Multi-cancer screening in peritoneal dialysis patients: an explainable machine learning framework integrating bioelectrical impedance and clinical features
5 affiliations
- 1Renal Medicine Laboratory, Department of Medical Research, Changhua Christian Hospital, Changhua, 50006 Taiwan
- 2Department of Computer Science and Information Engineering, National Cheng Kung University, No. 1, University Road, Tainan, 701401 Taiwan
- 3Department of Post-Baccalaureate Medicine, College of Medicine, National Chung Hsing University, Taichung, 402202 Taiwan
- 4School of Medicine, Chung Shan Medical University, Taichung, 40201 Taiwan
- 5Division of Nephrology, Department of Internal Medicine, Changhua Christian Hospital, No. 135, Nanxiao St, Changhua, Changhua County 50006 Taiwan
Abstract
Background: Standard cancer screening guidelines are often suboptimal for patients on maintenance dialysis due to limited life expectancy and the reduced diagnostic accuracy of conventional tumor markers. Given the disproportionately high cancer incidence in this population, there is an urgent need for specialized, non-invasive screening modalities. This study explores the potential of combining bioelectrical impedance spectroscopy (BIS) with artificial intelligence (AI) to identify malignant signatures in peritoneal dialysis (PD) patients.
Methods: This retrospective nested case-control study analyzed 329 PD patients (9.4% with cancer). Multi-frequency BIS parameters and clinical features were utilized to train four machine learning algorithms (XGBoost, random forest, SVM, and logistic regression). To prevent information leakage, recursive feature elimination and synthetic minority over-sampling technique were strictly embedded within a 10-fold repeated stratified train-test split validation framework. Model interpretability was evaluated using the SHapley Additive exPlanations framework.
Results: The XGBoost model outperformed the other algorithms, delivering a mean area under the receiver operating characteristic curve of 0.759 ± 0.071, an area under the precision-recall curve of 0.379 ± 0.120, and a high negative predictive value of 0.970 ± 0.021. Subgroup analysis showed the highest detection rate for urinary system cancers (80.0%). C-reactive protein and overhydration emerged as the most critical predictors, demonstrating a distinct synergistic interaction effect. Misclassification analysis revealed that the model was particularly adept at capturing a cancer-driven systemic inflammatory and cachectic phenotype.
Conclusion: This work establishes a non-invasive, explainable AI framework that enables opportunistic multi-cancer screening within routine PD workflows. By decoding the hidden biophysical synergy between systemic inflammation and fluid dynamics, the model serves as an efficient clinical sentinel to rule out non-cancer instances and identify aggressive malignancies without added diagnostic burden.
Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1186/s13040-026-00594-6.
Related
- MambaGLA-NPC: a 3D framework integrating CNN and state space models for robust multi-center nasopharyngeal carcinoma segmentation
- Explainable computational framework for subtype classification and biomarker identification in head and neck squamous cell carcinoma using machine learning
- Explainable AI for Network Pharmacology: A Critical Framework for Interpreting Multi-Target Drug Discovery on Biological Graphs
- An Explainable MRI-Conditioned Generative Framework for Gene-Feature Representations in Alzheimer’s Disease
- Bacterial Cytochrome C Orchestrates Larval Metamorphosis Through Bioelectrical Control of Extracellular Vesicles Biogenesis
BibTeX
@article{tsai2026multicancer,
title = {{Multi-cancer screening in peritoneal dialysis patients: an explainable machine learning framework integrating bioelectrical impedance and clinical features}},
author = {Yi-Liang Tsai and Chia-Lin Wu and Jung-Hsien Chiang},
journal = {BioData Mining},
year = {2026},
month = aug,
doi = {10.1186/s13040-026-00594-6},
url = {https://doi.org/10.1186/s13040-026-00594-6},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - JOUR
TI - Multi-cancer screening in peritoneal dialysis patients: an explainable machine learning framework integrating bioelectrical impedance and clinical features
AU - Yi-Liang Tsai
AU - Chia-Lin Wu
AU - Jung-Hsien Chiang
PY - 2026
DA - 2026/08/06
JO - BioData Mining
DO - 10.1186/s13040-026-00594-6
UR - https://doi.org/10.1186/s13040-026-00594-6
AB - Background: Standard cancer screening guidelines are often suboptimal for patients on maintenance dialysis due to limited life expectancy and the reduced diagnostic accuracy of conventional tumor markers. Given the disproportionately high cancer incidence in this population, there is an urgent need for specialized, non-invasive screening modalities. This study explores the potential of combining bioelectrical impedance spectroscopy (BIS) with artificial intelligence (AI) to identify malignant signatures in peritoneal dialysis (PD) patients. Methods: This retrospective nested case-control study analyzed 329 PD patients (9.4% with cancer). Multi-frequency BIS parameters and clinical features were utilized to train four machine learning algorithms (XGBoost, random forest, SVM, and logistic regression). To prevent information leakage, recursive feature elimination and synthetic minority over-sampling technique were strictly embedded within a 10-fold repeated stratified train-test split validation framework. Model interpretability was evaluated using the SHapley Additive exPlanations framework. Results: The XGBoost model outperformed the other algorithms, delivering a mean area under the receiver operating characteristic curve of 0.759 ± 0.071, an area under the precision-recall curve of 0.379 ± 0.120, and a high negative predictive value of 0.970 ± 0.021. Subgroup analysis showed the highest detection rate for urinary system cancers (80.0%). C-reactive protein and overhydration emerged as the most critical predictors, demonstrating a distinct synergistic interaction effect. Misclassification analysis revealed that the model was particularly adept at capturing a cancer-driven systemic inflammatory and cachectic phenotype. Conclusion: This work establishes a non-invasive, explainable AI framework that enables opportunistic multi-cancer screening within routine PD workflows. By decoding the hidden biophysical synergy between systemic inflammation and fluid dynamics, the model serves as an efficient clinical sentinel to rule out non-cancer instances and identify aggressive malignancies without added diagnostic burden. Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1186/s13040-026-00594-6.
SN - 1756-0381
ER -