# Predicting lymphovascular space invasion in early-stage cervical squamous cell carcinoma using heart rate variability

## TL;DR

This study explores using heart rate variability to predict lymphovascular space invasion in early-stage cervical cancer, offering a non-invasive alternative to biopsies.

## Contribution

The study introduces a novel logistic model using HRV parameters to predict LVSI in early-stage cervical squamous cell carcinoma.

## Key findings

- The model achieved an AUC of 0.845 in predicting LVSI status.
- Sensitivity and specificity were 0.871 and 0.750, respectively.
- The model included HRV features and clinical variables like age and BMI.

## Abstract

Accurate preoperative assessment of lymphovascular space invasion (LVSI) in patients with early-stage cervical squamous cell carcinoma (ECSCC) is clinically significant for guiding treatment decisions and predicting prognosis. However, current LVSI assessment of ECSCC mainly relies on the invasive method of pathological biopsy, which needs to be further improved in terms of convenience. The main objective of this study is to verify the value of preoperative heart rate variability (HRV) parameters in predicting ECSCC LVSI.

A total of 79 patients with ECSCC confirmed by postoperative pathology were enrolled in this study at the Department of Gynecologic Oncology of the First Affiliated Hospital of Bengbu Medical University. Patients were classified as LVSI-positive (LVSI+) or LVSI-negative (LVSI-) based on pathological examination. Preoperative 5-minute electrocardiogram (ECG) data were collected from all patients, and their HRV parameters were analysed, including 7 time-domain parameters, 5 frequency-domain parameters, and 2 nonlinear parameters. Ten HRV features were selected through univariate analysis, and a logistic model was constructed using age, body mass index, menopausal status, and mean heart rate to predict LVSI status. The model performance was evaluated by the area under the receiver operating characteristic curve (AUC), accuracy, precision, sensitivity, and specificity.

The constructed model showed good predictive performance, with an AUC of 0.845 (95% CI: 0.761 - 0.930), sensitivity of 0.871, specificity of 0.750, precision of 0.690, and accuracy of 0.747.

The Logistic model constructed based on HRV features has a relatively good diagnostic performance in predicting the LVSI status of ECSCC, but further research is still needed through larger datasets, more features, and the combination of machine learning models.