Abstract
Background
Tumor budding (TB) is recognized as an independent prognostic factor in rectal cancer (RC). This study aimed to evaluate the ability of preoperative synthetic MRI (SyMRI) quantitative parameters to predict high-grade TB (≥ 10 buds, Bd 3) in RC.
Methods
A total of 105 patients with RC were enrolled, including 42 with high-grade TB and 63 with non–high-grade TB (< 10 buds, Bd 1 and 2) patients. Parameters T1, T2 and PD values of SyMRI were obtained on a GE AW 4.7 workstation. Continuous variables were compared by independent-samples t tests, and categorical variables were analyzed using the chi-square test or Fisher’s exact test, as appropriate. Binary logistic regression was performed to identify independent predictors of high-grade TB. ROC analysis was used to assess the predictive performance of T1, T2, and their combined model. Model fit was compared using the Akaike information criterion (AIC), and calibration and overall predictive accuracy were evaluated using the Hosmer–Lemeshow test and the Brier score, respectively.
Results
Compared with non–high-grade TB group, high-grade TB group showed significantly higher T1 value (p < 0.001) and lower T2 value (p < 0.001). Multivariable logistic regression confirmed that both T1 and T2 values were independent predictors of high-grade TB. The combined T1 and T2 values model achieved an AUC of 0.926, outperforming the T1 value alone model (AUC = 0.883) and the T2 value alone model (AUC = 0.838). In addition, the combined model showed improved fit (AIC = 72.22 vs. 92.77 and 108.18) and better overall accuracy (Brier score = 0.107 vs. 0.135 and 0.168).
Conclusions
Parameters T1 and T2 values of SyMRI are independent predictors of high-grade TB in RC. A simplified model integrating T1 and T2 values provides superior discrimination and improved model fit and accuracy compared with single-parameter models, suggesting that T1 and T2 values may have potential for noninvasive assessment of TB status.
Trial registration
Not applicable.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12880-026-02367-3.
Keywords: Rectal cancer, Tumor budding, Synthetic MRI, Imaging biomarker, Prognostic factors
Introduction
Tumor budding (TB) in rectal cancer (RC) is defined as the presence of single tumor cells or small clusters of up to four tumor cells at the invasive front. Accumulating evidence has demonstrated that high-grade TB (≥ 10 buds, Bd 3) is closely associated with adverse pathological features, including a higher TNM stage, poorer tumor differentiation, lymphovascular invasion, and perineural invasion [1]. More importantly, an increasing body of literature indicates that high-grade TB serves as an independent prognostic risk factor in patients with RC [2–4].
Traditionally, patients with endoscopically resectable pT1 RC were considered to have outcomes comparable to those undergoing radical surgery. However, recent studies have suggested that, even among endoscopically resectable pT1 RC, the presence of high-grade TB is indicative of an increased risk of lymph node metastasis [5, 6]. What’s more, previous studies have shown that high-grade TB is one of the factors associated with poorer disease-free survival (DFS) in patients with stage II RC after curative resection, and that the 5-year DFS is significantly lower in patients with high-grade TB than in those with non-high-grade TB (< 10 buds, Bd 1 and 2) [7–10]. In addition, high-grade TB is regarded as a key factor in determining whether adjuvant chemotherapy should be recommended after curative resection for stage II RC. Evidence has suggested that adjuvant chemotherapy may effectively improve outcomes in patients with stage II RC with high-grade TB [11]. Therefore, accurate preoperative assessment of TB status is of great significance, as it may contribute to more informed preoperative planning and hold the potential to improve patient outcomes. Nevertheless, at present, TB can only be diagnosed based on histopathological evaluation.
Synthetic MRI (SyMRI) is an emerging quantitative MRI technique that uses a multi-dynamic multi-echo sequence to generate longitudinal relaxation time (T1), transverse relaxation time (T2), and proton density (PD) maps within a single acquisition. T1, T2, and PD values reflect fundamental tissue properties and exhibit relatively stable measurements, theoretically being less dependent on scanner settings and acquisition parameters under a given field strength [12]. Prior studies have reported that preoperative SyMRI shows favorable diagnostic performance in predicting perineural invasion, lymph node metastasis, extramural venous invasion, and T staging in RC [13, 14]. Moreover, SyMRI has demonstrated potential utility in evaluating response to neoadjuvant therapy in RC; for example, T1 and T2 values were reported to be significantly lower in patients achieving pathological complete response and those with T downstaging than in their counterparts [15]. However, to the best of our knowledge, no study has investigated the use of preoperative SyMRI for assessing TB status in RC.
Based on these considerations, we hypothesized that quantitative SyMRI parameters (T1, T2, and PD values) differ among RC patients with different TB grades. Therefore, this study aimed to investigate the feasibility and value of preoperative SyMRI-derived quantitative parameters for non-invasive assessment of TB status in RC.
Materials and methods
Patients
The retrospective study was approved by the ethics committee of the Seventh Affiliated Hospital, Sun Yat-sen University (registration number: KY-2024-360-01) and adhered to the principles of the Declaration of Helsinki. Requirement for informed consent was waived. Between March 2021 and June 2024, a total of 134 patients were included in the study. These patients underwent either total mesorectal excision surgery or endoscopic resection and were pathologically confirmed to have RC. All patients underwent rectal MRI examinations, including SyMRI sequences. The exclusion criteria were: (1) patients who had received RC-related treatment before the surgery or rectal MRI examination (n = 14); (2) an interval of more than 4 weeks between MRI examination and surgery (n = 5); and (3) images with artifacts that affected image analysis (n = 10). Ultimately, 105 RC patients were included in the study, including 63 non-high-grade TB patients and 42 high-grade TB patients.
MRI acquisition
MRI was performed using a 3.0-T whole-body MRI system (Signa Architect, GE Healthcare) with a 30-channel AIR coils. Participants were asked to undergo bowel preparation but did not receive anti-peristaltic medication before MRI. The conventional MRI protocol included T2 weighted imaging (T2WI), diffusion weighted imaging (DWI), T1 weighted imaging (T1WI) and contrast-enhanced T1WI. For SyMRI, a commercially available MAGnetic resonance image Compilation sequence with two echo times (19.8/99.0 msec) and four saturation delay times (210/610/1810/3810 msec) was used. At the level of the tumor, oblique axial scan positioning should be performed perpendicular to the long axis of the rectal. Imaging parameters were as follows: repetition time, 4000 ms; echo time, 18.7/90.2 ms; slice thickness, 3 mm; field of view, 40 mm × 40 mm; acquisition matrix, 320 × 256; number of excitation, 1; echo train length, 16; bandwidth (kHz), 27.78; scan time, 4 min 32 s. The detailed scanning parameters are shown in Table 1.
Table 1.
Parameters of rectal MRI
| Sequence | T1WI | T2WI | DWI | SyMRI | DCE |
|---|---|---|---|---|---|
| Imaging plane | Oblique axial | Oblique axial, Oblique coronal, sagittal | Oblique axial | Oblique axial | Oblique axial |
| TR (ms) | 497 | 3765, 4898, 3732 | 3815 | 4000 | 4.9 |
| TE (ms) | 9 | 95, 95, 95 | 83 | 18.7 | 1.1 |
| Slice thickness/ gap(mm) | 6/0 | 3/0 | 5/0 | 3/0 | 1/0 |
| FOV | 18 × 18 | 18 × 18, 24 × 24, 24 × 24 | 40 × 40 | 40 × 40 | 40 × 40 |
| Matrix | 384 × 256 | 320 × 320, 320 × 256, 320 × 256 | 192 × 194 | 320 × 256 | 320 × 192 |
| Echo train length | 3 | 21, 15, 23 | - | 16 | - |
| Bandwidth(kHz) | 50 | 50, 62.5, 50 | 250 | 27.8 | 142.9 |
TR, repetition time; TE, echo time; FOV, field of view; T1WI, T1 weighted imaging; T2 weighted imaging; DWI, diffusion weighted imaging; SyMRI, synthetic Magnetic Resonance Imaging; DCE, dynamic contrast enhanced imaging
Imaging analysis and measurement
All quantitative data measurements are performed on GE AW 4.7 workstation by two radiologists (KP. Z, 5 years of experience in abdominal imaging; JW. L,13 years of experience in abdominal imaging). Given that histopathological evaluation of TB is usually performed on the postoperative specimen at the invasive front of the tumor, particularly in the “hotspot” with the highest budding density [1], the study aimed to ensure that the quantitative parameters derived from SyMRI, including T1, T2, and PD values, would reflect the corresponding pathological changes as accurately as possible. Therefore, during imaging analysis, efforts were made to match the slice containing the macroscopic imaging region of interest (ROI) as closely as possible to the histopathologically evaluated slice, rather than correspondence to the “hotspot”. Specifically, the slice (recorded as slice 1) corresponding to the tumor invasive front was first identified on conventional T2WI and synthetic T2WI. Subsequently, a ROI was manually drawn along the tumor outline on synthetic T2WI, and the mean T1, T2, and PD values of the lesion were extracted and recorded from the corresponding SyMRI parametric maps. Considering the potential mismatch between the invasive-front slice selected on imaging and the actual slice assessed on histopathology, ROIs were further manually drawn on the nearest slices above (recorded as slice 2) and below (recorded as slice 3) slice 1 using the same method, and the corresponding SyMRI quantitative parameters were measured. Finally, the T1, T2, and PD values obtained from these multiple slices were averaged separately and used as the final quantitative metrics of the lesion for statistical analysis. The detailed process of acquiring quantitative data is shown in Fig. 1. The agreement of measurements for the same parameter across different slices has been evaluated, and details showed in Supplementary Materials.
Fig. 1.
Diagram illustrating the quantitative data acquisition process (GE AW 4.7 workstation, GE Healthcare, USA). Step 1, Preoperative magnetic resonance imaging is used to evaluate the position of rectal cancer (distance from the inferior part of the tumor to the transitional skin), T staging, the anal complex (including the sphincters and puborectal muscles), nodal staging, circumferential resection margin, and extramural vascular invasion. Step 2, Pathological evaluation included pT stage, pN stage, tumor differentiation, vascular invasion, nerve invasion and tumor budding. Step 3, on the synthetic T2WI, the regions of interest were manually drawn along the tumor outline at the slice (slice 2) with the greatest depth of extramural spread of rectal cancer, as well as the nearest slices above (slice 1) and below (slice 3) it. T1, T2, and PD values were measured at the slice 1, slice 2 and slice 3, separately. And then, the average values of T1, T2, and PD were calculated by the formulas on the illustration diagram, separately
Pathological analysis
All resection specimens were stained with hematoxylin and eosin (H&E) and staged according to the 8th edition TNM staging system of the American Joint Committee on Cancer (AJCC) guidelines [16]. The final pathology report included pathological T-stage, N-stage, tumor differentiation, TB, perineural invasion, lymphovascular invasion, tumor deposits, and circumferential resection margin.
TB of RC was evaluated by a pathologist (K.G) with more than 10 years of experience in pathological diagnosis following the recommendations of the ITBCC [17]. The area with the highest TB density (“hotspot”) was selected under a 20 × objective lens for evaluation. TB was then counted within a 0.785 mm² area in the hotspot. The ITBCC conversion table was used to score TB (Table 2). The scoring system is as follows: low budding (0–4 buds, Bd 1), intermediate budding (5–9 buds, Bd 2), and high budding (≥ 10 buds, Bd 3). As Bd 3 RC had a significantly worse prognosis than Bd 1 and Bd 2 RC [10]. Therefore, we defined RC in Bd 3 group as high-grade TB and RC in Bd 1 and Bd 2 group as non-high-grade TB.
Table 2.
Conversion table to adjust and standardize the tumor bud count for different microscope types. Objective magnification: 20
| Eyepiece FN Diameter (mm) | Specimen Area (mm2) | Normalization Factor |
|---|---|---|
| 18 | 0.636 | 0.810 |
| 19 | 0.709 | 0.903 |
| 20 | 0.785 | 1.000 |
| 21 | 0.866 | 1.103 |
| 22 | 0.950 | 1.210 |
| 23 | 1.039 | 1.323 |
| 24 | 1.131 | 1.440 |
| 25 | 1.227 | 1.563 |
| 26 | 1.327 | 1.690 |
Statistical analysis
Statistical analyses were performed using Statistical Package for the Social Sciences (SPSS, version 25.0, IBM) and GraphPad Prism (version 8.4.2). Continuous variables were compared between the high-grade TB and non-high-grade TB groups using independent-samples t test, whereas categorical variables were analyzed using the chi-square test or Fisher’s exact test, as appropriate. Univariable binary logistic regression was first performed to examine the associations of clinical characteristics, histopathological characteristics, and parameters of SyMRI with TB, and variables showing statistical significance were subsequently entered into a multivariable binary logistic regression model to identify independent predictors of high-grade TB. Receiver operating characteristic (ROC) curve analysis was then used to evaluate the predictive performance of each independent predictor alone and in combination, and DeLong’s test was applied to compare differences in the areas under the ROC curves (AUC). The combined model was internally adjusted using cross-validation, from which the corresponding AUC was derived, and the Akaike information criterion (AIC) was used to compare the goodness-of-fit between the combined model and models based on individual predictors. Model calibration and overall predictive accuracy were assessed using the Hosmer–Lemeshow test and the Brier score, respectively, and decision curve analysis (DCA) was finally conducted to compare the clinical net benefit and utility of the different models. All tests were two-sided with statistical significance set at p < 0.05.
Results
Clinical characteristics
A total of 105 patients with RC were included, comprising 42 patients in high-grade TB group and 63 in non-high-grade TB group. The mean age was 53.33 ± 11.81 years in high-grade TB group and 53.86 ± 12.36 years in non-high-grade TB group, with no significant difference between groups (p = 0.829). Between-group comparisons showed significant differences in tumor differentiation (p < 0.001), pathological T-stage (p < 0.001), pathological N-stage (p = 0.002), vascular invasion (p < 0.001), and nerve invasion (p = 0.003) (Table 3). Detail showed in Table 3. Results of univariate binary logistic regression analysis analyses indicated that, compared with non-high-grade TB group, high-grade TB group was more likely to exhibit poor tumor differentiation (OR [95% CI], 8.64 [3.06–24.35]; p < 0.001), higher pathological T-stage (OR [95% CI], 12.59 [3.52–45.02]; p < 0.001), and higher pathological N-stage (OR [95% CI], 2.73 [1.17–6.37]; p = 0.020); moreover, the risks of vascular invasion (OR [95% CI], 4.46 [1.93–10.31]; p < 0.001) and nerve invasion (OR [95% CI], 3.33 [1.47–7.57]; p = 0.004) were also significantly increased in patients with high-grade TB. Detail showed in Table 4.
Table 3.
Characteristics of the study participants
| Parameters | High-grade TB (n = 42) | non- High-grade TB (n = 63) | p |
|---|---|---|---|
| Age (year) | 53.33 ± 11.81 | 53.86 ± 12.36 | 0.829 |
| Gender | |||
| Male | 20 (47.62%) | 29 (46.03%) | 0.873 |
| Female | 22 (52.38%) | 34 (53.97%) | |
| Tumor differentiation | |||
| Well | 2 (4.76%) | 13 (20.63%) | <0.001 |
| Moderate | 20 (47.62%) | 44 (69.84%) | |
| Poor | 20 (47.62%) | 6 (9.52%) | |
| Pathological T-stage | |||
| T1 | 0 (0%) | 4 (6.35%) | <0.001 |
| T2 | 3 (7.14%) | 27 (42.86%) | |
| T3 | 31 (73.81%) | 29 (46.03%) | |
| T4 | 8 (19.05%) | 3 (4.76%) | |
| Pathological N-stage | |||
| N0 | 11 (26.19%) | 31 (49.21%) | 0.002 |
| N1 | 26 (61.91%) | 32 (50.79%) | |
| N2 | 5 (11.90%) | 0 (0%) | |
| Vascular invasion | |||
| positive | 29 (69.05%) | 21 (33.33%) | <0.001 |
| negative | 13 (30.95%) | 42 (66.67%) | |
| Nerve invasion | |||
| positive | 24 (57.14%) | 18 (28.57%) | 0.003 |
| negative | 18 (42.86%) | 45 (71.43%) |
TB, tumor budding
Table 4.
Results of binary logistic regression analysis
| Parameters | Univariable analysis | Multivariable analysis | |||
|---|---|---|---|---|---|
| OR (95% CI) | p | OR (95% CI) | p | ||
| T1 | 7.96 (3.70 ~ 17.12) | <0.001 | 10.03 (3.39 ~ 29.65) | <0.001 | |
| T2 | 0.21 (0.11 ~ 0.39) | <0.001 | 0.14 (0.04 ~ 0.45) | <0.001 | |
| PD | 1.12 (0.76 ~ 1.66) | 0.571 | 0.83 (0.40 ~ 1.74) | 0.624 | |
| Age (year) | 0.96 (0.65 ~ 1.42) | 0.827 | 0.60 (0.28 ~ 1.28) | 0.187 | |
| Gender | 1.07 (0.49 ~ 2.33) | 0.873 | 0.71 (0.16 ~ 3.14) | 0.656 | |
| T-stage | T1 ~ T2 | Reference | |||
| T3 ~ T4 | 12.59 (3.52 ~ 45.02) | <0.001 | 6.53 (0.64 ~ 66.93) | 0.114 | |
| N-stage | N0 | Reference | |||
| N1 ~ N2 | 2.73 (1.17 ~ 6.37) | 0.020 | 1.55 (0.25 ~ 9.55) | 0.637 | |
| Tumor differentiation | Well and Moderate | Reference | |||
| Poor | 8.64 (3.06 ~ 24.35) | <0.001 | 1.75 (0.32 ~ 9.42) | 0.515 | |
| Vascular invasion | 4.46 (1.93 ~ 10.31) | <0.001 | 0.91 (0.17 ~ 4.95) | 0.911 | |
| Nerve invasion | 3.33 (1.47 ~ 7.57) | 0.004 | 2.20 (0.45 ~ 10.66) | 0.328 | |
PD, proton density; T1, longitudinal relaxation time; T2, transverse relaxation time
SyMRI characteristics
Independent-samples t-tests showed that the T1 value was significantly higher in high-grade TB group (1528.1 ± 104.8 ms) than in non-high-grade TB group (1359.5 ± 97.9 ms) (p < 0.001), whereas the T2 value was significantly lower in high-grade TB group (80.9 ± 5.6 ms) compared with non-high-grade TB group (89.17 ± 6.5 ms) (p < 0.001). Although PD value was slightly higher in high-grade TB group (67.9 ± 12.1 pu) than in non-high-grade TB group (66.6 ± 11.6 pu), the difference was not statistically significant (p = 0.575) (Figs. 2 and 3). ROC analysis demonstrated that a T1 value cutoff of ≥ 1448.8 ms yielded an AUC of 0.883 for predicting high-grade TB, with a sensitivity of 83.3%, specificity of 84.1%, positive predictive value (PPV) of 77.8%, and negative predictive value (NPV) of 88.3%; using a T2 value cutoff of ≤ 84.4 ms resulted in an AUC of 0.838, sensitivity of 81.0%, specificity of 82.5%, PPV of 75.6%, and NPV of 86.7% (Fig. 4 Part A). DeLong’s test further indicated no significant difference between the AUCs of T1 and T2 values for predicting high-grade TB (Z = 0.796, p = 0.426).
Fig. 2.
A patient with midrectal cancer (A-C, T2 weighted imaging). T2N0, moderate differentiation, vascular invasion (negative), nerve invasion (negative) and non-high-grade tumor budding (6 buds, Bd 2 group). (D, T1 map) T1 value, 1427.4 ms; (E, T2 map) T2 value, 91.7 ms, (F, PD map) PD value, 67.1 pu
Fig. 3.
A patient with upper rectal cancer (A-C, T2 weighted imaging). T4aN0, poor differentiation, vascular invasion (negative), nerve invasion (positive) and high-grade tumor budding (12 buds, Bd 3 group). (D, T1 map) T1 value, 1549.2 ms; (E, T2 map) T2 value, 81.5 ms, (F, PD map) PD value, 66.4 pu
Fig. 4.
(Part A) Receiver operating characteristic curves for predicting high-grade tumor budding using T1 value alone, T2 value alone, and the combined T1 and T2 values model, with areas under the curve of 0.883, 0.838, and 0.926, respectively. (Part B) Calibration curves and decision curve analysis for the T1value alone model, the T2 value alone model, and the combined T1 and T2 values model in predicting high-grade tumor budding
Results of multivariate binary logistic regression analysis
Multivariable binary logistic regression identified T1 value (p < 0.001) and T2 value (p = 0.002) as independent predictors of high-grade tumor budding (TB), with an OR of 10.03 (95% CI, 3.29–29.65) for T1 value and 0.14 (95% CI, 0.04–0.45) for T2 value (Table 4). After combining T1 and T2 values and applying cross-validation, the model demonstrated good discriminatory performance for high-grade TB, achieving an AUC of 0.926; compared with the model T1 value alone (AIC = 92.77) or T2 value alone (AIC = 108.18), the combined model yielded a lower akaike information criterion (AIC) (AIC = 72.22), indicating a better fit. Calibration assessment further showed a higher calibration slope for the combined model (slope, 0.924; intercept, − 0.013; Eavg, 0.0656; Emax, 0.1262) than for the T1 value alone model (slope, 0.885; intercept, − 0.034; Eavg, 0.0665; Emax, 0.1801) and the T2 value alone model (slope, 0.877; intercept, − 0.024; Eavg, 0.1075; Emax, 0.2978). The Hosmer–Lemeshow test and Brier score indicated no significant lack of calibration for any model, including the combined model (HL χ² = 6.601, p = 0.580; Brier score = 0.107), the T1 value alone model (HL χ² = 6.127, p = 0.633; Brier score = 0.135), and the T2 value alone model (HL χ² = 12.851, p = 0.117; Brier score = 0.168); nevertheless, the combined model showed the best overall predictive accuracy, as reflected by the lowest Brier score. In terms of clinical utility, decision curve analysis (DCA) demonstrated that the combined T1 and T2 values model provided the highest net benefit across most threshold probabilities, outperforming the T1 value alone and T2 value alone models and yielding greater benefit than the treat-all and treat-none strategies. Detail showed in Table 5; Fig. 4.
Table 5.
Differences of quantitative parameters among different tumor budding group
| Parameters | Combined T1 and T2 values | T1value | T2 value | Treat-all |
|---|---|---|---|---|
| pt = 0.05 | 0.3739 | 0.3644 | 0.3624 | 0.3684 |
| pt = 0.10 | 0.3630 | 0.3450 | 0.3302 | 0.3333 |
| pt = 0.20 | 0.3214 | 0.3000 | 0.2738 | 0.2500 |
| pt = 0.30 | 0.2993 | 0.2680 | 0.2531 | 0.1429 |
| pt = 0.40 | 0.2730 | 0.2635 | 0.2222 | 0.0000 |
| pt = 0.50 | 0.2476 | 0.2286 | 0.2095 | -0.2000 |
T1, longitudinal relaxation time; T2, transverse relaxation time
Discussion
The potential value of preoperative TB prediction is not to replace postoperative histopathological evaluation, but rather to provide additional information on tumor aggressiveness before treatment, beyond what TNM staging alone can capture, thereby offering potential support for individualized patient management. The study demonstrated that parameters T1 and T2 values of SyMRI were both independent predictors of high-grade TB. A simplified combined model incorporating T1 and T2 values showed favorable predictive performance for high-grade TB and overall outperformed the single-parameter models (T1 value or T2 value alone) in terms of model fit, calibration performance, and clinical net benefit.
Previous studies have shown that high-grade TB is related not only to poorer tumor differentiation but also to higher T- and N-stages, and higher rates of vascular invasion and perineural invasion [18–21]. Our findings are consistent with these observations and further support the close association between high-grade TB and increased tumor aggressiveness.
In the present study, high-grade TB group showed significantly lower T2 value and significantly higher T1 value than non–high-grade TB group. Previous studies have shown that poorly differentiated or lymph node-positive RC tend to exhibit lower T2 value and higher T1 value, which indirectly supports our findings [22–25]. A possible explanation is that more aggressive tumors are often characterized by increased cellularity, enhanced inflammatory infiltration, and alterations in extracellular components, leading to reduced free-water content and relatively increased macromolecular content, ultimately resulting in lower T2 and higher T1 values [26–30]. In our cohort, the high-grade TB group was also associated with poorer differentiation, higher T- and N-stages, and higher rates of vascular and perineural invasion, which provides a plausible explanation for these findings. Notably, Zhao et al. [14]. reported lower T1 value in RC with positive lymph node metastasis and extramural venous invasion, whereas Ma et al. [24]. found no significant association between T1 value and relevant prognostic subgroups. These findings are not entirely consistent with our results. One possible explanation is the difference in measurement strategy across studies. Zhao et al. [14]. used a whole-tumor approach, and Ma et al. [24]. delineated ROI on the slice with the largest tumor diameter, whereas our study focused on the invasive front to better correspond to the region of TB evaluated by histopathology. Whole-tumor or largest-slice measurements may reflect more global tumor characteristics, whereas invasive-front assessment may be more sensitive to localized microenvironmental changes associated with tumor invasiveness.
Regarding PD value, previous studies have suggested its potential value in the diagnosis and grading of several malignancies, but findings in RC remain inconsistent. Ma et al. [24]. reported lower PD value in poorly differentiated RC, whereas Zhao et al. [14]. found no clear association between PD and tumor differentiation. In the present study, PD value did not differ significantly between high-grade TB and non–high-grade TB groups, suggesting that the incremental value of PD for TB stratification may be limited and more susceptible to cohort composition and measurement-related factors.
At the model level, multivariable binary logistic regression analysis identified both T1 and T2 values as independent predictors of high-grade TB. Although the area under the curve of the T1 model was numerically higher than that of the T2 model, the difference was not statistically significant. In contrast, the simplified combined model integrating T1 and T2 values demonstrated better discriminative ability after cross-validation and overall outperformed the single-parameter models in terms of goodness-of-fit, calibration performance, and clinical net benefit. These findings suggest that multiparametric integration may capture the histopathological heterogeneity associated with tumor budding more comprehensively than a single parameter alone, thereby improving model robustness and potential clinical utility.
Several limitations should be acknowledged. First, a key limitation of this study is the discrepancy between microscopic pathological assessment and macroscopic MRI-based ROI placement. TB is evaluated histopathologically in a microscopic hotspot field of 0.785 mm², whereas such a hotspot cannot be directly visualized or precisely localized on preoperative MRI. Therefore, ROI placement in the present study was based on a macroscopic imaging strategy within the visible tumor region rather than true hotspot localization. This limitation may affect reproducibility and may partly contribute to the diagnostic performance observed in this retrospective cohort. Second, despite efforts to match SyMRI measurements with pathological slices, imperfect spatial correspondence may persist (e.g., differences between histological section thickness and MRI slice thickness), potentially affecting accuracy and reproducibility. Third, only patients undergoing total mesorectal excision or endoscopic resection were included, whereas those receiving neoadjuvant chemoradiotherapy were excluded. While this design may improve internal consistency by minimizing treatment-related confounding in pathological assessment, it may introduce selection bias and reduce external applicability. Fourth, this study did not distinguish between peritumoral budding and intratumoral budding for separate analysis, which may have limited the comprehensiveness of the pathological evaluation and reduced the clinical applicability of the findings to some extent. Fifth, this was a single-center study with a relatively small sample size; although cross-validation was applied to mitigate overfitting, validation in larger cohorts is still needed. Sixth, the lack of external validation limits the generalizability of our findings. Future multi-center, prospective studies with external validation and more rigorous spatial co-registration strategies are warranted to further strengthen reliability and facilitate clinical translation.
Conclusion
The study demonstrates that compared with non–high-grade TB group, patients with high-grade TB exhibited significantly higher T1 value and significantly lower T2 value. Multivariable binary logistic regression further confirmed that both T1 and T2 values were independent predictors of high-grade TB. A simplified combined model incorporating T1 and T2 value achieved superior discrimination for high-grade TB and showed overall advantages over single-parameter models (T1 or T2 values alone) in terms of model fit, calibration performance, and clinical net benefit. Collectively, these preliminary findings suggest that parameters of T1 and T2 values of SyMRI, particularly when combined, may serve as potential preoperative indicators for identifying patients at higher risk of high-grade TB. However, further prospective studies and external validation are required.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
Thanks are due to Jia-Wei Hu for assistance with the experiments.
Abbreviations
- TB
Tumor budding
- RC
Rectal cancer
- SyMRI
Synthetic MRI
- T1
Longitudinal relaxation time
- T2
Transverse relaxation time
- PD
Proton density
- ROC
Receiver operating characteristic
- AUC
Under the ROC curves
- AIC
Akaike information criterion
- DCA
Decision curve analysis
Author contributions
Conceptualization, KP Z, QY L and DM L; Data curation, HB H, XW C and WJ L; Formal analysis, J B; Investigation, KP Z, LP L and XW C; Methodology, KP Z, QY L, HB Hand DM L; Project administration, KP Z and QY L; Software, ZX L and LH C; Supervision, KP Z and QY L; Validation, XW C, WJ L and J B; Visualization, XW C; Writing – original draft, KP Z, HB H, XW C and WJ L; Writing – review & editing, KP Z, QY L, DM L and WJ L. All authors reviewed the manuscript.
Funding
The authors did not receive support from any organization for the submitted work.
Data availability
The data that support the findings of this study are available from the authors but restrictions apply to the availability of these data, which were used under license from the Seventh Affiliated Hospital, Sun Yat-sen University for the current study, and so are not publicly available. Data are, however, available from the authors upon reasonable request and with permission from the Seventh Affiliated Hospital, Sun Yat-sen University.
Declarations
Ethics approval and consent to participate
This retrospective study was conducted in accordance with the ethical principles of the Declaration of Helsinki and approved by the ethics committee of the Seventh Affiliated Hospital, Sun Yat-sen University (registration number: KY-2024-360-01). Requirement for informed consent was waived.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Kun-Peng Zhou, Hua-Bin Huang and Xian-Wen Cheng contributed equally to this work.
Contributor Information
Wei-Jian Lai, Email: laiweijian@sysush.com.
Di-Min Liu, Email: liudimin@sysush.com.
Qing-Yu Liu, Email: liuqingyu@sysush.com.
References
- 1.Lugli A, Kirsch R, Ajioka Y, Bosman F, Cathomas G, Dawson H, et al. Recommendations for reporting tumor budding in colorectal cancer based on the International Tumor Budding Consensus Conference (ITBCC) 2016. Mod Pathol. 2017;30(9):1299–311. [DOI] [PubMed]
- 2.De Smedt L, Palmans S, Sagaert X. Tumour budding in colorectal cancer: what do we know and what can we do? Virchows Arch. 2016;468(4):397–408. [DOI] [PubMed] [Google Scholar]
- 3.Lugli A, Karamitopoulou E, Zlobec I. Tumour budding: a promising parameter in colorectal cancer. Br J Cancer. 2012;106(11):1713–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Rogers AC, Winter DC, Heeney A, Gibbons D, Lugli A, Puppa G, et al. Systematic review and meta-analysis of the impact of tumour budding in colorectal cancer. Br J Cancer. 2016;115(7):831–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Chida K, Watanabe J, Hirasawa K, Inayama Y, Misumi T, Kunisaki C, et al. A novel risk-scoring system for predicting lymph node metastasis of rectal neuroendocrine tumors. Ann Gastroenterol Surg. 2020;4(5):562–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Koelzer VH, Zlobec I, Lugli A. Tumor budding in colorectal cancer–ready for diagnostic practice? Hum Pathol. 2016;47(1):4–19. [DOI] [PubMed] [Google Scholar]
- 7.Barresi V, Reggiani Bonetti L, Ieni A, Branca G, Tuccari G. Histologic prognostic markers in stage IIA colorectal cancer: a comparative study. Scand J Gastroenterol. 2016;51(3):314–20. [DOI] [PubMed] [Google Scholar]
- 8.Noguchi M, Shitara K, Kawazoe A, Yamamoto D, Takii Y, Saito Y, et al. Short-term safety of adjuvant chemoradiotherapy after local resection for patients with high-risk submucosal invasive rectal cancer: a single-arm, multicenter phase II trial. Jpn J Clin Oncol. 2021;51(5):707–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Ammendola S, Turri G, Marconi I, Burato G, Pecori S, Tomezzoli A, et al. The presence of poorly differentiated clusters predicts survival in stage II colorectal cancer. Virchows Arch. 2021;478(2):241–8. [DOI] [PubMed] [Google Scholar]
- 10.Kyong Shin J, Ah Park Y, Wook Huh J, Hyeon Yun S, Cheol Kim H, Yong Lee W, et al. Is High-Grade Tumor Budding an Independent Prognostic Factor in Stage II Colon Cancer? Dis Colon Rectum. 2023;66(8):e801–8. [DOI] [PubMed] [Google Scholar]
- 11.Kim S, Huh JW, Lee WY, Yun SH, Kim HC, Cho YB, et al. Prognostic Impact of Lymphatic Invasion, Venous Invasion, Perineural Invasion, and Tumor Budding in Rectal Cancer Treated With Neoadjuvant Chemoradiotherapy Followed by Total Mesorectal Excision. Dis Colon Rectum. 2023;66(7):905–13. [DOI] [PubMed] [Google Scholar]
- 12.Warntjes JB, Leinhard OD, West J, Lundberg P. Rapid magnetic resonance quantification on the brain: Optimization for clinical usage. Magn Reson Med. 2008;60(2):320–9. [DOI] [PubMed] [Google Scholar]
- 13.Zhao L, Liang M, Wu PY, Yang Y, Zhang H, Zhao X. A preliminary study of synthetic magnetic resonance imaging in rectal cancer: imaging quality and preoperative assessment. Insights Imaging. 2021;12(1):120. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Zhao L, Liang M, Xie L, Yang Y, Zhang H, Zhao X. Prediction of pathological prognostic factors of rectal cancer by relaxation maps from synthetic magnetic resonance imaging. Eur J Radiol. 2021;138:109658. [DOI] [PubMed] [Google Scholar]
- 15.Lian S, Liu H, Meng T, Ma L, Zeng W, Xie C. Quantitative synthetic MRI for predicting locally advanced rectal cancer response to neoadjuvant chemoradiotherapy. Eur Radiol. 2023;33(3):1737–45. [DOI] [PubMed] [Google Scholar]
- 16.Amin MB, Greene FL, Edge SB, Compton CC, Gershenwald JE, Brookland RK, et al. 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(2):93–9. [DOI] [PubMed] [Google Scholar]
- 17.Dawson H, Galuppini F, Trager P, Berger MD, Studer P, Brugger L, et al. Validation of the International Tumor Budding Consensus Conference 2016 recommendations on tumor budding in stage I-IV colorectal cancer. Hum Pathol. 2019;85:145–51. [DOI] [PubMed]
- 18.Mitrovic B, Schaeffer DF, Riddell RH, Kirsch R. Tumor budding in colorectal carcinoma: time to take notice. Mod Pathol. 2012;25(10):1315–25. [DOI] [PubMed] [Google Scholar]
- 19.Lugli A, Zlobec I, Berger MD, Kirsch R, Nagtegaal ID. Tumour budding in solid cancers. Nat Rev Clin Oncol. 2021;18(2):101–15. [DOI] [PubMed] [Google Scholar]
- 20.van Wyk HC, Roseweir A, Alexander P, Park JH, Horgan PG, McMillan DC, et al. The Relationship Between Tumor Budding, Tumor Microenvironment, and Survival in Patients with Primary Operable Colorectal Cancer. Ann Surg Oncol. 2019;26(13):4397–404. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.van Wyk HC, Park JH, Edwards J, Horgan PG, McMillan DC, Going JJ. The relationship between tumour budding, the tumour microenvironment and survival in patients with primary operable colorectal cancer. Br J Cancer. 2016;115(2):156–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Ge YX, Hu SD, Wang Z, Guan RP, Zhou XY, Gao QZ, et al. Feasibility and reproducibility of T2 mapping and DWI for identifying malignant lymph nodes in rectal cancer. Eur Radiol. 2021;31(5):3347–54. [DOI] [PubMed] [Google Scholar]
- 23.Zhang J, Ge Y, Zhang H, Wang Z, Dou W, Hu S. Quantitative T2 Mapping to Discriminate Mucinous from Nonmucinous Adenocarcinoma in Rectal Cancer: Comparison with Diffusion-weighted Imaging. Magn Reson Med Sci. 2022;21(4):593–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Ma L, Lian S, Liu H, Meng T, Zeng W, Zhong R, et al. Diagnostic performance of synthetic magnetic resonance imaging in the prognostic evaluation of rectal cancer. Quant Imaging Med Surg. 2022;12(7):3580–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Zhao L, Liang M, Shi Z, Xie L, Zhang H, Zhao X. Preoperative volumetric synthetic magnetic resonance imaging of the primary tumor for a more accurate prediction of lymph node metastasis in rectal cancer. Quant Imaging Med Surg. 2021;11(5):1805–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Fernandes JL, Rochitte CE. T1 mapping: technique and applications. Magn Reson Imaging Clin N Am. 2015;23(1):25–34. [DOI] [PubMed] [Google Scholar]
- 27.Wang S, Li J, Zhu D, Hua T, Zhao B. Contrast-enhanced magnetic resonance (MR) T1 mapping with low-dose gadolinium-diethylenetriamine pentaacetic acid (Gd-DTPA) is promising in identifying clear cell renal cell carcinoma histopathological grade and differentiating fat-poor angiomyolipoma. Quant Imaging Med Surg. 2020;10(5):988–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Meng T, He N, He H, Liu K, Ke L, Liu H, et al. The diagnostic performance of quantitative mapping in breast cancer patients: a preliminary study using synthetic MRI. Cancer Imaging. 2020;20(1):88. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Adams LC, Jurmeister P, Ralla B, Bressem KK, Fahlenkamp UL, Engel G, et al. Assessment of the extracellular volume fraction for the grading of clear cell renal cell carcinoma: first results and histopathological findings. Eur Radiol. 2019;29(11):5832–43. [DOI] [PubMed] [Google Scholar]
- 30.Li J, Gao X, Dominik Nickel M, Cheng J, Zhu J. Native T1 mapping for differentiating the histopathologic type, grade, and stage of rectal adenocarcinoma: a pilot study. Cancer Imaging. 2022;22(1):30. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are available from the authors but restrictions apply to the availability of these data, which were used under license from the Seventh Affiliated Hospital, Sun Yat-sen University for the current study, and so are not publicly available. Data are, however, available from the authors upon reasonable request and with permission from the Seventh Affiliated Hospital, Sun Yat-sen University.