In this study, we aimed to explore immune markers predicting locoregional recurrence/distant metastasis (R/M) for patients with esophageal squamous cell carcinoma (ESCC) post-surgical intervention by using a novel high-throughput spatial tool to quantify multiple immune proteins expressed in ESCC and lymphocytes in tumor microenvironment (TME-L). First, formalin-fixed paraffin-embedded tissues from surgical patients with ESCC (n=94) were constructed on a microarray，which were then divided into discovery (n=36) and validation cohorts (n=58). Using a newly developed GeoMx digital spatial profiling tool, 31 immune proteins in paired ESCC and TME-L, morphologically segmented by PANCK and CD45, respectively, from the discovery cohort were quantified, releasing 2232 variables. Next, the correlation matrix was analyzed using the Corrplot package in R Studio, resulting in six closely correlated clusters. The Least Absolute Shrinkage and Selection Operator regression scoring model predictive of R/M risk with superior specificity was successfully established based on three hierarchically clustered immune proteins: ARG1 in ESCC/PANCK+, STING and IDO1 in TME-L/CD45+. Moreover, the expression of IDO1 in TME-L, rather than in ESCC, significantly predicted the R/M risk score with an area under curve of 0.9598. Besides, its correlation with R/M status was further validated by dual immunohistochemistry staining of IDO1 and CD45 in discovery and validation cohorts. Above all, our findings not only provide a more accurate scoring approach based on quantitative immune proteins for the prediction of R/M risk, but also propose that IDO1 in TME-L potentially plays a driving role in mediating R/M in ESCC.