CT-Based Deep Learning Radiomics Analysis for Evaluation of Serosa Invasion in Advanced Gastric Cancer

Abstract

Background: This work aimed to develop and validate a deep learning radiomics model for evaluating serosa invasion in gastric cancer. Methods: First we retrospective enrolled 428 patients (252 in the training set and 176 in the test set I) with pathological confirmed T3 or T4a gastric cancer. Subsequently, a total of 144 consecutive patients who were clinically diagnosed cT3 or cT4a were allocated to the test set II. The contrast enhanced CT images of three phases were manually segmented. Conventional hand-crafted features and deep learning features were extracted based on CT images automatically and were utilized to build radiomics signatures via machine learning methods. Multivariable logistic regression analysis was used to develop a diagnostic model (radiomics nomogram) incorporating the radiomics signatures and subjective CT findings. Its diagnostic ability was measured using receiver operating characteristic curve analysis. Findings: The three radiomics signatures were built with support vector machine or random forest, and showed good performance for discriminating T4a in the test I and II sets with area under curves (AUCs) of 0.728-0.800 and 0.786-0.847. The nomogram had powerful diagnostic ability in all training, test I and II sets with AUCs of 0.891 (95% CI, 0.853-0.930), 0.868 (95% CI, 0.813-0.923) and 0.880 (95% CI, 0.823-0.936) respectively. The net reclassification index revealed that the radiomics nomogram had significantly better performance than the clinical characteristics-based model (p-values < 0.05). Interpretation: The deep learning radiomics model based on CT images is effective at discriminating serosa invasion in gastric cancer. Funding: This work was funded by the Beijing Municipal Administration of Hospitals Clinical Medicine Development of Special Funding Support (ZYLX201803), Beijing Natural Science Foundation (7172049, L182061), National Key R&D Program of China (2017YFC1308700, 2017YFA0205200, 2017YFC1309100), National Natural Science Foundation of China (81771924, 81501616, 81227901), the Bureau of International Cooperation of Chinese Academy of Sciences (173211KYSB20160053), and the Youth Innovation Promotion Association CAS (2017175). Declaration of Interest: The authors declare no conflicts of interest. Ethical Approval: The phase I retrospective study was approved by the institutional review board of our institution, and the requirement of informed consent was waived. The phase II validation study was approved by our institution review board, and informed consent was obtained from all patients.