The value of machine learning based radiomics model in preoperative detection of perineural invasion in gastric cancer: a two-center study

Xujie Gao; Jingli Cui; Lingwei Wang; Qiuyan Wang; Tingting Ma; Jilong Yang; Zhaoxiang Ye

doi:10.3389/fonc.2023.1205163

The value of machine learning based radiomics model in preoperative detection of perineural invasion in gastric cancer: a two-center study

Front Oncol. 2023 Jun 14:13:1205163. doi: 10.3389/fonc.2023.1205163. eCollection 2023.

Authors

Xujie Gao^{1

2

3

4}, Jingli Cui^{1

2

3

4

5}, Lingwei Wang^{1

2

3

4}, Qiuyan Wang⁶, Tingting Ma^{1

2

3

4

7}, Jilong Yang^{2

3

4

8}, Zhaoxiang Ye^{1

2

3

4}

Affiliations

¹ Department of Radiology, Tianjin Medical University Cancer Institute and Hospital, Tianjin, China.
² Department of Radiology, National Clinical Research Center for Cancer, Tianjin, China.
³ Department of Radiology, Tianjin's Clinical Research Center for Cancer, Tianjin, China.
⁴ The Key Laboratory of Cancer Prevention and Therapy, Tianjin, China.
⁵ Department of General Surgery, Weifang People's Hospital, Weifang, Shandong, China.
⁶ Department of Radiology, Weifang People's Hospital, Weifang, Shandong, China.
⁷ Department of Radiology, Tianjin Cancer Hospital Airport Hospital, Tianjin, China.
⁸ Department of Bone and Soft Tissue Tumor, Tianjin Medical University Cancer Institute and Hospital, Tianjin, China.

Abstract

Purpose: To establish and validate a machine learning based radiomics model for detection of perineural invasion (PNI) in gastric cancer (GC).

Methods: This retrospective study included a total of 955 patients with GC selected from two centers; they were separated into training (n=603), internal testing (n=259), and external testing (n=93) sets. Radiomic features were derived from three phases of contrast-enhanced computed tomography (CECT) scan images. Seven machine learning (ML) algorithms including least absolute shrinkage and selection operator (LASSO), naïve Bayes (NB), k-nearest neighbor (KNN), decision tree (DT), logistic regression (LR), random forest (RF), eXtreme gradient boosting (XGBoost) and support vector machine (SVM) were trained for development of optimal radiomics signature. A combined model was constructed by aggregating the radiomic signatures and important clinicopathological characteristics. The predictive ability of the radiomic model was then assessed with receiver operating characteristic (ROC) and calibration curve analyses in all three sets.

Results: The PNI rates for the training, internal testing, and external testing sets were 22.1, 22.8, and 36.6%, respectively. LASSO algorithm was selected for signature establishment. The radiomics signature, consisting of 8 robust features, revealed good discrimination accuracy for the PNI in all three sets (training set: AUC = 0.86; internal testing set: AUC = 0.82; external testing set: AUC = 0.78). The risk of PNI was significantly associated with higher radiomics scores. A combined model that integrated radiomics and T stage demonstrated enhanced accuracy and excellent calibration in all three sets (training set: AUC = 0.89; internal testing set: AUC = 0.84; external testing set: AUC = 0.82).

Conclusion: The suggested radiomics model exhibited satisfactory prediction performance for the PNI in GC.

Keywords: computed tomography; gastric cancer; nomogram; perineural invasion; radiomics.

Grants and funding

This work was supported by National Key Research and Development Program of China (2021YFC2500402), National Natural Science Foundation of China (82001909, 82171932) and Tianjin Binhai New Area Health Committee Science and Technology Projects (2022BWKY017).