Ultrasound-Based Radiomics for Preoperative Lymph Node Metastasis Prediction in Pancreatic Ductal Adenocarcinoma
Researchers developed and compared ultrasound-based intratumoral, peritumoral, clinical, and combined models for lymph node metastasis prediction. The study explored the complementary value of multi-regional imaging in p
Accurate preoperative assessment of lymph node metastasis (LNM) in pancreatic ductal adenocarcinoma (PDAC) remains challenging due to the heterogeneity of tumor biology and the difficulty in distinguishing benign from malignant nodes. This study addresses this gap by evaluating ultrasound-based imaging features to improve diagnostic accuracy before surgical intervention, which is critical for treatment planning and patient outcomes.
The authors retrospectively enrolled ninety-nine patients with pathologically confirmed PDAC who underwent preoperative ultrasound. Intratumoral and 3-mm peritumoral regions of interest (ROIs) were manually delineated on imaging scans. Radiomics features were extracted using PyRadiomics software, selected via reproducibility filtering, correlation analysis, and LASSO regression. Nine machine-learning algorithms were evaluated to identify the optimal classifier for each region, with a decision-level combined model constructed by integrating regional outputs with clinical information.
The intratumoral and peritumoral models achieved area under the curve (AUC) values of 0.815 and 0.792, respectively. The combined model yielded the highest performance (AUC = 0.898, 95% CI: 0.770–1.000) with good calibration and the greatest net benefit on decision curve analysis. DeLong tests showed no statistically significant AUC differences among models, while intratumoral and peritumoral outputs were moderately correlated (Pearson's r = 0.711).
These findings suggest that combining ultrasound-based radiomics with clinical data enhances predictive capability for LNM in PDAC. However, the study was retrospective and limited to a single institution, which may affect generalizability. Additionally, the authors did not validate these models on an independent cohort or assess long-term outcomes.
This research highlights the potential of non-invasive imaging techniques to support surgical decision-making in pancreatic cancer management. While promising, further validation is required before clinical implementation. The study emphasizes the importance of multi-regional analysis and machine learning integration in radiomics for improving diagnostic precision.