proBNP. Model performance was evaluated by accuracy.
The multimodal framework achieved its best performance under the three-class classification model. For iPTH, both ResNet50 and CNN achieved an accuracy of 0.90 under broad criteria, with 0.78 and 0.81 under strict criteria, respectively, and execution times ranging from 13 to 28 seconds. For NT-proBNP, ResNet50 achieved accuracies of 0.64 (broad) and 0.56 (strict), while CNN reached 0.61 (broad) and 0.75 (strict), with execution times between 16 and 41 seconds. Overall, CNN demonstrated slightly superior discriminative performance for NT-proBNP under strict classification, whereas both models performed comparably well for iPTH. The inclusion of both fundus image and clinical data significantly improved predictive accuracy compared with unimodal models. These findings highlight the potential of multimodal deep learning in predicting biochemical abnormalities related to CKD.
This study demonstrates that integrating fundus imaging with clinical data through machine learning enables accurate, noninvasive prediction of biochemical abnormalities associated with CKD. The proposed multimodal framework effectively combines image-derived and clinical features to enhance prediction of iPTH and NT-proBNP abnormalities. This approach provides a scalable and clinically applicable tool for early disease detection, risk stratification, and longitudinal monitoring of CKD, contributing to improved clinical decision support and personalized patient care. Future work should focus on expanding datasets, balancing class distributions, and validating the model across multiple centers to enhance robustness and generalizability.
