Editorial Open Access
Copyright ©The Author(s) 2025. Published by Baishideng Publishing Group Inc. All rights reserved.
World J Gastrointest Oncol. Jul 15, 2025; 17(7): 108175
Published online Jul 15, 2025. doi: 10.4251/wjgo.v17.i7.108175
Deep learning radiomics: Redefining precision oncology through noninvasive insights into the tumor immune microenvironment
Mesut Tez, Department of Surgery, University of Health Sciences, Ankara City Hospital, Ankara 06800, Türkiye
ORCID number: Mesut Tez (0000-0001-5282-9492).
Author contributions: Tez M wrote and finalized the manuscript.
Conflict-of-interest statement: All the authors report no relevant conflicts of interest for this article.
Open Access: This article is an open-access article that was selected by an in-house editor and fully peer-reviewed by external reviewers. It is distributed in accordance with the Creative Commons Attribution NonCommercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: https://creativecommons.org/Licenses/by-nc/4.0/
Corresponding author: Mesut Tez, Department of Surgery, University of Health Sciences, Ankara City Hospital, No. 1 Bilkent Street, District of Universities, Ankara 06800, Türkiye. mesuttez@yahoo.com
Received: April 7, 2025
Revised: April 17, 2025
Accepted: May 13, 2025
Published online: July 15, 2025
Processing time: 98 Days and 8.6 Hours

Abstract

Computed tomography-based deep learning radiomics provides a novel, noninvasive approach to predicting the tumor immune microenvironment in colorectal cancer, revolutionizing precision oncology. The retrospective study by Zhou et al analyzed preoperative computed tomography scans from 315 patients using convolutional neural networks, achieving robust predictive performance (area under the curve: 0.851-0.892) for critical tumor immune microenvironment features, such as tumor-stroma ratio and lymphocyte infiltration, without requiring invasive biopsies. This editorial explores how this technique advances personalized immunotherapy, chemotherapy, and targeted therapies; challenges conventional oncology practices; and paves the way for a future of precision medicine. By integrating advanced imaging with immune profiling, deep learning radiomics redefines colorectal cancer management, highlighting the need to re-evaluate the interplay of technology, biology, and ethics in gastrointestinal oncology.

Key Words: Colorectal cancer; Radiomics; Tumor immune microenvironment; Therapy; Immunotherapy

Core Tip: Computed tomography-based deep learning radiomics provides a noninvasive, scalable approach to predict the tumor immune microenvironment in colorectal cancer, achieving high accuracy (area under the curve: 0.851-0.892) in the study by Zhou et al. By overcoming biopsy limitations, the approach revolutionizes personalized immunotherapy, chemotherapy, and targeted therapies, integrating advanced imaging with immune profiling to redefine precision oncology and improve patient outcomes.



INTRODUCTION

Cancer is a complex ecosystem driven by the tumor microenvironment (TME), extending beyond genetic alterations. The TME encompasses cancer cells, immune cells, cancer-associated fibroblasts, endothelial cells, pericytes, and the extracellular matrix, with dynamic interactions shaping tumor initiation, progression, and therapeutic response[1]. The tumor immune microenvironment (TIME), a critical TME subset, includes cytokines, cytotoxic T cells, and immunosuppressive elements like regulatory T cells and myeloid-derived suppressor cells, exhibiting spatial and temporal heterogeneity[2]. These immune components modulate antitumor immunity and treatment efficacy, as evidenced by studies highlighting their pivotal role in colorectal cancer (CRC) progression and response to therapies such as immune checkpoint inhibitors[2].

CRC, a leading gastrointestinal malignancy, is significantly influenced by the TIME, which drives tumor development, progression, and metastasis. Angiogenesis, regulatory T cells, myeloid-derived suppressor cells, and cytokine-mediated modulation within the TIME suppress antitumor immunity, promoting tumor growth[3]. Despite infiltration by effector memory lymphocytes, immunotherapy advances in CRC remain limited, underscoring the need for precise TIME characterization to optimize treatments[3].

Conventional TIME assessments rely on invasive, costly, and spatially limited biopsies[4]. The groundbreaking study by Zhou et al[5] introduces computed tomography (CT)-based deep learning radiomics to noninvasively predict TIME features, tumor-stroma ratios, lymphocyte infiltration, and immune scoring, with high accuracy (area under the curve: 0.851-0.892). Leveraging convolutional neural networks, the study analyzed preoperative CT scans from 315 CRC patients, extracting radiomic features such as first-order statistics (e.g., intensity, mean), shape-based metrics (e.g., volume), and texture-based patterns (e.g., gray-level co-occurrence matrix). These features capture tumor heterogeneity and immune microenvironment dynamics, validated through calibration and decision curve analyses[5]. By eliminating the need for invasive procedures, this approach surpasses biopsy limitations, enabling preoperative personalization of immunotherapy and improving outcomes[4,5].

Compared to other noninvasive modalities, CT-based radiomics benefits from widespread availability, cost-effectiveness, and high spatial resolution, making it a practical choice for TIME prediction in CRC. Magnetic resonance imaging radiomics offers superior soft-tissue contrast, potentially improving stromal and immune cell characterization, but its limited accessibility and longer scan times pose challenges. Positron emission tomography imaging excels in metabolic profiling but is constrained by radiation exposure and lower resolution. These modalities may complement the CT-based approach, but the scalability thereof positions it as a leading tool to democratize precision oncology[4].

CLINICAL APPLICATIONS AND CHALLENGES

Beyond immunotherapy, deep learning radiomics holds promise for broader cancer management. By predicting chemosensitivity, it could guide chemotherapy dosing, while profiling molecular pathways may optimize targeted therapies. Longitudinal radiomic analysis could monitor the tumor response or detect early recurrence, enhancing proactive care[4,5]. These applications underscore the potential of the technology to transform treatment decision-making across oncology, extending its impact beyond TIME prediction.

Implementing deep learning radiomics in clinical practice faces practical hurdles, particularly in resource-limited settings where advanced CT scanners and computational infrastructure may be scarce. Cloud-based radiomics platforms could mitigate these barriers by enabling remote data processing. Clinicians may require 6-12 months of training to interpret radiomic features, though automated decision-support tools could streamline this process. Addressing these challenges through scalable infrastructure and user-friendly interfaces is critical to ensuring broad adoption[4].

Ethical concerns in deep learning radiomics include safeguarding patient privacy and ensuring algorithmic fairness. Anonymizing imaging data and complying with regulations like the General Data Protection Regulation and Health Insurance Portability and Accountability Act are essential to protect patient confidentiality. Secure data storage and encryption further mitigate risks. To ensure transparency and fairness, explainable artificial intelligence frameworks should be adopted to clarify algorithmic decision-making, reducing biases and fostering trust in clinical applications. These ethical safeguards are critical as radiomics advances[4].

CONCLUSION

This innovation challenges traditional gastrointestinal malignancy management. Does it merely enhance existing methods, or does it redefine cancer as a dynamic, relational process? Its potential to democratize precision medicine by broadening access to advanced diagnostics is evident[4-6]. However, challenges such as standardization, model transparency, and interdisciplinary collaboration must be addressed. As gastrointestinal oncology stands now, the study by Zhou et al[5] calls for a collaborative effort among researchers, clinicians, and ethicists to harness noninvasive tools and reshape the future of cancer care.

Footnotes

Provenance and peer review: Invited article; Externally peer reviewed.

Peer-review model: Single blind

Specialty type: Oncology

Country of origin: Türkiye

Peer-review report’s classification

Scientific Quality: Grade B

Novelty: Grade B

Creativity or Innovation: Grade A

Scientific Significance: Grade A

P-Reviewer: Xu SS S-Editor: Wu S L-Editor: Filipodia P-Editor: Zhao S

References
1.  de Visser KE, Joyce JA. The evolving tumor microenvironment: From cancer initiation to metastatic outgrowth. Cancer Cell. 2023;41:374-403.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in RCA: 1449]  [Reference Citation Analysis (0)]
2.  Binnewies M, Roberts EW, Kersten K, Chan V, Fearon DF, Merad M, Coussens LM, Gabrilovich DI, Ostrand-Rosenberg S, Hedrick CC, Vonderheide RH, Pittet MJ, Jain RK, Zou W, Howcroft TK, Woodhouse EC, Weinberg RA, Krummel MF. Understanding the tumor immune microenvironment (TIME) for effective therapy. Nat Med. 2018;24:541-550.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 2019]  [Cited by in RCA: 3761]  [Article Influence: 537.3]  [Reference Citation Analysis (0)]
3.  Khan MM, Kalim UU, Khan MH, Lahesmaa R. PP2A and Its Inhibitors in Helper T-Cell Differentiation and Autoimmunity. Front Immunol. 2021;12:786857.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 12]  [Cited by in RCA: 12]  [Article Influence: 4.0]  [Reference Citation Analysis (0)]
4.  Bernatowicz K, Amat R, Prior O, Frigola J, Ligero M, Grussu F, Zatse C, Serna G, Nuciforo P, Toledo R, Escobar M, Garralda E, Felip E, Perez-Lopez R. Radiomics signature for dynamic monitoring of tumor inflamed microenvironment and immunotherapy response prediction. J Immunother Cancer. 2025;13:e009140.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Cited by in Crossref: 2]  [Reference Citation Analysis (0)]
5.  Zhou C, Zhang YF, Yang ZJ, Huang YQ, Da MX. Computed tomography-based deep learning for preoperative prediction of tumor immune microenvironment in colorectal cancer. World J Gastrointest Oncol. 2025;17:106103.  [PubMed]  [DOI]
6.  Jia Q, Wang A, Yuan Y, Zhu B, Long H. Heterogeneity of the tumor immune microenvironment and its clinical relevance. Exp Hematol Oncol. 2022;11:24.  [RCA]  [PubMed]  [DOI]  [Full Text]  [Full Text (PDF)]  [Cited by in Crossref: 4]  [Cited by in RCA: 124]  [Article Influence: 41.3]  [Reference Citation Analysis (0)]