# The value of metabolic parameters on dynamic 18F-FDG PET/CT for predicting lymph node metastasis in non-small cell lung cancer

## TL;DR

This study shows that combining dynamic and static PET/CT scans with a blood marker improves accuracy in predicting lymph node metastasis in lung cancer.

## Contribution

The study introduces a composite model using dynamic PET/CT parameters and CA125 for improved lymph node metastasis prediction in NSCLC.

## Key findings

- TLG, slope10-30min, and CA125 were independent predictive factors for lymph node metastasis.
- The composite model outperformed PET/CT and clinical models with an AUC of 0.867.
- Combining dynamic and static PET/CT with CA125 improves N-staging accuracy in NSCLC patients.

## Abstract

To evaluate the predictive value of dynamic 18F-FDG PET/CT metabolic parameters of the primary tumor for mediastinal lymph node metastasis (LNM) in non-small cell lung cancer (NSCLC).

A total of 316 patients with clinically suspected but untreated lung lesions who underwent dynamic PET/CT and static PET/CT scans from May 2021 to November 2024 were retrospectively collected in this study. Quantitative parameters including K1, k2, k3, and Ki of each lesion, were obtained by applying the irreversible two-tissue compartmental modeling using an in-house Matlab software. Time-activity curves (TACs) at the primary tumor were extracted from each dynamic 18F-FDG PET/CT scan. The TAC signal was then decomposed into metabolism and blood flow components through kinetic modeling. Dynamic features including area under the curve (AUC), time-to-peak (tpeak), and slopes were then extracted from each component. Predictive analyses were performed using multivariate logistic regression to determine the predictive factors for LNM. Receiver-operating characteristic (ROC) analysis was performed to evaluate the predictive performance of models.

One hundred and fifteen patients who obtained LN biopsy within one month were enrolled in this study. Based on the results of the pathology, the patients were divided into LNM and non-LNM groups. The multivariate logistic regression analyses showed that the TLG, slope10-30min, and CA125 were independent predictive factors for LNM (P < 0.05, respectively). For the model comparison, composite model achieved the highest diagnostic efficacy (AUC of 0.867, sensitivity 75.5%, specificity 84.5%, accuracy 80.2%) compared with PET/CT model (AUC of 0.822, sensitivity 80.0%, specificity 72.4%, accuracy 75.7%) and clinical model (AUC of 0.792, sensitivity 49.1%, specificity 96.7%, accuracy 73.9%).

The metabolic parameters based on dynamic and static 18F-FDG PET/CT combined with CA125 can improve N-staging accuracy in patients with NSCLC.