Analyzing the Contribution of Bioimpedance and Laboratory Feature Groups to Machine Learning-Based Gallstone Disease Classification

Gallstone disease classification using structured patient data can benefit from different measurement sources, but the contribution of complete bioimpedance and laboratory feature groups beyond routine variables remains unclear. This study compared predefined measurement groups, rather than selecting individual predictors, to assess their incremental predictive value and consistency across models. The UCI Gallstone Dataset contained 319 participants and 38 predictors assigned to demographic, clinical, bioimpedance, and laboratory groups. Four scenarios combined a demographic and clinical baseline with bioimpedance, laboratory measurements, or both. Logistic Regression, Support Vector Machine, Random Forest, and XGBoost were tuned using GridSearch and Sand Cat Swarm Optimization (SCSO). Each optimizer used 100 fitness evaluations based on stratified five-fold cross-validation within the training data, followed by evaluation on the same held-out test set of 64 participants. Stability was examined across five repeated stratified partitions for GridSearch and five independent SCSO seeds on the original partition, while uncertainty was assessed using 4,000 stratified paired bootstrap resamples. Averaged across classifiers, adding bioimpedance improved held-out accuracy by 0.195 with GridSearch and 0.211 with SCSO; corresponding gains for laboratory measurements were 0.242 and 0.254. Mean ROC-AUC gains were 0.240 and 0.261 for bioimpedance and 0.303 and 0.327 for laboratory measurements. The highest held-out accuracy was 0.797 for Logistic Regression in Scenario C under both optimizers and Scenario D under SCSO; the best F1-score (0.817) and ROC-AUC (0.886) occurred in Scenario D with SCSO and GridSearch, respectively. Bootstrap intervals supported both additions to the baseline, but did not establish that laboratory gains were statistically greater than bioimpedance gains. The incremental value of bioimpedance after laboratory measurements remained uncertain. Neither optimizer consistently outperformed the other on held-out data. These findings require external validation because they derive from a small, single-center dataset.

Authors

Institutions

Publication Details

Journal
Indonesian Journal of Electronics Electromedical Engineering and Medical Informatics
Published
2026-11-16
DOI
https://doi.org/10.35882/ijeeemi.v8i4.407
Primary Topic
Artificial Intelligence in Healthcare
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Analyzing the Contribution of Bioimpedance and Laboratory Feature Groups to Machine Learning-Based Gallstone Disease Classification

Yuni Yamasari, Leo Cristian Ferry
4 links
Indonesian Journal of Electronics Electromedical Engineering and Medical Informatics
Artificial Intelligence in Healthcare
article

Analyzing the Contribution of Bioimpedance and Laboratory Feature Groups to Machine Learning-Based Gallstone Disease Classification

Yuni Yamasari, Leo Cristian Ferry
article en

Abstract

Gallstone disease classification using structured patient data can benefit from different measurement sources, but the contribution of complete bioimpedance and laboratory feature groups beyond routine variables remains unclear. This study compared predefined measurement groups, rather than selecting individual predictors, to assess their incremental predictive value and consistency across models. The UCI Gallstone Dataset contained 319 participants and 38 predictors assigned to demographic, clinical, bioimpedance, and laboratory groups. Four scenarios combined a demographic and clinical baseline with bioimpedance, laboratory measurements, or both. Logistic Regression, Support Vector Machine, Random Forest, and XGBoost were tuned using GridSearch and Sand Cat Swarm Optimization (SCSO). Each optimizer used 100 fitness evaluations based on stratified five-fold cross-validation within the training data, followed by evaluation on the same held-out test set of 64 participants. Stability was examined across five repeated stratified partitions for GridSearch and five independent SCSO seeds on the original partition, while uncertainty was assessed using 4,000 stratified paired bootstrap resamples. Averaged across classifiers, adding bioimpedance improved held-out accuracy by 0.195 with GridSearch and 0.211 with SCSO; corresponding gains for laboratory measurements were 0.242 and 0.254. Mean ROC-AUC gains were 0.240 and 0.261 for bioimpedance and 0.303 and 0.327 for laboratory measurements. The highest held-out accuracy was 0.797 for Logistic Regression in Scenario C under both optimizers and Scenario D under SCSO; the best F1-score (0.817) and ROC-AUC (0.886) occurred in Scenario D with SCSO and GridSearch, respectively. Bootstrap intervals supported both additions to the baseline, but did not establish that laboratory gains were statistically greater than bioimpedance gains. The incremental value of bioimpedance after laboratory measurements remained uncertain. Neither optimizer consistently outperformed the other on held-out data. These findings require external validation because they derive from a small, single-center dataset.

Indonesian Journal of Electronics Electromedical Engineering and Medical InformaticsVol. 8(4)
Universitas Negeri Surabaya (ID)
Good health and well-being
Artificial Intelligence in Healthcare
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon