Multiparametric MRI-derived biomarkers for preoperative prediction of recurrence and/or metastasis after neoadjuvant chemoradiotherapy in locally advanced rectal cancer

Introduction

Colorectal cancer is among the most common malignancies globally, with its incidence continuing to rise in recent years (1). Locally advanced rectal cancer (LARC) typically refers to tumors staged as T3–T4 or N1–N2 (2). It is estimated that approximately 70% of newly diagnosed rectal cancer patients present with LARC (3). Neoadjuvant chemoradiotherapy (nCRT) is a standard treatment for LARC, aiming to reduce tumor burden, downstage disease, and increase the likelihood of sphincter-preserving surgery (4). However, due to tumor heterogeneity, patients exhibit varying sensitivity to nCRT. Although 50–75% of LARC patients experience some degree of tumor regression after nCRT (5), around 30% show limited response, potentially missing the optimal treatment window and experiencing worse prognosis (6). Total mesorectal excision (TME) significantly improves survival (7), and postoperative adjuvant therapies have been widely adopted. Nonetheless, the rate of postoperative recurrence and/or metastasis (RM) remains high (8), and RM is associated with poor clinical outcomes (9). Therefore, accurate preoperative identification of patients at high or low risk of RM is crucial. Tailored surveillance and treatment plans for high-risk individuals may yield survival benefits, aligning with the goals of precision medicine.

Currently, the pathological examination of resected specimens is a key prognostic tool in rectal cancer (10). However, it is invasive and cannot be performed preoperatively. Multiparametric magnetic resonance imaging (mpMRI), which includes diffusion-weighted imaging (DWI) and dynamic contrast-enhanced MRI (DCE-MRI), offers a promising noninvasive alternative. The apparent diffusion coefficient (ADC), derived from DWI, reflects tumor cellularity and response to therapy (11,12). DCE-MRI provides insight into tumor microvascular perfusion and permeability by analyzing the temporal behavior of contrast agent dynamics (13). By combining DWI and DCE-MRI, both diffusion characteristics and hemodynamic features of the tumor can be evaluated, potentially capturing subtle biological changes associated with treatment response (14).

In recent years, mpMRI has gained increasing attention in oncology research. Through multimodal scanning and comprehensive analysis, mpMRI enables a more holistic, multiparametric characterization of tumors. It shows promise for tumor diagnosis, treatment response assessment, and prognosis prediction—including recurrence and metastasis—thus assisting clinicians in formulating more personalized management strategies (15).

This study aimed to investigate whether mpMRI-derived parameters obtained around nCRT can preoperatively predict RM risk in patients with LARC. Identifying effective imaging biomarkers for RM risk stratification could help to personalize treatment and follow-up strategies, ultimately improving patient outcomes and advancing the goals of precision oncology. We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-2040/rc).

Methods

Study design and patients

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of Tongde Hospital of Zhejiang Province (No. 2025-397 (K) -F1), and the requirement for informed consent was waived due to its retrospective nature.

A total of 145 patients with surgically and pathologically confirmed LARC were identified from January 2016 to December 2021. Of these, 126 completed the full course of nCRT. The inclusion criteria were as follows: (I) primary rectal adenocarcinoma confirmed by colonoscopic biopsy; (II) patients who underwent nCRT after pathological confirmation; (III) MRI examinations had been conducted within 1 week before nCRT and within one week before TME after nCRT, and the preoperative MRI staging was T3–T4 or N1–N2 rectal cancer; (IV) no anti-tumor treatment was received before preoperative MRI examination. The exclusion criteria were as follows: (I) patients with incomplete clinical data and postoperative pathological data; (II) patients who could not tolerate nCRT and stopped midway; (III) patients with incomplete MRI images or poor image quality affecting evaluation; (IV) patients who did not undergo TME within 4–6 weeks after nCRT; (V) patients lost to follow-up. A flow diagram of patient recruitment is shown in Figure 1.

Figure 1 Flow diagram of this study. LARC, locally advanced rectal cancer; MRI, magnetic resonance imaging; nCRT, neoadjuvant chemoradiotherapy; RM, recurrence and/or metastasis; TME, total mesorectal excision.

Equipment and scanning methods

All examinations were performed on a 3.0-T Siemens Verio scanner (Siemens, Erlangen, Germany) with a body array coil, patient positioned head-first supine. A MEDRAD MR automatic high-pressure injector produced by Bayer (Leverkusen, Germany) was used; the contrast agent was gadopentetate dimeglumine injection (Gd-DTPA; Beilu Pharmaceutical Co., Ltd., Beijing, China). All patients underwent MRI examinations within one week before nCRT (pre-nCRT) and within one week before TME after nCRT (post-nCRT). The post-nCRT MRI was performed at a median of 4 weeks (range, 3–5 weeks) after nCRT, which was determined by the TME scheduling (4–6 weeks post-nCRT) minus the 1-week pre-TME MRI interval to ensure optimal evaluation of tumor response to nCRT. The selected sequences in this study included T2-weighted imaging (T2WI), DWI, and contrast enhancement T1-weighted imaging (CE-T1) sequence with conventional axial scans. The scanning parameters are shown in Table 1.

Clinical treatment

All patients in this group were treated with nCRT before surgery. The surgical treatment plan was radical surgery in line with the principle of TME, and the surgical target was R0 resection. The treatment plan is shown in Table 2.

Image analysis and data measurement

All regions of interest (ROIs) delineation and quantitative parameter measurement were performed independently in a blinded manner by two abdominal radiologists with more than 5 years of imaging experience, and a third senior radiologist (15 years of experience) was consulted for arbitration in case of significant discrepant results. The unified protocol was as follows.

• Slice selection: for both ADC and DCE-MRI, the maximal cross-sectional plane of the lesion was set as the ROI placement plane, confirmed by multi-sequence fusion (T2WI for anatomical localization, DWI for viable tumor identification, and DCE-MRI for vascularized region detection) to target the tumor core accurately. Heterogeneous tissue exclusion: prior to ROI placement, necrotic/hemorrhagic areas (T2WI hypointensity/heterointensity, DCE-MRI non-enhancement) and mucin-rich regions (T2WI hyperintensity, DWI hypointensity, high ADC value, mild/no enhancement) were identified and excluded. ROIs were only placed in enhanced viable tumor areas, at least 2 mm from lesion edges, blood vessels, and heterogeneous tissues. ROI shape and size: ROIs were circular/elliptical with a consistent area of 25–60 mm² for both ADC and DCE-MRI. Quantitative measurement: For ADC, three non-overlapping ROIs were evenly placed in the tumor core, and the mean of the three measurements was taken as the final single-measurement value; for DCE-MRI, a single ROI covering the main viable tumor core (excluding non-enhanced areas) was placed, with the mean value calculated as the final single-measurement value.
• Consistency verification: intra- and inter-observer agreements were assessed using the intraclass correlation coefficient (ICC). Intraobserver agreement was determined by the ICC between the initial measurement and a 1-month re-measurement by one radiologist, whereas interobserver agreement was calculated by the ICC between the initial measurements of the two radiologists. All parameters exhibited excellent consistency (all ICC >0.80), as detailed in Table S1.

For DCE-MRI parameter measurement, data were imported into the MMWP version workstation (Siemens) and analyzed using Tissue4D software (Siemens), with detailed steps as follows: Step 1: data preprocessing: import raw DCE-MRI dynamic sequence data, which were automatically converted to Digital Imaging and Communications in Medicine (DICOM) standard format. The motion correction results from the scanning phase were applied to align all phase images consistently, and signals were normalized using adjacent muscle tissue as a reference to eliminate scanning noise. Step 2: image loading: load the preprocessed dynamic sequence data by double-clicking the patient folder or dragging it into the Tissue 4D post-processing module. Step 3: motion correction verification: review the alignment of each phase image with the first phase; no manual adjustment was required for all patients in this study due to effective pre-scan breath-hold training and three-dimensional (3D) rigid registration. Step 4: registration: register morphological images (T2WI/CE-T1) with dynamic scanning images to lock the scope and shape of the lesion, ensuring consistent ROI positioning across all phases. Step 5: ROI delineation: manually outline a circular or elliptical ROI (area: 25–60 mm²) on the maximal cross-sectional plane of the lesion, avoiding blood vessels, bleeding, necrosis, calcifications, and mucin-rich regions. Step 6: arterial input function calibration: Import the pre-extracted and manually corrected descending aorta arterial input function curve, which was automatically matched with the lesion’s time-signal curve via the software to eliminate abnormal peaks caused by injection artifacts. Step 7: parameter calculation: a two-compartment Tofts model was used to generate perfusion pseudo-color maps of the ROI, and quantitative parameters including volume transfer constant (Ktrans), rate constant (Kep), and extracellular volume ratio (Ve) were calculated. The fitting goodness of parameters was evaluated by the coefficient of determination (R²) built into Tissue4D software, and all parameters had R²>0.80, indicating reliable model fitting. Step 8: result confirmation: each lesion was measured 3 times (with a 1-week interval between measurements by the same radiologist) to reduce intra-observer variability, and the average value was taken as the final parameter result.

ADC change percentage (ΔADC%) = [ADC after nCRT (post-ADC) − ADC before nCRT (pre-ADC)] / ADC before nCRT (pre-ADC)] ×100%.

Ktrans change percentage (ΔKtrans%) = [(Ktrans before nCRT (pre-Ktrans) − Ktrans after nCRT (post-Ktrans) / Ktrans before nCRT (pre-Ktrans)] ×100%, as well as extracellular volume fraction (Ve) change percentage (ΔVe%) and Kep change percentage (ΔKep%).

Follow-up data collection

Patient follow-up data were collected by reviewing inpatient and outpatient medical records, imaging reports, and laboratory results. The collected information included quality of life (e.g., bowel function, intestinal obstruction), time to tumor recurrence, and time to distant metastasis. According to standard medical protocols, follow-up was conducted every 3 months for the first 2 years after surgery and every 6 months thereafter. Surviving patients were followed for up to 3 years. The primary endpoint was disease-free survival, defined as the time from surgery to tumor recurrence, progression, or death from any cause. Follow-up assessments included evaluation of clinical symptoms, laboratory tests, imaging studies, and colonoscopy. The disease status was comprehensively assessed based on the integration of all available data. For patients with imaging findings suggesting RM, further pathological examination was performed for confirmation. According to postoperative imaging and colonoscopy of rectal cancer, patients without recurrence at the original surgical resection site, lymph node metastasis, or distant organ metastasis were included in the non-recurrence and non-metastasis group (non-RM group), and vice versa in the RM group.

Statistical analysis

Statistical analysis was performed using PASW23.0 software (IBM Corp., Armonk, NY, USA). Measurement data conforming to normal distribution were expressed as mean ± standard deviation (SD). For intragroup paired sample comparison, paired t-test was used; for intergroup independent sample comparison, independent sample t-test was used. Measurement data not conforming to normal distribution were expressed as median (M) and interquartile range (Q). For intragroup paired sample comparison, Wilcoxon signed-rank test was used; for intergroup independent sample comparison, Mann-Whitney U test was used. Count data were expressed as cases or percentages, and the comparison was performed using the Chi-squared test. For intragroup paired sample comparison, McNemar test was used; for intergroup independent sample comparison, Pearson chi-square test was used. Univariate logistic regression analysis was used to screen for risk factors related to RM after nCRT treatment; to assess potential multicollinearity among variables (ΔADC%, post-Ktrans, ΔKtrans%) in the combined model, variance inflation factor (VIF) was calculated. Results showed that VIF (ΔADC%) =1.01, VIF (post-Ktrans) =1.21, VIF (ΔKtrans%) =1.21, and the average VIF of the model =1.14 (all VIF <3), indicating no severe multicollinearity. The receiver operating characteristic (ROC) curve was drawn to analyze the predictive efficacy of each model; the calibration curve (CC) was used to detect the calibration degree of each model, and decision curve analysis (DCA) and clinical impact curve (CIC) were used to evaluate the clinical practicability of each model.

Results

Patient characteristics and representative cases

This study enrolled a total of 86 patients with rectal cancer. Based on 3-year follow-up results, patients were divided into the non-RM group (n=61, 70.9%) and the RM group (n=25, 29.1%). Baseline clinical characteristics of the two groups were compared (Table S2), with only the circumferential resection margin and magnetic resonance tumor regression grade showing statistically significant intergroup differences (P<0.05).

All patients underwent DWI and DCE-MRI within one week before nCRT (pre-nCRT) and within one week before total TME surgery after nCRT (post-nCRT). Histopathological examination of the surgical specimens confirmed rectal adenocarcinoma in all cases. All 86 patients completed the 3-year follow-up. Among the 25 RM cases, 3 (12.0%) were isolated local recurrence (limited to the pelvic cavity, original tumor bed, or regional lymph nodes), 21 (84.0%) cases were distant metastasis events, and 1 (4.0%) was concurrent local recurrence and distant metastasis. The detailed breakdown confirms that distant metastasis is the main form of RM in this cohort, with only a small number of local recurrence events. Representative imaging examples are provided below (Figures 2,3).

Figure 2 A 56-year-old male patient in the non-recurrence and non-metastasis group. Pre-nCRT MRI (A) T2WI shows obvious thickening of the middle and lower rectal wall with luminal stenosis (red arrow); (B) ADC map demonstrates restricted diffusion of the tumor, with an ADC value of 1.077×10⁻³ mm²/s; (C) DCE-MRI reveals significant enhancement of the tumor, indicating abundant blood supply; (D) PWI shows Ktrans, Kep, and Ve values of 0.235 min⁻¹, 0.562 min⁻¹, and 0.415, respectively. Two months after nCRT, follow-up MRI (E) T2WI shows reduced tumor volume and improved luminal stenosis (red arrow); (F) ADC map demonstrates reduced restriction of tumor diffusion, with an ADC value increased to 1.638×10⁻³ mm²/s; (G) DCE-MRI reveals decreased degree of tumor enhancement compared with baseline, indicating reduced blood supply; (H) PWI shows Ktrans, Kep, and Ve values of 0.201 min⁻¹, 0.310 min⁻¹, and 0.662, respectively. ADC, apparent diffusion coefficient; DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; Kep, rate constant; Ktrans, volume transfer constant; MRI, magnetic resonance imaging; nCRT, neoadjuvant chemoradiotherapy; PWI, perfusion-weighted imaging; T2WI, T2-weighted imaging; Ve, extracellular volume fraction.

Figure 3 A 63-year-old male patient in the recurrence and metastasis group. Pre-nCRT MRI (A) T2WI shows local thickening of the middle rectal wall with luminal stenosis (red arrow); (B) ADC map demonstrates restricted diffusion of the tumor, with an ADC value of 1.142×10⁻³ mm²/s; (C) DCE-MRI reveals significant enhancement of the tumor, indicating rich blood supply; (D) PWI shows Ktrans, Kep, and Ve values of 0.208 min⁻¹, 0.473 min⁻¹, and 0.497, respectively. Six months after nCRT, follow-up MRI (E) T2WI shows increased tumor volume and obvious luminal stenosis (red arrow); (F) ADC map demonstrates aggravated restriction of tumor diffusion, with an ADC value decreased to 0.738×10⁻³ mm²/s; (G) DCE-MRI reveals significantly increased degree of tumor enhancement compared with baseline, indicating increased blood supply; (H) PWI shows Ktrans, Kep, and Ve values of 0.417 min⁻¹, 0.640 min⁻¹, and 0.666, respectively. ADC, apparent diffusion coefficient; DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; Kep, rate constant; Ktrans, volume transfer constant; MRI, magnetic resonance imaging; nCRT, neoadjuvant chemoradiotherapy; PWI, perfusion-weighted imaging; T2WI, T2-weighted image; Ve, extracellular volume fraction.

Group comparisons and predictors

Comparison of the mpMRI quantitative parameters between the two groups revealed statistically significant differences in ΔADC%, post-Ktrans, and ΔKtrans% (P<0.05; Table 3). The correlation between each quantitative parameter is shown in Figure 4A,4B. Furthermore, univariate logistic regression analysis concluded that ΔADC%, post-Ktrans, and ΔKtrans% are predictive factors for the RM risk of patients after nCRT (Figure 5).

Figure 4 Correlation clustering heatmap and standardized expression heatmap of MRI quantitative parameters. (A) The heatmap displays the Pearson correlation coefficients between all quantitative MRI parameters, with a color gradient from light areas (0, no correlation) to dark areas (1, perfect positive correlation). Dendrograms on the vertical and horizontal axes represent the hierarchical clustering structure of parameters based on their correlation patterns. Values within the heatmap indicate the specific correlation coefficients between individual parameter pairs. VIF analysis [VIF (post-Ktrans) =1.21, VIF (ΔKtrans%) =1.21] indicates no severe multicollinearity. (B) This heatmap visualizes the standardized expression levels of MRI quantitative parameters across the study cohort. Each row represents an individual patient, and each column represents a specific MRI feature (pre-treatment, post-treatment, or percentage change). The color gradient from red (low expression) to dark blue (high expression) illustrates the relative distribution of parameter values, enabling the identification of distinct expression patterns among the cohort. ADC, apparent diffusion coefficient; ID, depth of invasion; Kep, rate constant; Ktrans, volume transfer constant; MRI, magnetic resonance imaging; Ve, extracellular volume fraction; VIF, variance inflation factor.

Figure 5 Forest plot of logistic regression analysis. The analysis demonstrated that ΔADC% (OR =0.032, 95% CI: 0.002–0.569), post-Ktrans (OR =0.840, 95% CI: 0.623–1.134), and ΔKtrans% (OR =0.812, 95% CI: 0.063–0.713) are predictive factors for the RM risk of LARC patients after nCRT treatment. ADC, apparent diffusion coefficient; CI, confidence interval; Ktrans, volume transfer constant; LARC, locally advanced rectal cancer; nCRT, neoadjuvant chemoradiotherapy; OR, odds ratio; RM, recurrence and/or metastasis.

Comparative performance of prediction models

ROC curve analysis was performed on the single models (ΔADC%, post-Ktrans, ΔKtrans%) and the combined model (ΔADC% + post-Ktrans + ΔKtrans%); ROC analysis demonstrated the superior predictive efficacy, in descending order: combined model > ΔKtrans% > post-Ktrans > ΔADC%. (Figure 6, Table 4).

Figure 6 ROC curves of single models (ΔADC%, post-Ktrans, ΔKtrans%) and combined model (ΔADC% + post-Ktrans + ΔKtrans%) for predicting RM risk of LARC patients after nCRT. ADC, apparent diffusion coefficient; AUC, area under the curve; CI, confidence interval; Ktrans, volume transfer constant; LARC, locally advanced rectal cancer; nCRT, neoadjuvant chemoradiotherapy; RM, recurrence and/or metastasis; ROC, receiver operating characteristic.

Model performance and clinical utility

In the study, a nomogram was constructed based on the combined model (ΔADC% + post-Ktrans + ΔKtrans%) (Figure 7). The calibration curve showed that the nomogram had good consistency in predicting the RM risk of LARC patients after nCRT (Figure 8A). DCA showed that the nomogram had a greater net benefit than the single models (Figure 8B). The CIC showed that the nomogram had higher clinical practicability than the single models (Figure 8C).

Figure 7 Nomogram constructed based on the combined model (ΔADC% + post-Ktrans + ΔKtrans%), which can correspond the scores of each variable to the corresponding values, and finally sum them up to obtain the total score for predicting the RM risk of patients after nCRT. ADC, apparent diffusion coefficient; Ktrans, volume transfer constant; nCRT, neoadjuvant chemoradiotherapy; RM, recurrence and/or metastasis.

Figure 8 Calibration, decision curve, and clinical impact analysis of the nomogram and individual MRI biomarkers for predicting recurrence and/or metastasis. (A) Calibration curves evaluating the agreement between predicted and observed probabilities of recurrence and/or metastasis. The 45° dashed line represents ideal prediction. Solid lines depict the performance of the nomogram (red), ΔADC% (green), post-Ktrans (blue), and ΔKtrans% (orange), respectively. (B) DCA assessing the clinical net benefit of predictive models across a range of high-risk threshold probabilities (X-axis). The Y-axis represents the net benefit. Curves for the nomogram (red), ΔADC% (green), post-Ktrans (blue), and ΔKtrans% (orange) are compared against two extreme strategies: treating all patients (gray line, “All”) and treating no patients (black line, “None”). (C) Clinical impact curves simulating model performance in a hypothetical cohort of 1,000 patients. The upper X-axis indicates the high-risk threshold probability, and the lower X-axis displays the corresponding cost-benefit ratio. The Y-axis shows the number of patients (per 1,000). Solid lines represent the number of patients classified as high-risk by each model, while dashed lines indicate the number of high-risk patients who experienced actual recurrence and/or metastasis events. ADC, apparent diffusion coefficient; DCA, decision curve analysis; Ktrans, volume transfer constant; MRI, magnetic resonance imaging.

Discussion

This study systematically evaluated the risk of RM in patients with LARC by integrating changes in tumor characteristics observed on MRI and the dynamic evolution of DCE-MRI quantitative parameters before and after nCRT. Among the investigated metrics—ΔADC%, post-Ktrans, and ΔKtrans%—the combined model incorporating all three demonstrated the highest diagnostic accuracy [area under the curve (AUC) =0.908]. The nomogram derived from this model provides an intuitive and clinically applicable tool for individualized RM risk assessment, with potential to guide follow-up and adjuvant treatment strategies.

Traditional morphological features, such as tumor volume, degree of luminal stenosis, and margin sharpness, remain valuable for evaluating treatment response and predicting prognosis (16). Larger tumors tend to be associated with higher recurrence rates, although this correlation may be confounded by factors such as tumor grade, tumor, node, metastasis (TNM) stage, and lymph node involvement (17). Importantly, increased tumor burden is linked to hypoxia-driven angiogenesis, which upregulates proangiogenic factors such as VEGF, promoting neovascularization and metastatic potential (18).

Functional MRI provides more than anatomical insights. ADC values reflect water molecule diffusion and are typically reduced in highly cellular, aggressive tumors (19,20). DCE-MRI parameters, such as Ktrans and Kep, represent vascular permeability and interstitial diffusion (21,22). Elevated Ktrans is often associated with leaky vasculature and heightened metastatic capacity (23-25), whereas Ve may be more susceptible to variability due to tumor microenvironment heterogeneity (26).

In our longitudinal analysis, distinct imaging trends were observed between groups. In the non-RM group, ADC values increased while perfusion parameters decreased after nCRT, indicating tumor cytoreduction and vascular normalization. In contrast, the RM group exhibited a slight decrease in ADC and increased perfusion parameters, suggestive of residual tumor viability and ongoing angiogenesis. This discrepancy is not merely a radiological finding but a direct reflection of tumors’ heterogeneous pathological responses to nCRT and inherent differences in their biological behaviors—ADC is inherently tightly correlated with tumor cellularity and tissue composition (27). Specifically, the elevated ΔADC% in the non-RM group is attributable to the effective induction of cancer cell death by nCRT and subsequent reduction in tumor cellularity, whereas the significantly lower ΔADC% in the RM group arises from nCRT resistance in a subset of cancer cells, with residual cells sustaining high proliferative activity and thereby constraining water molecule diffusion. Among all evaluated parameters, ΔADC% and ΔKtrans% were the most effective in differentiating RM risk, likely due to their capacity to capture dynamic changes and reduce inter-patient baseline variability (28). Given the time dependency of functional MRI parameters (ADC, Ktrans, Kep), post-nCRT imaging timing is critical to their validity. In this study, post-nCRT MRI was standardized to a 3–5-week imaging window (median: 4 weeks) after nCRT. Prior studies have shown quantitative MRI parameters stabilize after the 2nd or 3rd week post-nCRT (29); our 3–5-week imaging window falls within this stable phase, thus minimizing parameter distortion caused by temporal variability. These findings underscore the value of time-dependent imaging biomarkers over static pre- or post-treatment metrics.

The advantage of the multimodal model likely stems from the biological complementarity of DWI and DCE-MRI: DWI assesses microstructural integrity and cellularity, whereas DCE-MRI characterizes vascular perfusion and permeability (30). Similar multimodal approaches have enhanced diagnostic performance in breast imaging. For instance, Muthuvel et al. (31) reported improved specificity and accuracy when combining DWI and DCE-MRI for breast lesion classification, and others have confirmed their additive value in characterizing BI-RADS 4 lesions (32). However, Pinker et al. (33) noted potential specificity limitations when combining modalities in benign cases, highlighting the importance of disease-specific model calibration.

Beyond tumor size and burden, RM is increasingly recognized as systemic, microenvironment-mediated processes. The “pre-metastatic niche” hypothesis suggests that tumor-secreted exosomes and cytokines may precondition distant organs for metastatic colonization before clinical detection (34). Functional imaging may indirectly capture such early biological phenomena through subtle alterations in diffusion and perfusion patterns.

Despite its promising findings, this study has several limitations. As a single-center retrospective study with a relatively small cohort (n=86), potential selection bias and limited generalizability must be acknowledged. Furthermore, imaging quantification—especially for small or irregular lesions—may be affected by measurement variability. Although we implemented a standardized ROI-based protocol to mitigate sampling bias, whole-volume analysis—which theoretically better captures intratumoral heterogeneity by evaluating the entire tumor volume rather than a single cross-sectional ROI—was not performed due to practical constraints (e.g., no specialized volume segmentation software). Our primary endpoint was 3-year disease-free survival; future studies should include longer follow-up durations and external multicenter validation. Furthermore, systematic assessment of clinical variable-mpMRI parameter interactions may validate the independence of mpMRI parameters, and the integration of clinical variables, radiomic signatures, and molecular biomarkers could further boost the model’s robustness and predictive performance.

Conclusions

This study suggests that preoperative mpMRI-derived quantitative parameters, including ΔADC%, post-Ktrans, and ΔKtrans%, may serve as effective predictors of RM in LARC patients following nCRT. The combined model incorporating these features achieved high predictive accuracy (AUC =0.908), and the derived nomogram may provide a clinically practical and intuitive tool for individualized risk stratification. These findings highlight the potential of mpMRI as a noninvasive, biomarker-driven imaging approach that may assist in guiding personalized treatment and follow-up strategies in rectal cancer management.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-2040/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-2040/dss

Funding: This work was supported by the Medical Science and Technology Project of Zhejiang Province (No. 2024KY052), the Project of Zhejiang Provincial Administration of Traditional Chinese Medicine (No. 2024ZL040), and the Zhejiang Provincial Natural Science Foundation (grant No. TGY24H220006). The funder played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-2040/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 Ethics Committee of Tongde Hospital of Zhejiang Province (No. 2025-397 (K) -F1), and the requirement for informed consent was waived due to its retrospective nature.

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. Yang M, Lin SQ, Liu XY, Tang M, Hu CL, Wang ZW, et al. Association between C-reactive protein-albumin-lymphocyte (CALLY) index and overall survival in patients with colorectal cancer: From the investigation on nutrition status and clinical outcome of common cancers study. Front Immunol 2023;14:1131496. [Crossref] [PubMed]
2. Amin MB, Greene FL, Edge SB, Compton CC, Gershenwald JE, Brookland RK, Meyer L, Gress DM, Byrd DR, Winchester DP. The Eighth Edition AJCC Cancer Staging Manual: Continuing to build a bridge from a population-based to a more "personalized" approach to cancer staging. CA Cancer J Clin 2017;67:93-9.
3. Chen BY, Xie H, Li Y, Jiang XH, Xiong L, Tang XF, Lin XF, Li L, Cai PQ. MRI-Based Radiomics Features to Predict Treatment Response to Neoadjuvant Chemotherapy in Locally Advanced Rectal Cancer: A Single Center, Prospective Study. Front Oncol 2022;12:801743. [Crossref] [PubMed]
4. Wang Y, Zhang L, Jiang Y, Cheng X, He W, Yu H, Li X, Yang J, Yao G, Lu Z, Zhang Y, Yan S, Zhao F. Multiparametric magnetic resonance imaging (MRI)-based radiomics model explained by the Shapley Additive exPlanations (SHAP) method for predicting complete response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer: a multicenter retrospective study. Quant Imaging Med Surg 2024;14:4617-34. [Crossref] [PubMed]
5. Yuval JB, Garcia-Aguilar J. Watch-and-wait Management for Rectal Cancer After Clinical Complete Response to Neoadjuvant Therapy. Adv Surg 2021;55:89-107. [Crossref] [PubMed]
6. Zhang RX, Zhou ZG, Lu SX, Lu ZH, Wan DS, Pan ZZ, Wu XJ, Chen G. Pim-3 as a potential predictor of chemoradiotherapy resistance in locally advanced rectal cancer patients. Sci Rep 2017;7:16043. [Crossref] [PubMed]
7. Hamamoto H, Yamamoto M, Masubuchi S, Ishii M, Osumi W, Tanaka K, Okuda J, Uchiyama K. Male sex and anterior wall tumor location as risk factors for urinary dysfunction after laparoscopic rectal surgery. Surg Endosc 2020;34:3567-73. [Crossref] [PubMed]
8. Weiser MR, Chou JF, Keshinro A, Chapman WC Jr, Bauer PS, Mutch MG, Parikh PJ, Cercek A, Saltz LB, Gollub MJ, Romesser PB, Crane CH, Shia J, Markowitz AJ, Garcia-Aguilar J, Gönen MColorectal Cancer Disease Management Team of Memorial Sloan Kettering Cancer Center. Development and Assessment of a Clinical Calculator for Estimating the Likelihood of Recurrence and Survival Among Patients With Locally Advanced Rectal Cancer Treated With Chemotherapy, Radiotherapy, and Surgery. JAMA Netw Open 2021;4:e2133457. [Crossref] [PubMed]
9. Li M, Zhu YZ, Zhang YC, Yue YF, Yu HP, Song B. Radiomics of rectal cancer for predicting distant metastasis and overall survival. World J Gastroenterol 2020;26:5008-21. [Crossref] [PubMed]
10. Nikberg M, Kindler C, Chabok A, Letocha H, Shetye J, Smedh K. Circumferential resection margin as a prognostic marker in the modern multidisciplinary management of rectal cancer. Dis Colon Rectum 2015;58:275-82. [Crossref] [PubMed]
11. Hu X, Ye K, Bo S, Xiao Z, Ma M, Pan J, Zhong X, Zhang D, Mo X, Yu X, Chen M, Luo L, Shi C. Monitoring imatinib decreasing pericyte coverage and HIF-1α level in a colorectal cancer model by an ultrahigh-field multiparametric MRI approach. J Transl Med 2024;22:712. [Crossref] [PubMed]
12. Gui B, Miccò M, Valentini AL, Cambi F, Pasciuto T, Testa A, Autorino R, Zannoni GF, Rufini V, Gambacorta MA, Giordano A, Scambia G, Manfredi R. Prospective multimodal imaging assessment of locally advanced cervical cancer patients administered by chemoradiation followed by radical surgery-the "PRICE" study 2: role of conventional and DW-MRI. Eur Radiol 2019;29:2045-57. [Crossref] [PubMed]
13. El Homsi M, Yildirim O, Gangai N, Shia J, Gollub MJ, Mazaheri Y. Contrast-enhanced pelvic magnetic resonance imaging (MRI) for the prediction of treatment response in mucinous rectal cancer. Quant Imaging Med Surg 2024;14:4110-22. [Crossref] [PubMed]
14. Mohamed RM, Panthi B, Adrada BE, Boge M, Candelaria RP, Chen H, et al. Multiparametric MRI-based radiomic models for early prediction of response to neoadjuvant systemic therapy in triple-negative breast cancer. Sci Rep 2024;14:16073. [Crossref] [PubMed]
15. Arita Y, Kwee TC, Akin O, Shigeta K, Paudyal R, Roest C, Ueda R, Lema-Dopico A, Nalavenkata S, Ruby L, Nissan N, Edo H, Yoshida S, Shukla-Dave A, Schwartz LH. Multiparametric MRI and artificial intelligence in predicting and monitoring treatment response in bladder cancer. Insights Imaging 2025;16:7. [Crossref] [PubMed]
16. Song Z, Xing F, Jiang H, He Y, Lv J. Long non-coding RNA TP73-AS1 predicts poor prognosis and regulates cell proliferation and migration in cervical cancer. Arch Med Sci 2022;18:523-34. [Crossref] [PubMed]
17. Wang HX, Ding C, Huang JC, Ma YW, Lyu SC, Lang R. Prognostic Value for Perioperative Serum Total Cholesterol Level on Postoperative Long-Term Prognosis of Pancreatic Cancer: A Retrospective Clinical Study. Diagnostics (Basel) 2023;13:1402. [Crossref] [PubMed]
18. Feng K, Ma R, Zhang L, Li H, Tang Y, Du G, Niu D, Yin D. The Role of Exosomes in Thyroid Cancer and Their Potential Clinical Application. Front Oncol 2020;10:596132. [Crossref] [PubMed]
19. Wang Q, Mao N, Liu M, Shi Y, Ma H, Dong J, Zhang X, Duan S, Wang B, Xie H. Radiomic analysis on magnetic resonance diffusion weighted image in distinguishing triple-negative breast cancer from other subtypes: a feasibility study. Clin Imaging 2021;72:136-41. [Crossref] [PubMed]
20. Abdelhafez AH, Musall BC, Adrada BE, Hess K, Son JB, Hwang KP, et al. Tumor necrosis by pretreatment breast MRI: association with neoadjuvant systemic therapy (NAST) response in triple-negative breast cancer (TNBC). Breast Cancer Res Treat 2021;185:1-12. [Crossref] [PubMed]
21. Chen H, Min Y, Xiang K, Chen J, Yin G. DCE-MRI Performance in Triple Negative Breast Cancers: Comparison with Non-Triple Negative Breast Cancers. Curr Med Imaging 2022;18:970-6. [Crossref] [PubMed]
22. Musall BC, Abdelhafez AH, Adrada BE, Candelaria RP, Mohamed RMM, Boge M, et al. Functional Tumor Volume by Fast Dynamic Contrast-Enhanced MRI for Predicting Neoadjuvant Systemic Therapy Response in Triple-Negative Breast Cancer. J Magn Reson Imaging 2021;54:251-60. [Crossref] [PubMed]
23. Pelissier M, Ambarki K, Salleron J, Henrot P. Maximum slope using ultrafast breast DCE-MRI at 1.5 Tesla: a potential tool for predicting breast lesion aggressiveness. Eur Radiol 2021;31:9556-66. [Crossref] [PubMed]
24. Ramtohul T, Tescher C, Vaflard P, Cyrta J, Girard N, Malhaire C, Tardivon A. Prospective Evaluation of Ultrafast Breast MRI for Predicting Pathologic Response after Neoadjuvant Therapies. Radiology 2022;305:565-74. [Crossref] [PubMed]
25. Kim JJ, Kim JY, Suh HB, Hwangbo L, Lee NK, Kim S, Lee JW, Choo KS, Nam KJ, Kang T, Park H. Characterization of breast cancer subtypes based on quantitative assessment of intratumoral heterogeneity using dynamic contrast-enhanced and diffusion-weighted magnetic resonance imaging. Eur Radiol 2022;32:822-33. [Crossref] [PubMed]
26. Kim YE, Lim JS, Choi J, Kim D, Myoung S, Kim MJ, Kim KW. Perfusion parameters of dynamic contrast-enhanced magnetic resonance imaging in patients with rectal cancer: correlation with microvascular density and vascular endothelial growth factor expression. Korean J Radiol 2013;14:878-85. [Crossref] [PubMed]
27. Wang X, Xie Z, Wang X, Song Y, Suo S, Ren Y, Hu W, Zhu Y, Cao M, Zhou Y. Preoperative prediction of IDH genotypes and prognosis in adult-type diffuse gliomas: intratumor heterogeneity habitat analysis using dynamic contrast-enhanced MRI and diffusion-weighted imaging. Cancer Imaging 2025;25:11. [Crossref] [PubMed]
28. Lee JH, Yoo GS, Yoon YC, Park HC, Kim HS. Diffusion-weighted and dynamic contrast-enhanced magnetic resonance imaging after radiation therapy for bone metastases in patients with hepatocellular carcinoma. Sci Rep 2021;11:10459. [Crossref] [PubMed]
29. Borggreve AS, Heethuis SE, Boekhoff MR, Goense L, van Rossum PSN, Brosens LAA, van Lier ALHMW, van Hillegersberg R, Lagendijk JJW, Mook S, Ruurda JP, Meijer GJ. Optimal timing for prediction of pathologic complete response to neoadjuvant chemoradiotherapy with diffusion-weighted MRI in patients with esophageal cancer. Eur Radiol 2020;30:1896-907. [Crossref] [PubMed]
30. Gollub MJ, Blazic I, Felder S, Knezevic A, Gonen M, Garcia-Aguilar J, Paty PP, Smith JJ. Value of adding dynamic contrast-enhanced MRI visual assessment to conventional MRI and clinical assessment in the diagnosis of complete tumour response to chemoradiotherapy for rectal cancer. Eur Radiol 2019;29:1104-13. [Crossref] [PubMed]
31. Muthuvel D, Mohakud S, Deep N, Muduly D, Kumar P, Mishra P, Naik S. Usefulness of Combined Advanced Dynamic Contrast-Enhanced and Diffusion-Weighted MRI Over Ultrasonography in Differentiating Cancer From Benign Lesions in Dense Breasts. Cureus 2024;16:e69634. [Crossref] [PubMed]
32. Xu W, Zheng B, Li H. Identification of the Benignity and Malignancy of BI-RADS 4 Breast Lesions Based on a Combined Quantitative Model of Dynamic Contrast-Enhanced MRI and Intravoxel Incoherent Motion. Tomography 2022;8:2676-86. [Crossref] [PubMed]
33. Pinker K, Bogner W, Baltzer P, Gruber S, Bickel H, Brueck B, Trattnig S, Weber M, Dubsky P, Bago-Horvath Z, Bartsch R, Helbich TH. Improved diagnostic accuracy with multiparametric magnetic resonance imaging of the breast using dynamic contrast-enhanced magnetic resonance imaging, diffusion-weighted imaging, and 3-dimensional proton magnetic resonance spectroscopic imaging. Invest Radiol 2014;49:421-30. [Crossref] [PubMed]
34. Li Y, Wang H, Mao D, Che X, Chen Y, Liu Y. Understanding pre-metastatic niche formation: implications for colorectal cancer liver metastasis. J Transl Med 2025;23:340. [Crossref] [PubMed]