ABSTRACT
Objective:To evaluate the value of radiomics in predicting the efficacy of non-operative treatment of esophageal cancer by meta-analysis.Methods:Search terms included "Esophageal Neoplasms", "Esophageal Neoplasms", "Neoplasm, Esophageal", "Esophagus Neoplasm", "Esophagus Neoplasm", "Neoplasm, Esophagus", "Neoplasms, Esophagus", "Neoplasms, Esophageal", "Cancer of Esophagus", "Cancer of the Esophagus", "Esophagus Cancer", "Cancer, Esophagus", "Cancers, Esophagus", "Esophagus Cancers", "Esophageal Cancer", "Cancer, Esophageal", "Cancers, Esophageal", "Esophageal Cancers" and "radiomics", "radiomics features", "radiomic", "texture", "texture analysis", "textural analysis", "histogram", "machine learning", "artificial intelligence", both in English and corresponding Chinese. The Chinese and English literatures related to radiomics prediction of the efficacy of non-surgical treatment of esophageal cancer published in PubMed, Web of Science, Embase, China National Knowledge Internet, Wanfang Medical Online and VIP Chinese Journal Service Platform from the establishment of the database to November 2022 were searched, and screening, quality evaluation and data extraction were carried out. Meta analysis was performed by using Stata 15.1, Meta-disc 1.4 and Review Manager 5.3 software.Results:Seventeen literatures of Chinese and English with 1389 patients with esophageal cancer who received non-surgical treatment were included. There was no significant threshold effect in predicting the effect of non-operative treatment of esophageal cancer by radiomics ( r=0.103, P=0.694), and there was high heterogeneity ( I2>50%). The combined sensitivity of all included literatures was 0.86 (95% CI: 0.81-0.89), specificity was 0.81 (95% CI: 0.76-0.85), positive likelihood ratio was 4.4 (95% CI: 3.5-5.6), and negative likelihood ratio was 0.18 (95% CI: 0.13-0.24). The diagnostic odds ratio was 25 (95% CI: 16-39) and the AUC was 0.90 (95% CI: 0.87-0.92). Conclusions:Radiomics can better predict the efficacy of non-surgical treatment of esophageal cancer, MRI and PET/CT radiomics has higher accuracy in predicting the efficacy of esophageal cancer, and machine learning can also improve the accuracy of prediction. It is helpful to make individualized treatment plan and improve the efficiency of treatment by effectively predicting the curative effect of patients with esophageal cancer before treatment.