# Predictive classification and regression models for bioimpedance vector analysis: Insights from a southern Cuban cohort

## TL;DR

This study uses predictive models to analyze bioimpedance data from a Cuban cohort, revealing insights into health and cancer-related differences.

## Contribution

The study introduces predictive models for bioimpedance vector analysis that can assess health and detect cancer-related changes.

## Key findings

- Cancer patients showed higher Zc and lower θc and Xcc values, indicating cell membrane deterioration.
- Females had higher Zc and Xcc values, suggesting better cell membrane integrity.
- Predictive models showed strong agreement with experimental values across various categories.

## Abstract

This study used predictive models to explore the link between bioparameters at characteristic frequency and their positions within tolerance ellipses in a southern Cuban cohort. The database includes 367 individuals (235 females, 132 males) aged 18–86. Among them, 61 had cancer, while 306 were healthy. After balancing the data, the analysis used 16 bioimpedance-based characteristics along with other anthropometric and location factors. The results showed that characteristic frequency bioparameters (Zc, θc, Xcc, and Rc) are key for assessing health and location. There was a strong agreement between experimental and predicted values for Zc, θc, Xcc, and Rc across various categories. Cancer patients showed higher Zc and slightly lower θc and Xcc values, attributed to unbalanced body composition and cell membrane deterioration. Females exhibited higher Zc and Xcc values, indicating better cell membrane integrity. Predictions are consistent across quartiles and percentiles, with lower θc observed in higher quartiles and centiles where more cancer patients are located. Variations in Rc values across different BIVA statuses demonstrated the model's robustness in estimating impedance parameters in diverse physiological conditions. These predictive models are significant for assigning locations without developing BIVA methods, enhancing clinical assessments and health monitoring.