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.
Peritoneal dialysis (PD) adequacy (Kt/V) and peritoneal membrane transport status (PET) guide prescriptions but require time and labour-intensive testing. We designed DETECT-PD to evaluate whether routinely available, simple measurements from the latest PD outflow and a spot urine/blood panel can support AI models to estimate PET and Kt/V, potentially reducing burden of testing and monitoring.
A prospective diagnostic study was set up at Tuen Mun Hospital, Hong Kong. Adults on maintenance PD provided an extra sample of spot urine before their appointment for Kt/V and PET tests; demographics, PD details, and routine labs were abstracted from clinic records. Models were trained for (1) Kt/V regression (Elastic Net, CatBoost, XGBoost, NGBoost) and (2) PET ordinal classification (Random Forest, CatBoost, XGBoost, LightGBM); Feature selection was performed with LASSO-regularized regression with L1 penalty. Discrimination and calibration power for each model were evaluated; the importance of each feature was presented as SHAP. Target sample size is 350; this abstract reports an initial pilot (n=57, including 12 isolated test samples).
The feature selection process selected 6 features for Kt/V regression (serum creatinine, protein, dialysate urea/creatinine/protein, BMI) and 10 features for PET ordinal classification (serum creatinine, albumin, dwell time, urine protein-creatinine ratio, dialysate osmolarity/urea/creatinine, BMI, age, Charlson index), consistent with greater biological heterogeneity in membrane transport(Figure 1a-b). CatBoost performed best in Kt/V regression: R²=0.451, Mean Absolute Error(MAE)=0.241, Mean Squared Error (MSE)=0.105, while XGBoost achieved the best Quadratic Weighted Kappa (QWK)=0.325 in PET ordinal classification, with the lowest Brier score (0.666)(Figure 2a-b, Table 1a-b).
Table 1a: Comparison of Fitness of different models in Kt/V prediction; MAE: Mean Absolute Error; MSE: Mean Squared Error
Table 1b: Comparison of calibration power of different models in PET classification; ECE: expected Calibration Error; QWK Quadratic Weighted Kappa
Using a clinic-friendly single-point sampling panel involving spot blood, urine and dialysate analysis, DETECT-PD delivered promising Kt/V estimation via gradient-boosted trees and feasible ordinal PET classification, suitable for low-burden surveillance and triage, e.g. flagging patients who need confirmatory PET. A larger ongoing cohort(target n≈350) will enable robust model training and validation. DETECT-PD shows the potential to lower patient and administrative workload for adequacy checks and transporter monitoring. The data pipeline and code repo is avaliable at https://github.com/leungkcofficial/detect-pd for open contribution, testing and evaluation.
