CKDX-Net: A Novel Cross-Domain Knowledge Distillation Framework from Tree-Based to Neural Architectures for Chronic Kidney Disease Staging with Adaptive Computational Optimization

DOI

10.2991/978-94-6239-664-7_17How to use a DOI?

Keywords

Biomedical applications; Chronic kidney disease; Computational efficiency; Deep neural network; Knowledge distillation; Temperature-scaled soft labelling; XGBoost

Abstract

The staging of chronic kidney disease (CKD) in low resource settings requires the right balance between diagnostic accuracy and computational efficiency. In this study, we propose CKDX-Net, a novel knowledge distillation pipeline transferring the patterns learned from a high-capacity XGBoost ensemble to a lightweight deep neural network in tasks of six-stage CKD classification. This framework adopts the dual-loss architecture consisting of hard-label cross-entropy and temperature-scaled Kullback-Leibler divergence, accompanied by an adaptive soft label scheduler balancing teacher-student guidance actively in training. We evaluated on 4,000 anonymized patient records sourced from the publicly-available Kaggle CKD dataset, stage 0-5, with a macro-F1 score of 0.918 and retaining the performance over 98.7% compared to the teacher model (99% accuracy); the distilled student model achieves 95% accuracy. The distilled student network speeds up the inference from 105.3 ms to 6.7 ms per record, thus facilitating real-time deployment in clinical pipelines. Temperature-scaled soft labeling achieves stronger knowledge transfer than training from scratch. CKDX Net constitutes the first unified solution to cross-architecture distillation in CKD staging and offers a viable option for deploying high-capacity diagnostic AI in resource-stricken healthcare environments.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Md. Musfiqur Rahman Akib
AU  - Habibur Rahaman
AU  - Rosni Akter
AU  - Pial Paul
PY  - 2026
DA  - 2026/06/08
TI  - CKDX-Net: A Novel Cross-Domain Knowledge Distillation Framework from Tree-Based to Neural Architectures for Chronic Kidney Disease Staging with Adaptive Computational Optimization
BT  - Proceedings of the International Conference on Intelligent Data Analysis and Applications (IDAA 2025)
PB  - Atlantis Press
SP  - 221
EP  - 234
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-664-7_17
DO  - 10.2991/978-94-6239-664-7_17
ID  - Akib2026
ER  -