J Cancer 2021; 12(20):6050-6057. doi:10.7150/jca.61101 This issue

Research Paper

A radiomics model that predicts lymph node status in pancreatic cancer to guide clinical decision making: A retrospective study

Xiaoyuan Liang1, Wei Cai3, Xingyu Liu1, Ming Jin1, Lingxiang Ruan2✉, Sheng Yan1✉

1. Department of Hepatobiliary and Pancreatic Surgery, The Second Affiliated Hospital, School of Medicine, Zhejiang University, Hangzhou, Zhejiang 310009, China.
2. Radiology Department, The First Affiliated Hospital, School of Medicine, Zhejiang University, Hangzhou, Zhejiang 310003, China.
3. The First Affiliated Hospital of USTC, Division of Life Sciences and Medicine, University of Science and Technology of China, Hefei, Anhui, 230001, China.

This is an open access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/). See http://ivyspring.com/terms for full terms and conditions.
Citation:
Liang X, Cai W, Liu X, Jin M, Ruan L, Yan S. A radiomics model that predicts lymph node status in pancreatic cancer to guide clinical decision making: A retrospective study. J Cancer 2021; 12(20):6050-6057. doi:10.7150/jca.61101. Available from https://www.jcancer.org/v12p6050.htm

File import instruction

Abstract

Graphic abstract

Purpose: To construct a radiomics-based model for predicting lymph node (LN) metastasis status in pancreatic ductal adenocarcinoma (PDAC) before therapy and to evaluate its prognostic clinical value.

Materials and Methods: We retrospectively collected preoperative CT scans of 130 PDAC patients who underwent original tumor resection and LN dissection in the entire cohort between January 2014 and December 2017. Radiomics features were systematically extracted and analyzed from CT scans of 89 patients in the primary cohort. To construct a radiomics signature, the least absolute shrinkage and selection operator methods were employed with LN metastasis status as classification labels. Pathological analysis of LN status which were assessed by experienced pathologists was used as the evaluation label. We subjected the clinical nomogram to multivariable logistic regression analysis and conducted performance evaluation based on its discrimination, calibration, and clinical value. The model was tested and validated in 41 patients with PDAC in a separate validation cohort.

Results: Four radiomics features closely associated with LN metastasis were selected in the primary and validation cohorts (P < 0.01). Following the integration of CT-reported results and radiomics signatures into the radiomics nomogram, we reported better performance in the primary (area under the curve, 0.80) and validation (area under the curve, 0.78) cohorts.

Conclusion: The noninvasive tool constructed from the portal venous phase CT based on radiomics showed better performance for LN metastasis prediction than traditional approaches in pancreatic cancer. It may assist surgeons in crafting detailed procedures before treatment, this subsequently improves tumor staging and resection of patients.

Keywords: Pancreatic ductal adenocarcinoma, Noninvasive tool, Lymph nodes, Radiomics technology