Extensive studies have suggested tumor immune microenvironment are of clinical importance in cancer progression, metastasis, therapeutic response (1–4). The type, density, and location of immune cells in multiple cancers had a prognostic value that was superior to and independent of those of the TNM stage (1, 5, 6). An international consortium of 14 centers in 13 countries assessed the Immunoscore assay in patients with TNM stage I-III colon cancer, and the results supported the implementation of the consensus Immunoscore as a new component of a TNM-Immune classification of cancer (1). For gastric cancer (GC), a tumor immune microenvironment (TME) signature (7) of gastric cancer based on seven features, including CD3 _{invasive margin (IM)}, CD8 _IM, CD45RO _{center of tumor (CT)}, CD66b _IM, CD34, periostin, and cyclooxygenase-2, which could be important for predicting survival and selecting appropriate patients for chemotherapy. However, the TME signature was mainly determined on postoperative tissue specimens, it is necessary to develop a noninvasive pretreatment tools for prediction of immune infiltrates.

Computational medical imaging, known as radiomics, is an emerging field that converts medical images into a high dimensional quantitative feature space using a large number of automatically extracted data-characterization algorithms (8–10). These imaging features may capture in-depth characterization of tumor distinct phenotypes, with the underlying hypothesis that imaging reflects not only macroscopic but also the cellular and molecular properties of tissues. The goal of radiomics is to develop image-driven biomarkers that serve as instruments that generate a further understanding of cancer biology to facilitate better clinical decision-making (10–12). Radiomics features are complementary to biopsies and have the advantages of being non-invasive and repeated during treatment in routine practice, contrary to genomics or proteomics, which are still challenging to apply in clinical routine (12, 13). Connecting radiomics features to the molecular biological processes active in a tumor could provide deeper information that may complement the molecular data (14). Hence, in some circumstances, the radiomics features could be apply to infer the molecular biological underpinnings of tumor in individual patients.

The association between imaging features and tumor infiltrating immune cell density has been explored (13, 15–17). Many image features extracted by radiomics, not visually observed, were closely related to specific microscopic features at the molecular level and could characterize the tumor and its tumor microenvironment (13, 18–20). Several studies also found that pre-existing tumor immune infiltration correlates with patient response to anti-programmed cell death protein (PD)-1 and anti-programmed cell death ligand 1 (PD-L1) immunotherapy (21, 22).

Gastric cancer is one of the most common cancer and leading cause of cancer death worldwide (23). Imaging with fluorine 18 (18F) fluorodeoxyglucose (FDG) positron emission tomography (PET) was used to detect of distant and lymph nodes metastases, which could stage GC accurately (24). PET/CT could reflect the tumor metabolically active. Chemotherapy have improved survival of GC patients (25, 26). However, for most of GC patients the survival rates were still limited despite initial high response rates (26, 27). Therefore, precise classification of GC that could be necessary to predict patient survival and chemotherapy responses. Radiomics could allow evaluation of a tumor and its microenvironment, thus, may lead to the identification of novel predictors for prognosis and treatment efficiency. We aimed to develop a PET/CT image based radiomics signature of tumor immune microenvironment and to assess the ability of this signature to predict survival and adjuvant chemotherapy benefits.

Radiomics approaches, when combined with tumor biopsies and genomics, could improve treatment selection. Radiomic features from tumor can provide information on both the tumor and its microenvironment (18, 36). In our study, we developed and validated a radiomic signature (RTIMS) of the tumor immune microenvironment, and the RTIMS signature was able to predict survival. Moreover, the RTIMS signature might help to identify stage II and III patients who could benefit from adjuvant chemotherapy.

The importance of radiomics is shown by the increasing number of oncological clinical trials being done that use radiomics. To date, more than 30 clinical studies are registered in ClinicalTrials.gov, including one prospective study of pembrolizumab (NCT02644369). Notably, Braman et al. (18) that evaluated radiomic features in the context of neoadjuvant chemotherapy for breast cancer found radiomics features could strongly predict pCR independent of choice of classifier, suggesting their robustness as response predictors. Moreover, pathological complete response was associated with infiltration of lymphocytes into tumors (18) Jiang et al. measure immune score based on the radiomics features from CT image (17). And, the CT image signature could predict survival and treatment response, which is concordant with our results.

Notably, several points strengthening the biological and clinical relevance of this radiomic signature have been identified. Grossmann et al. discovered a connection between the radiomic phenotype of a tumor, the signaling pathways inside cells that drive how cancer develops, and clinical treatment outcomes (14). They also did IHC staining of CD3 in 22 tumors that were predicted to show relatively high or low immune response based on one radiomic feature, and they found agreement between radiomic features and pathology. Thus, they deemed that radiomic approaches permit noninvasive assessment of both molecular and clinical characteristics of tumors, however, the results still should be interpreted with caution. Ferté et al. also developed a radiomic signature for the gene expression signature of CD8 cells, which could predict clinical outcomes in patients treated with anti-PD-1 or anti-PD-L1 immunotherapy (15).

At present, the standard treatment for advanced gastric cancer includes adjuvant chemotherapy after surgery to prevent disease recurrence and improve survival, however, many studies have reported that a subgroup of patients could not benefit from adjuvant chemotherapy (25, 26, 28). Moreover, the criteria for the selection of candidates who are more likely to benefit from adjuvant chemotherapy remain controversial. Thus, the accurate identification of subgroups of patients will improve the prognostic system and lead to more personalized therapy. Recently, several studies reported that radiomics signatures based on CT/MRI/PET images were associated with chemotherapy response in several types of cancers (37–39). Besides, Ferté et al. and Jiang et al. previous study showed that imaging biomarkers could be used to estimate tumor-infiltrating lymphocytes (15, 16), which were associated with chemotherapy response (28, 40). In this study, we found that adjuvant chemotherapy provided a better survival benefit to patients with stage II and III GC patients classified as high-RTIMS, whereas low-RTIMS patients did not obtain benefits from adjuvant chemotherapy; further use of the radiomics signature might allow for better identification of patients who are most likely to benefit from adjuvant therapy. Thus, we suggest that patients with low-RTIMS should be treated with new combinations of more tolerable medication as an adjunct to potentiate the efficacy of systemic approaches. Therefore, our PET/CT image-based RTIMS signature for patients with stage II and III gastric cancer is both a prognostic and predictive method, in that patients with a high- RTIMS have a clear benefit from adjuvant chemotherapy.

Interpretation of complex images by radio-pathology/genomic approaches is currently changing several fields in medical imaging, but clinical application of this method is still in its infancy. Validating this approach in a separate cohort, we confirmed the prognostic power of this approach. Thus, A novel biomarker that can be incorporated into existing clinical workflows because it only relies on PET/CT images, which are non-invasive and widely available. Compared to the gene-based signatures, a major advantage of our radiomics method is the ubiquitous availability of PET/CT images, which are available for every cancer patient, and analyzing them is not very costly. Besides, abdominal PET/CT scans are effective tools to diagnose and guide treatment for patients with gastric cancer. Whereas, those gene-based signatures have not been widely led into clinical application as initially expected owing to the variability of measurements in microarray and sequencing assays, inconsistencies in assay platforms, and the requirement for analytical expertise (41–43). In addition, our radiomic approach of the TME signature is reproducible: If presented with the same image twice, the algorithm will export the same result. These points make this new approach well suited for a clinical application.

According to the molecular classification of The Cancer Genome Atlas (TCGA) project, gastric cancer was divided into four subtypes: Epstein-Barr virus (EBV)-positive, microsatellite instability (MSI), genomic stability, and chromosomal instability (44). High lymphocytes infiltration were frequently observed in certain molecular subtypes of GC, such as EBV-positive and MSI subtypes (45, 46). According to a phase II study, mismatch repair deficiency or MSI-high renders different solid tumors highly sensitive to immune checkpoint blockade with the PD-1 inhibitor pembrolizumab, and these tumors contain prominent lymphocytes infiltration (47, 48). Recently, several studies showed that the tumor imaging biomarkers could provide a promising way to predict the immune phenotype of tumors and to infer clinical outcomes for patients with cancer who had been treated with PD-1 and PD-L1 inhibitor (15, 16). Therefore, future studies should investigate the association between the radiomic signature of tumor immune microenvironment and molecular classification, and explore whether the radiomic signature could predict the responses of patients with gastric cancer to immunotherapy.

There are some limitations to this study. First, the study was conducted retrospectively, which was susceptible to the inherent biases of such a study format. Secondly, the decision to accept postoperative chemotherapy or not was made by the patients and clinicians together, that was not within a randomized assignment. Thirdly, the model was developed and validated using data from East Asian patients, and its generalizability in western populations remains to be determined. Therefore, a prospective, international, multicenter clinical trial will be needed to further validate our findings.

In conclusion, we developed a PET/CT image based radiomic signature that allows the noninvasive evaluation of tumor immune microenvironment. Moreover, the RTIMS might be a useful predictive tool to identify stage II and III patients benefit from adjuvant chemotherapy. Thus, the RTIMS potentially may offer clinical value in directing individualized therapeutic regimen selection for patients with stage II and III gastric cancer.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

The studies involving human participants were reviewed and approved by Ethics Committee of Henan Provincial Hospital. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

Conception and design: JL and JB. Collection and assembly of data: JL, HW, HZ, and SL. Data analysis and interpretation: CZ, YY, HG, and PZ. Manuscript writing: all authors. Final approval of manuscript: all authors.

This study was supported by the Project of Tackle Key Problems in Science and Technology of Henan Province (No.212102310679), and the Project Construction of National Key Disciplines (General Surgery Department, 2021).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.