A study on screening of postoperative chemotherapy beneficial populations in stage II colorectal cancer based on diffusion weighted imaging and ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography

Introduction

Background on stage II colorectal cancer (CRC) risk stratification

CRC is one of the most common malignant tumors and the fifth leading cause of cancer-related death worldwide (1,2). In recent years, its incidence and mortality have shown an upward trend in China, posing a serious threat to patients’ quality of life, health, and survival. CRC exhibits obvious heterogeneity, and the formulation of treatment strategies and prognosis are related to a variety of clinicopathological factors, including histopathological type, tumor-node-metastasis (TNM) stage, grade, differentiation, and tumor location (3,4). According to the latest version of the Chinese Society of Clinical Oncology (CSCO) guidelines (5) and several landmark studies (6,7), all patients with high-risk stage II disease should receive adjuvant chemotherapy with the CAPEOX regimen for 3 months after surgery, whereas follow-up is recommended for patients with low-risk stage II and stage I disease after surgery. Low-risk stage II is defined as T3N0M0 with deficient mismatch repair (dMMR), regardless of the presence of high-risk factors. High-risk stage II includes T3N0M0 with proficient mismatch repair (pMMR) accompanied by high-risk factors, or T4N0M0. High-risk factors include T4 stage, poorly differentiated histology [high grade, excluding microsatellite instability-high (MSI-H)], lymphovascular invasion, perineural invasion, preoperative intestinal obstruction or tumor perforation, positive or uncertain surgical margins, insufficient clearance margin distance, and fewer than 12 harvested lymph nodes. Therefore, the identification of high-risk status in patients with stage II CRC after surgery is crucial. Nevertheless, current risk stratification strategies for stage II CRC still have several inherent limitations. Traditional high-risk judgment depends entirely on postoperative pathological indicators, which can only be obtained after surgical resection. Pathological evaluation is cumbersome and time-consuming, and local sampling cannot completely represent the overall biological characteristics of the tumor, resulting in inevitable heterogeneity bias (8). More importantly, most existing studies only focus on prognostic risk stratification rather than screening for chemotherapy-beneficial populations, which cannot effectively guide individualized adjuvant treatment decision-making (9). In contrast, preoperative imaging examination can non-invasively and comprehensively reflect the macroscopic and quantitative features of the entire tumor, partially overcoming the limitations of postoperative single-point pathological evaluation (10). Imaging-based quantitative analysis is low-cost, repeatable, and widely applicable in clinical practice, making it an ideal tool for preoperative individualized risk assessment (11).

Rationale for imaging modalities and biologic markers

As the most commonly used and mature functional imaging technique in magnetic resonance imaging (MRI), diffusion-weighted imaging (DWI) has been applied in multiple systems to distinguish benign from malignant tumors, evaluate tumor grade, differentiation, and chemotherapeutic efficacy, and assess the expression of immunohistochemical markers (12-17). ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT), as a mainstream technique in nuclear medicine, fuses functional information from PET with anatomical information from CT. ¹⁸F-FDG, as a tracer, reflects glucose uptake. Tumor cells are usually metabolically active with increased glucose uptake, showing hypermetabolic foci. The standardized uptake value (SUV), as a semi-quantitative parameter of ¹⁸F-FDG PET/CT, can reflect the malignant degree and metabolic activity of tumors. ¹⁸F-FDG PET/CT shows certain advantages in the early diagnosis, differential diagnosis, staging, differentiation, evaluation of chemotherapeutic efficacy, and assessment of immunohistochemical expression in tumors (18-23). Given artificial intelligence (AI) including deep learning (DL), machine learning (ML), large language model (LLM), etc. is aggressively entering in cancer study (24-26), However, these methods suffer from complex feature extraction, poor repeatability, and low interpretability, which hinders their rapid translation into clinical practice (27). MRI and ¹⁸F-FDG PET/CT are often performed preoperatively in patients with CRC, although they focus on different aspects. However, whether these two imaging techniques can complement each other in evaluating the benefit of postoperative chemotherapy in CRC patients remains unclear.

Study hypothesis/objective

Therefore, this study aims to investigate the value of DWI and ¹⁸F-FDG PET/CT in screening patients with CRC who may benefit from postoperative chemotherapy. We present this article in accordance with the STARD reporting checklist (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0384/rc).

Methods

General clinicopathological data

This study was designed as a retrospective analysis, which was approved by the Institutional Ethics Committee of Zhuhai People’s Hospital (No. 2023-KT-33), and informed consent was waived because it was a retrospective analysis of pre-existing de-identified data. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Eligible patients who underwent radical resection for CRC at Zhuhai People’s Hospital (The Affiliated Hospital of Beijing Institute of Technology, Zhuhai Clinical Medical College of Jinan University) from February 2019 to February 2026 were retrospectively enrolled in the research.

Inclusion criteria were defined as follows: (I) pathologically confirmed CRC after initial radical resection; (II) preoperative MRI and ¹⁸F-FDG PET/CT examinations both completed in Zhuhai People’s Hospital (The Affiliated Hospital of Beijing Institute of Technology, Zhuhai Clinical Medical College of Jinan University) within one month before surgery; (III) complete clinical and follow-up data available for analysis.

Exclusion criteria included: (I) history of local or systemic anti-tumor therapy (radiotherapy, chemotherapy, biotherapy, etc.) before imaging examinations; (II) suboptimal bowel preparation or inadequate visualization of lesions on imaging images; (III) incomplete postoperative pathological information leading to failure of accurate staging and grouping.

All enrolled patients with stage II CRC were stratified into high-risk and non-high-risk groups strictly according to the 2025 CSCO Colorectal Cancer Guidelines. In detail, the non-high-risk group was defined as patients with T3N0M0 tumors and dMMR status, regardless of the presence of conventional pathological high-risk factors. The high-risk group included two categories of patients: (I) T3N0M0 tumors with pMMR status accompanied by at least one pathological high-risk factor; and (II) patients with T4N0M0 disease. The high-risk factors defined by the CSCO guideline included poorly differentiated histology (high grade, excluding MSI-H tumors), lymphovascular invasion, perineural invasion, preoperative intestinal obstruction or tumor perforation, positive or uncertain surgical resection margins, insufficient resection margin distance, and retrieval of fewer than 12 regional lymph nodes. Notably, the risk stratification in the present study relied entirely on standardized clinicopathological and molecular indicators recommended by the official guideline, without incorporating any imaging features, self-built scoring models, or artificial weighted variables. Consistent guideline-based grouping criteria were applied to all patients to ensure objectivity, transparency, and reproducibility of patient grouping.

Finally, a total of 106 eligible patients were enrolled, among whom 47 were stratified to the high-risk stage II CRC group and 59 to the non-high-risk stage II CRC group (including low-risk stage II cases). The flowchart of patient selection is shown in Figure 1.

Figure 1 Flowchart of patient selection. CRC, colorectal cancer; DWI, diffusion-weighted imaging; FDG, fluorodeoxyglucose; PET-CT, positron emission tomography-computed tomography.

Clinical data within one week before surgery were extracted from the Hospital Information System (HIS) of Zhuhai People’s Hospital (The Affiliated Hospital of Beijing Institute of Technology, Zhuhai Clinical Medical College of Jinan University), including age, gender, serum levels of CEA and carbohydrate antigen 199 (CA199), and cancer family history. The definitions of clinical risk factors were standardized as follows:

Definitions: alcohol consumption history: continuous drinking for 5 years with a daily alcohol intake of ≥40 g in men or ≥20 g in women; or heavy drinking with a daily alcohol intake of ≥80 g within two weeks. Smoking history: A cumulative smoking index (cigarettes per day × years of smoking) of ≥400. Hypertension history: definite clinical diagnosis of hypertension, regardless of the current blood pressure control status. Diabetes history: definite clinical diagnosis of hyperglycemia, regardless of the current blood glucose control status. Family history of cancer: At least one immediate family member diagnosed with malignant tumors.

Postoperative pathological reports were retrospectively collected to extract pathological parameters, including maximum tumor diameter, tumor type (protruding, infiltrative, ulcerative), differentiation degree (well, moderately, poorly differentiated), number of harvested lymph nodes and lymph node metastases, as well as the presence or absence of lymphovascular invasion and perineural invasion.

MRI examination

All MRI scans were performed on a Signa HDxt 3.0T MR scanner (GE Healthcare, USA) with the following sequences and parameters set for axial scanning:

• Axial T1WI sequence: repetition time (TR)/echo time (TE) =5.4/2.5 ms, matrix =256×200, field of view (FOV) =380 mm × 340 mm, slice thickness =5 mm, slice interval =1.0 mm, scanning duration =1 min 16 s;
• Axial T2WI sequence: TR/TE =13,043/71 ms, matrix =320×320, FOV =380 mm × 380 mm, slice thickness =5 mm, slice interval =1.0 mm, scanning duration =2 min 37 s;
• Axial DWI sequence: TR/TE =5,455/82 ms, matrix =128×128, b values =0 and 1,500 s/mm², FOV =380 mm × 320 mm, slice thickness =5 mm, slice interval =1.0 mm, scanning duration =2 min 22 s;
• Axial and sagittal liver acquisition with volume acceleration (LAVA) dynamic enhanced scan: TR/TE =5.4/1.9 ms, flip angle =15°, matrix =256×200, FOV =380 mm × 350 mm, slice thickness =4.0 mm, slice interval =2.0 mm, acquisition time =3 min 16 s.

Magnevist (Bayer Medical, Guangzhou, China) was intravenously injected via the elbow vein using a double syringe at a dose of 0.1 mL/kg body weight and an injection rate of 2–3 mL/s for the enhanced scan.

¹⁸F-FDG PET/CT examination

¹⁸F-FDG PET/CT scans were completed on a standardized PET/CT scanner in Zhuhai People’s Hospital (The Affiliated Hospital of Beijing Institute of Technology, Zhuhai Clinical Medical College of Jinan University), with the imaging agent ¹⁸F-FDG supplied by Guangzhou Atomic High-tech Co., Ltd. (radiochemical purity >95%). All subjects were required to fast for more than 6 hours before scanning, with blood glucose controlled below 8.0 mmol/L. ¹⁸F-FDG was intravenously injected at a dose of 0.12 mCi/kg body weight, followed by whole-body image acquisition from the top of the cranium to the bilateral upper femurs after the recommended uptake time. The acquisition parameters of PET/CT were set as follows: 6–8 scanning beds with an acquisition time of 3 min per bed, FOV =600 mm, matrix =512×512, slice thickness =2 mm, and 3D acquisition mode. A non-contrast CT scan was first performed for attenuation correction and anatomical localization, with CT parameters: 140 kVp, 200 mAs, slice thickness =2 mm. PET data were reconstructed with attenuation correction and iterative reconstruction algorithm, and fused PET/CT images were generated for subsequent analysis.

Image processing and data analysis

Axial DWI and ¹⁸F-FDG PET/CT images were imported into the ADW4.5 workstation (GE Healthcare) for data measurement, which was independently completed by two senior radiologists 6 or more years of clinical experience in oncological imaging, under the blind condition of unknown pathological results and grouping information. The measurement method was standardized as follows: the axial slice with the largest tumor diameter and its two adjacent slices (upper and lower) were selected for region of interest (ROI) delineation. The ROI was placed in the solid component of the CRC lesion, accounting for approximately two-thirds of the axial area of the solid component with a minimum area of 10 mm². The edge of the ROI was at least 2 mm away from the tumor margin to avoid hemorrhagic, necrotic areas and intestinal contents (Figure 2A,2B). To ensure the consistency of ROI placement, the ROIs on ¹⁸F-FDG PET/CT and DWI images were delineated on the same anatomical plane with the same number and size (Figure 2C,2D). For the imperfect image registration caused by intestinal peristalsis, the radiologists finely adjusted the ROI position according to the specific anatomical location of the CRC lesion. The average value of the three slices was calculated for each parameter, and the mean value of the measurements from the two radiologists was used for subsequent statistical analysis.

Figure 2 A 51-year-old woman with CRC. The high b value (b =2,000) diffusion weighted image (A) shows high signal intensity in CRC with corresponding low signal in the ADC maps (B). ¹⁸F-FDG PET (C) and fused PET/DWI (D) images show strong ¹⁸F-FDG uptake. The delineation of ROI is achieved through the integration of different machines. ADC, apparent diffusion coefficient; CRC, colorectal cancer; DWI, diffusion-weighted imaging; FDG, fluorodeoxyglucose; PET, positron emission tomography; ROI, region of interest.

Statistical analysis

SPSS 22.0, MedCalc V20.0.3 and GraphPad Prism 9 software were used for all statistical analyses. Categorical variables were described as frequencies and percentages, and compared using the chi-square test. The Shapiro-Wilk (S-W) test was applied to assess the normality of continuous variables, which were expressed as mean ± standard deviation for normally distributed data and median (interquartile range) for non-normally distributed data. Intraclass correlation coefficient (ICC) was used to evaluate the inter-observer consistency of parameter measurements between the two radiologists, with the consistency level defined as: poor (ICC <0.40), moderate (0.40≤ ICC <0.75), and good (ICC ≥0.75). Independent-samples t-test (for normally distributed data) or Mann-Whitney U test (for non-normally distributed data) was used to compare the differences in continuous variables between the high-risk and non-high-risk groups. Binary logistic regression analysis was performed to construct a combined diagnostic model integrating apparent diffusion coefficient (ADC) value, SUVmax and CEA level. ROC curve analysis was used to evaluate the diagnostic efficacy of each parameter and the combined model, with the area under the curve (AUC) calculated. The optimal cutoff value, corresponding sensitivity and specificity were determined based on the Youden index. The Delong test was used to compare the differences in AUC values between the combined model and single parameters. Pearson correlation analysis (for normally distributed data) or Spearman correlation analysis (for non-normally distributed data) was used to explore the correlation between ADC value and SUVmax, with the correlation strength classified as: no/very weak (0≤|r|<0.20), weak (0.20≤|r|<0.40), moderate (0.40≤|r|<0.60), strong (0.60≤|r|<0.80), and extremely strong (0.80≤|r|≤1). A two-sided P value <0.05 was considered statistically significant for all analyses.

Results

General clinicopathological data

A total of 106 CRC patients were included in this study (47 patients in the high-risk stage II of CRC; 59 patients in the non-high-risk stage II) as shown in (Table 1).

Consistency test of the measurement results of two doctors’ parameters

The two doctors analyzed the lesion ADC value and SUVmax of the high risk/non-high risk stage II group. The ICC values were all greater than 0.80.

Comparison of parameters between the two groups

The ADC value of the high-risk group was lower than that of the non-high-risk group (P<0.05), and the SUVmax of the high-risk group was higher than that of the non-high-risk group (P<0.05) as shown in (Table 2).

Diagnostic efficiency evaluation of each parameter value

The AUC, 95% confidence interval (CI), threshold, sensitivity, and specificity were identified by each and combined parameter in the high risk/non-high risk groups, as shown in (Table 3). ROC curves are shown in Figure 3. The comparison of single parameter values and the AUC of the combined model is shown in Table 4 and Figure 3.

Figure 3 The ROC curves of the CEA, DWI, SUVmax, and CEA + DWI + SUVmax. CEA, carcinoembryonic antigen; DWI, diffusion-weighted imaging; ROC, receiver operating characteristic; SUVmax, maximum standardized uptake value.

Correlation analysis of SUVmax, and ADC values

There was a strong negative correlation between the SUVmax and ADC value in the high-risk group (r=−0.638) and non-high-risk group (r=−0.777), as shown in Figure 4.

Figure 4 Correlation analysis chart of ADC with SUVmax values in the two groups. ADC, apparent diffusion coefficient; SUVmax, maximum standardized uptake value.

Discussion

The results of the present study demonstrated that ADC values, SUVmax, and CEA could quantitatively evaluate the high-risk status of stage II CRC. Individual comparison of the two techniques alone or with CEA did not yield clinical benefit for patients. However, the combination of the two techniques and CEA was superior to any single technique and provided additional benefit for patients who would gain from chemotherapy. Furthermore, we found a strong negative correlation between ADC and SUVmax in the high-risk versus non-high-risk groups.

This study has several distinct strengths. First, the risk stratification of stage II CRC was strictly implemented in accordance with the latest 2025 CSCO clinical guidelines, with unified and standardized grouping criteria. This guideline-based classification strategy ensures good clinical consistency, reproducibility, and practical applicability of our subgroup analysis. Second, different from conventional risk evaluation that relies merely on postoperative pathological indicators, the present study incorporated preoperative quantitative imaging parameters to construct the predictive model. Imaging examination can comprehensively reflect the overall tumor characteristics, thereby compensating for the inherent limitations of postoperative pathological detection, including delayed acquisition and potential sampling bias caused by tumor heterogeneity. With the advantages of non-invasiveness, low cost, and repeatability, imaging-based prediction tools are more feasible for preoperative individualized clinical decision-making.

Serum CEA is an important biomarker for the diagnosis, prognosis, recurrence, metastasis monitoring, and evaluation of chemotherapy efficacy in CRC (28-30). As a serological tumor marker for CRC, CEA has been widely and commonly used in clinical practice. Higher serum CEA levels indicate more active tumors and may predict a poorer prognosis. In the present study, we found that CRC with elevated CEA was more frequently observed in patients with high-risk stage II disease, which may be related to the prognostic differences between high-risk and non-high-risk stage II CRC patients. Sonoda et al. (31) analyzed pathologically confirmed CRC patients and found that stage II/III CRC patients with elevated serum CEA levels often had a poor prognosis. Overall, the prognostic value of CEA in CRC has been well established, which was also confirmed in the present study.

Identification of high-risk status in stage II CRC is an important basis for guiding postoperative chemotherapy in clinical practice. Liu et al. (32) developed a more accurate risk stratification system based on a large sample size, integrating artificial intelligence-based CT radiomic analysis with pathological markers to optimize individualized therapeutic strategies for patients with stage II CRC. Meanwhile, their team also proposed multimodal data integration of bio-related artificial intelligence for guiding adjuvant chemotherapy in stage II CRC (33). Zhang et al. (34) found that a CT radiomics-based model could provide benefit for postoperative chemotherapy in stage II CRC. Ye et al. (35) demonstrated that the tumor enhancement ratio (TER) derived from contrast-enhanced CT could serve as a high-risk feature of stage II colon cancer to predict tumor recurrence and guide the formulation of clinical chemotherapy regimens. All the above studies confirmed that imaging-based methods can evaluate the high-risk status of stage II CRC and provide feasible clinical schemes for selecting postoperative chemotherapy.

The results of the present study yielded similar findings, indicating that DWI and ¹⁸F-FDG PET/CT can quantitatively assess the high-risk status of stage II CRC. The underlying mechanism may be that high-risk CRC is usually associated with higher malignancy, resulting in decreased ADC values and elevated SUVmax. Furthermore, using regression analysis, we found that the combined diagnosis integrating the two imaging techniques and CEA could bring clinical benefit to patients and is worthy of clinical promotion.

Two meta-analyses (36,37) demonstrated that ¹⁸F‑FDG PET/CT and MRI had comparable diagnostic efficacy in evaluating lymph node metastasis in breast cancer and therapeutic response in pancreatic cancer. Unfortunately, no combination or direct comparison of the two techniques was performed in those studies.In the present study, both SUVmax and ADC values enabled the quantitative assessment of high-risk status in stage II CRC. To date, there is still no consensus regarding which imaging modality is superior, and our results also indicated that the two imaging techniques exhibited equivalent diagnostic performance. However, we attempted to combine the two imaging techniques with CEA, and the combined model was superior to any single modality. Therefore, it is reasonable to believe that preoperative assessment of high-risk status in stage II CRC using the combination of the two techniques and CEA could benefit patients. Lee et al. (38) systematically compared the diagnostic efficacy of DSC-MRI perfusion metrics and ¹¹C-methionine PET parameters for post-treatment surveillance of gliomas, verifying that the combined application of the two functional imaging modalities yields more reliable differentiation of tumor recurrence from radiation injury than single-modality evaluation. Their results revealed that the combination of the two techniques improved diagnostic accuracy. The combination of ¹⁸FFDG PET/CT and MRI has also been shown to be superior to either modality alone in assessing staging and lymph node metastasis in rectal cancer (39,40). Findings from studies across multiple systems and the present study consistently indicate that the combination of ¹⁸F‑FDG PET/CT and MRI can benefit patients in terms of treatment and prognosis.

Emerging evidence has demonstrated significant correlations between quantitative parameters derived from ¹⁸F-FDG PET/CT and DWI in multiple malignancies (41-43). Specifically, ADC from DW-MRI exhibits a robust negative correlation with SUVmax from ¹⁸F-FDG PET/CT, reflecting the intrinsic link between tumor cellular density, glycolytic metabolic activity, and malignant proliferation potential. Nevertheless, single-modality imaging still has inherent limitations in comprehensively characterizing tumor heterogeneity and predicting therapeutic response. The integration of metabolic PET/CT and diffusion functional MRI can complement each other’s advantages, providing multidimensional quantitative tumor information. The findings of the present study also demonstrated a negative correlation between SUVmax and ADC values in the assessment of high-risk status in stage II CRC.

Our study has several limitations. First, we did not perform detailed subgrouping based on the proportions of various histological components (glandular, mucinous, solid components, etc.) in CRC pathological specimens. Prospective collection of tumor specimens is needed in future studies to explore tumor heterogeneity from a pathological perspective. Second, this study adopted binary high/non-high-risk stratification based on the 2025 CSCO guidelines, which is unified and clinically practicable. Nevertheless, the actual risk progression of stage II CRC is continuous rather than completely dichotomous. Some patients near the grouping threshold present mixed clinicopathological features and ambiguous risk status, showing the fuzzy transition of tumor risk. Such inherent defects of binary grouping may cause subtle classification bias and limit the accuracy of chemotherapy benefit stratification. Further large-sample studies constructing continuous quantitative risk scoring models are needed to optimize individualized risk assessment for borderline populations. Third, multiple ROI placements in MRI and ¹⁸F‑FDG PET/CT may not be completely aligned at the same level, leading to certain bias. Future research will expand the sample size and adopt multi-center prospective data for external validation to optimize and stabilize the predictive model. Moreover, we will attempt to construct a continuous quantitative risk scoring system to refine the risk stratification of borderline patients, eliminate the classification bias of binary grouping, and further improve the accuracy and individualization of adjuvant chemotherapy benefit prediction for stage II CRC.

Conclusions

DWI and ¹⁸F‑FDG PET/CT allow quantitative assessment of the high‑risk status of CRC and can guide the formulation of clinical chemotherapy regimens. Meanwhile, ADC values and SUVmax are significantly correlated, and the two techniques exert complementary effects. Combined with CEA, they significantly improve diagnostic efficacy and bring clinical benefits to patients.

Acknowledgments

The authors thank Lei Li, at Zhuhai People’s Hospital (The Affiliated Hospital of Beijing Institute of Technology, Zhuhai Clinical Medical College of Jinan University), for the support during the ¹⁸F‑FDG PET/CT data acquisition.

Footnote

Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0384/rc

Data Sharing Statement: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0384/dss

Peer Review File: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0384/prf

Funding: This work was supported by the Zhuhai Science and Technology Program Project, Guangdong Province, China (No. 2520004003551).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0384/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Ethics Committee of Zhuhai People’s Hospital (No. 2023-KT-33), and informed consent was waived because it was a retrospective analysis of pre-existing de-identified data.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Siegel RL, Wagle NS, Star J, et al. Colorectal cancer statistics, 2026. CA Cancer J Clin 2026;76:e70067. [Crossref] [PubMed]
2. Wu Y, He S, Cao M, et al. Comparative analysis of cancer statistics in China and the United States in 2024. Chin Med J (Engl) 2024;137:3093-100. [Crossref] [PubMed]
3. Savu E, Vasile L, Serbanescu MS, et al. Clinicopathological Analysis of Complicated Colorectal Cancer: A Five-Year Retrospective Study from a Single Surgery Unit. Diagnostics (Basel) 2023;13:2016. [Crossref] [PubMed]
4. Shaukat A, Levin TR. Current and future colorectal cancer screening strategies. Nat Rev Gastroenterol Hepatol 2022;19:521-31. [Crossref] [PubMed]
5. Lu D, Dong C, Wang K, et al. Updates of CSCO guidelines for colorectal cancer version 2025. Chin J Cancer Res 2025;37:297-302. [Crossref] [PubMed]
6. Iveson TJ, Sobrero AF, Yoshino T, et al. Duration of Adjuvant Doublet Chemotherapy (3 or 6 months) in Patients With High-Risk Stage II Colorectal Cancer. J Clin Oncol 2021;39:631-41. [Crossref] [PubMed]
7. Iveson T, Saunders MP, Kelly C, et al. Three Versus 6 Months of Adjuvant Oxaliplatin-Fluoropyrimidine Chemotherapy for Colorectal Cancer: Final Results of SCOT-An International, Randomized, Phase III, Noninferiority Trial. J Clin Oncol 2026;44:534-9. [Crossref] [PubMed]
8. Chen XY, Li HX, Cheng H, et al. Microsatellite instability in colorectal cancer. Indian J Pathol Microbiol 2025;68:255-60. [Crossref] [PubMed]
9. Kotani D, Oki E, Nakamura Y, et al. Molecular residual disease and efficacy of adjuvant chemotherapy in patients with colorectal cancer. Nat Med 2023;29:127-34. [Crossref] [PubMed]
10. Zhang L, Chen J, Wang Y. Application value of SOMATOM Force computed tomography in assisting the preoperative localization of colorectal cancer resection surgery. J Int Med Res 2024;52:1-9. Available online.
11. Huang QY, Zheng HD, Xiong B, et al. Preoperative prediction of multiple biological characteristics in colorectal cancer using MRI and machine learning. Heliyon 2025;11:e41852. [Crossref] [PubMed]
12. Mohammadzadeh S, Mohebbi A, Zare M, et al. Diagnostic value of quantitative DWI and IVIM parameters in differentiating intrahepatic cholangiocarcinoma and hepatocellular carcinoma: a systematic review and meta-analysis. Abdom Radiol (NY) 2026;51:1244-60. [Crossref] [PubMed]
13. Qu W, Wang J, Hu X, et al. MRI radiomic study on prediction of nonenlarged lymph node metastasis of rectal cancer: reduced field-of-view versus conventional DWI. Eur Radiol Exp 2025;9:34. [Crossref] [PubMed]
14. Phuttharak W, Wannasarnmetha M, Lueangingkasut P, et al. Differentiation between germinoma and other pineal region tumors using diffusion-and susceptibility-weighted MRI. Eur J Radiol 2023;159:110663. [Crossref] [PubMed]
15. Mohanasundaram P, Kumar J, Pradhan GS, et al. Role of resolve DWI and DCE MRI in differentiating benign and malignant skull base lesions. Neuroradiology 2025;67:2573-83. [Crossref] [PubMed]
16. Deng Y, Bi Q, Wang Q, et al. The performance of DWI and ADC values with different b-values for the diagnosis and staging of endometrial carcinoma at 3T. Cancer Imaging 2025;25:126. [Crossref] [PubMed]
17. Yuan L, Lin X, Zhao P, et al. Correlations between DKI and DWI with Ki-67 in gastric adenocarcinoma. Acta Radiol 2023;64:1792-8. [Crossref] [PubMed]
18. Salimi M, Vadipour P, Khosravi A, et al. Diagnostic accuracy of ¹⁸F-FDG PET/CT radiomics for non-invasive prediction of PD-L1 expression in non-small cell lung cancer: A systematic review and meta-analysis. Ann Nucl Med 2026;40:231-47. [Crossref] [PubMed]
19. Whi W, Moon SH, Oh D, et al. Diagnostic Value of Surveillance 18F-FDG PET/CT in Head and Neck Squamous Cell Carcinoma After Curative Therapy. Clin Nucl Med 2025;50:301-6. [Crossref] [PubMed]
20. Tao X, Zhang Q, Li N, et al. Prognostic Value of 18F-FDG PET/CT in Neoadjuvant PD-1 Inhibitor-treated NSCLC: A Five-year Follow-up Study. Clin Nucl Med 2025;50:577-87. [Crossref] [PubMed]
21. Tamsel I, Kocabeyoglu B, Hekimsoy I, et al. The radiologic and FDG uptake findings of osseous lesions incidentally detected on 18F-FDG-PET/CT imaging. Acta Radiol 2025;66:1029-35. [Crossref] [PubMed]
22. Qin J, Han C, Li H, et al. Relationship between PD-L1 expression and [18F]FAPI versus [18F]FDG uptake on PET/CT in lung cancer. Eur J Nucl Med Mol Imaging 2025;52:3211-22. [Crossref] [PubMed]
23. Xu H, Wang H, Yu SZ, et al. Prognostic and diagnostic value of [18F]FDG, 11C-acetate, and [68Ga]Ga-FAPI-04 PET/CT for hepatocellular carcinoma. Eur Radiol 2025;35:4121-31. [Crossref] [PubMed]
24. Yin Y, Sun Z, Deng X, et al. Integration of deep learning and radiomic features from multiplex immunohistochemistry images for reproducible Multi-Outcome prediction in a Multi-Center study of colorectal cancer. Int J Med Inform 2026;214:106436. [Crossref] [PubMed]
25. Kim HB, Tan HQ, Nei WL, et al. Impact of large language models and vision deep learning models in predicting neoadjuvant rectal score for rectal cancer treated with neoadjuvant chemoradiation. BMC Med Imaging 2025;25:306. [Crossref] [PubMed]
26. El Jiani L, El Filali S, Ben Lahmar EH, et al. Deep Learning-Based Approaches Using Medical Imaging for Therapy Response Prediction in Breast Cancer: A Systematic Literature Review. International Journal of Online & Biomedical Engineering 2024;20:37-54.
27. Shi D, Kou Q, Chen HY, et al. Quantitative analysis of enhanced CT in predicting microvascular invasion and pathological grading of hepatocellular carcinoma. Eur Radiol 2026;36:2778-93. [Crossref] [PubMed]
28. Diergaarde B, Young G, Hall DW, et al. Circulating Tumor DNA as a Marker of Recurrence Risk in Stage III Colorectal Cancer: The α-CORRECT Study. J Surg Oncol 2025;132:175-86. [Crossref] [PubMed]
29. Lepage C, Phelip JM, Cany L, et al. Effect of 5 years of CT-scan and CEA follow-up on survival endpoints in patients with colorectal cancer: the PRODIGE-13 FFCD phase III trial. Ann Oncol 2025;36:1468-79. [Crossref] [PubMed]
30. Yang Q, Bai C, Xu Y, et al. Construction of a postoperative liver metastasis prediction model for colorectal cancer based on spectral CT imaging, CEA, and CA19-9. Transl Cancer Res 2025;14:2113-24. [Crossref] [PubMed]
31. Sonoda H, Yamada T, Matsuda A, et al. Elevated serum carcinoembryonic antigen level after curative surgery is a prognostic biomarker of stage II-III colorectal cancer. Surg Today 2024;54:987-95. Available online.
32. Huang YQ, Chen XB, Cui YF, et al. Enhanced risk stratification for stage II colorectal cancer using deep learning-based CT classifier and pathological markers to optimize adjuvant therapy decision. Ann Oncol 2025;36:1178-89. [Crossref] [PubMed]
33. Xie C, Ning Z, Guo T, et al. Multimodal data integration for biologically-relevant artificial intelligence to guide adjuvant chemotherapy in stage II colorectal cancer. EBioMedicine 2025;117:105789. [Crossref] [PubMed]
34. Zhang L, Li J, Han Y, et al. Multi-center evaluation of machine learning-based radiomic model in predicting disease free survival and adjuvant chemotherapy benefit in stage II colorectal cancer patients. J Clin Med 2023;12:5321. Available online.
35. Ye Y, Lu W, Deng Q, et al. Tumor enhancement ratio on preoperative abdominal contrast-enhanced CT scan for predicting recurrence risk in stage II colon cancer. Abdom Radiol (NY) 2022;47:1265-75. [Crossref] [PubMed]
36. Han S, Huang Y. Comparison of the Performance of [18F]-FDG PET/CT and [18F]-FDG PET/MRI for Lymph Node Metastasis in Breast Cancer: A Systematic Review and Meta- Analysis. Curr Med Imaging 2024;20:e15734056304849. [Crossref] [PubMed]
37. Zhang Y, Wang H, Li X. Head-to-head comparison of 18F-FDG PET/CT and MRI for neoadjuvant chemotherapy response evaluation in pancreatic cancer: a meta-analysis. J Clin Imaging 2023;96:452-60. Available online.
38. Lee SH, Han JY, Shin NY. Comparative performance of DSC-MRI perfusion metrics and ¹¹C-methionine PET parameters in post-treatment glioma surveillance. Neuroradiology 2025;67:589-98. Available online.
39. Gong X, Zhang Y, Liu H, et al. Lymph nodes primary staging of colorectal cancer in 18F-FDG PET/MRI: a systematic review and meta-analysis. Quant Imaging Med Surg 2023;13:3214-24. Available online.
40. Li M, Wang L, Chen J, et al. FDG-PET/MRI outperforms standalone MRI and PET/CT for staging and nodal metastasis detection in locally advanced rectal cancer: an updated systematic review. J Clin Imaging 2024;98:88-96. Available online.
41. Morsy M, El-Sayed M, Khairy A, et al. Correlation between chest DW-MRI and 18F-FDG PET/CT parameters in newly diagnosed NSCLC and their role in predicting therapeutic response. J Cardiothorac Surg 2024;19:45. Available online.
42. Yilmaz S, Tekin B. Correlation of 3T Diffusion-weighted MRI and 18F-FDG-PET/CT in Liver Metastases: SUV Versus ADC. Mol Imaging Radionucl Ther 2024;33:112-8. Available online.
43. Abdel Razek AAK, Elshafey RA, Elawady M, et al. The Relationship Between Functional Parameters Derived from Diffusion-Weighted MRI and 18F-Fluorodeoxyglucose PET/CT in Head and Neck Squamous Cell Carcinoma: A Systematic Review and Meta-analysis. Diagnostics 2025;15:2418. Available online.