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

Area of Science:

• Biophysics
• Medical Diagnostics
• Health Informatics

Background:

• Bioimpedance analysis (BIVA) is used to assess body composition and cell membrane status.
• Characteristic frequency bioparameters (Zc, θc, Xcc, Rc) offer insights into physiological conditions.
• Predictive modeling can enhance the interpretation of BIVA data.

Purpose of the Study:

• To explore the relationship between characteristic frequency bioparameters and their positions within tolerance ellipses.
• To develop and validate predictive models for assessing health status and location using bioimpedance data.
• To investigate differences in bioparameters between cancer patients and healthy individuals.

Main Methods:

• Utilized a database of 367 individuals (61 cancer, 306 healthy) from a southern Cuban cohort.
• Employed predictive models analyzing 16 bioimpedance-derived characteristics, anthropometric data, and location factors.
• Balanced data and validated model predictions against experimental values for Zc, θc, Xcc, and Rc.

Main Results:

• Characteristic frequency bioparameters (Zc, θc, Xcc, Rc) proved crucial for health and location assessment.
• High agreement was observed between experimental and predicted impedance values.
• Cancer patients exhibited higher Zc and lower θc and Xcc values, linked to body composition and cell membrane changes.
• Females showed higher Zc and Xcc, suggesting better cell membrane integrity.
• Predictive models demonstrated consistency across data quartiles and percentiles, identifying trends related to cancer prevalence.

Conclusions:

• Predictive models accurately estimate impedance parameters, offering a robust tool for clinical assessment.
• Characteristic frequency bioparameters are valuable biomarkers for distinguishing health states and identifying individuals at risk.
• These models facilitate health monitoring and clinical assessments, potentially without requiring traditional BIVA methods.