Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Aug 25.
Published in final edited form as: Eur Radiol. 2025 Aug 25;36(3):1827–1837. doi: 10.1007/s00330-025-11945-y

Lymph nodes in MRI-based early-stage rectal cancer: can MRI accurately predict lymph node involvement?

Maria Clara Fernandes a, Yousef Mazaheri b, James Louis Fuqua a, Natally Horvat c, Emmanouil Pappou d, Junting Zheng e, Marinela Capanu e, Lee Rodriguez a, Marc J Gollub a, Jennifer S Golia Pernicka a
PMCID: PMC12945439  NIHMSID: NIHMS2117815  PMID: 40853592

Abstract

Objectives:

To investigate if quantitative DWI and qualitative T2WI parameters can predict lymph node involvement in clinical early rectal cancer.

Materials and methods:

This retrospective study included consecutive patients who had rectal MRI from January 1, 2010-March 31, 2021 showing T1-T2 rectal cancer before undergoing total mesorectal excision without neoadjuvant therapy. Visible mesorectal lymph nodes on DWI were segmented by a junior reader and ADC values were extracted. Additionally, the junior reader and a senior reader independently recorded qualitative T2WI parameters of the most suspicious lymph node per patient. Histopathology was the reference standard for malignant (pN+) and benign (pN−) lymph node status.

Results:

Of 69 patients (37 men and 32 women) (median age, 55 years; IQR: 48–66]), 23 (33%) had pN+ status; 46 (67%) had pN− status. ADC mean (in × 10−3 mm2/s) of the most suspicious lymph node was not different between the two patient groups (junior reader: 1,087 vs. 926, p=0.31; senior reader: 1,178 vs. 1,086, p=0.89). The Dutch criteria, based on the combination of T2WI size and morphologic parameters, showed better diagnostic performance for the senior vs. junior reader: accuracy, sensitivity, specificity, PPV, and NPV of 79.7% (95% CI: 68.3–88.4%), 56.5% (95% CI: 34.5–76.8%), 91.3% (95% CI: 79.2–97.6%), 76.5% (95% CI: 50.1–93.2%), and 80.8% (95% CI: 67.5–90.4%), vs. 69.6% (95% CI: 57.3–80.1%), 26.1% (95% CI: 12.2–48.4%), 91.3% (95% CI: 79.2–97.6%), 60% (95% CI: 26.2–97.8%), and 71.2% (95% CI: 57.9–82.2%).

Conclusions:

Additional research of alternative and more objective methods for lymph node characterization is needed.

Keywords: Rectal cancer, early-stage, nodal staging, magnetic resonance imaging, diffusion-weighted imaging

Introduction

Colorectal cancer is the third most common cancer and the second leading cause of death due to cancer worldwide [1]. In the United States, the widespread acceptance of colonoscopy in the 1990s led to the trend towards earlier stage diagnoses. For instance, the proportion of colorectal cancer cases diagnosed at a localized stage rose from 33% in 1995 to 41% in 2005 [2]. Presently, due to increasing screening programs and the growing aging population, a further rise in the number of early rectal cancer (ERC) diagnoses is anticipated [3; 4]. This shift toward earlier detection highlights the growing importance of accurate lymph node evaluation in ERCs, which is crucial for guiding treatment and improving patient outcomes.

To date, the definition of ERC has not been standardized, and the optimal approach to managing ERC remains undefined. In 2015, the European Association for Endoscopic Surgery and the European Society of Coloproctology developed a consensus [5], defining ERC as “rectal cancer with good prognostic features that might be safely removed preserving the rectum and that will have a very limited risk of relapse after local excision.” ERC is certainly distinct from “locally advanced rectal cancer” which is rectal cancer that has extended beyond the rectal wall (T3 and T4) or that has involved the lymph nodes [6]. Regarding T1 and T2 rectal cancer, while some T1 and T2 rectal cancers may be considered ERC, it must be noted that high rates of lymph node positivity and recurrence have been reported for these tumors [7].

Local excision procedures include transanal excision, polypectomies, and transanal endoscopic microsurgery; these procedures have surged as alternatives to total mesorectal excision (TME) in selected patients with ERC [8]. In patients with ERC, the oncological outcomes following TME have been outstanding, with a cancer-specific survival rate exceeding 80% and a local recurrence rate below 5% at 6-year follow-up [9]. However, TME exposes patients to the inherent risks of radical resection, including anastomotic leak or permanent colostomy; wound dehiscence; bowel, urinary, and sexual dysfunction; and increased risk of mortality [10; 11].

Compared with TME, local excision promises better quality of life in selected patients with ERC. However, during local excision, the mesorectum is not resected typically, and therefore the lymph nodes are also not removed. As such, the postoperative nodal status is uncertain and the procedures risk leaving positive lymph nodes behind that can increase the risk for local recurrence and negatively affect long-term survival [12]. On the other hand, in the preoperative setting, the presence of lymph node involvement is an indication for neoadjuvant therapy [13]. As neoadjuvant therapy adds potential toxicity, accurate preoperative nodal assessment is crucial.

MRI is routinely used for local staging of rectal cancer but has a limited sensitivity and specificity of 58–77% and 62–74% for nodal staging [14–16]. Presently, MRI-based nodal assessment primarily relies on lymph node size and qualitatively assessed morphologic criteria [17]. As such, other MRI criteria, such as diffusion-weighted imaging (DWI) criteria, have been investigated to assess lymph node status but have not been fully validated [18–21].

The primary aim of this study was to investigate if quantitative analysis of mesorectal lymph nodes on preoperative DWI could be helpful in predicting lymph node involvement in clinical ERC. The secondary aim was to investigate preoperative T2-weighted imaging (T2WI) lymph node size and morphologic criteria in predicting lymph node involvement in clinical ERC.

Materials and Methods

Study Sample

This retrospective single-center study was approved by the institutional review board at Memorial Sloan Kettering Cancer Center; the need for informed consent was waived. The institutional database was searched to identify consecutive patients meeting the following inclusion criteria: diagnosed with rectal cancer, had TME without neoadjuvant treatment, and had rectal MRI showing T1 or T2 rectal cancer within 3 months prior to TME, in the period from January 1, 2010, to March 31, 2021. The exclusion criteria included: mucinous adenocarcinoma, secondary involvement of the rectum from a non-rectal malignancy, histopathology other than adenocarcinoma, and the presence of artifacts on DWI or T2WI.

MRI Protocol

All rectal MRI scans were acquired on GE Healthcare platforms (1.5 T or 3.0 T GE Discovery MR750, GE Optima MR450w, GE Signa EXCITE, and GE Signa HDxt) using a body coil for excitation and a flexible surface phased-array coil for reception. The main sequences evaluated for this study included axial DWI and oblique T2WI without fat suppression perpendicular to the long axis of the rectum. The standard frequency encoding direction was set to the left–right direction. Axial DW images were acquired using the single-shot echo planar imaging technique with b-values of 0 and 800 s/mm2. The technical parameters of all DW sequences were TR = 3200–5730 ms, TE = 60–112 ms, slice thickness = 5 mm, slice gap = 1 mm, and matrix = 128 × 128.

To illustrate inter-unit variability in ADC values, Supplementary Figure 1 presents data from a small-scale comparison involving three healthy volunteers scanned on three different GE MRI units. Despite the limited sample size, the ADC values remained within a relatively narrow range, supporting the feasibility of ADC analysis across multiple units when acquisition protocols are standardized.

MRI Assessment

For the assessment of lymph node involvement on DWI, one radiologist (a junior reader with 4 years of experience in interpreting rectal MRI), blinded to pathology results, manually segmented the region of interest (ROI) of all mesorectal lymph nodes per patient that were visible on axial DW images using open-source software (ITK-SNAP, version 3.4.0; http://itksnap.org). ROIs were drawn by delineating the margin of the entire node and copied to the same location for all diffusion sequences and the ADC map. ADCmean of the most suspicious lymph node was determined, as well as ADCmean, ADCmin, and ADCmax across all segmented lymph nodes at the patient level.

For the assessment of lymph node involvement on T2WI, two radiologists (the junior reader [MCF] with 4 years of experience and a senior reader [JLF] with 16 years of experience) assessed oblique T2W images independently from each other and blinded to pathology results. The following features were recorded for the most suspicious lymph node per patient: short axis diameter (cm); morphologic features including internal signal characteristic (heterogeneous/homogeneous), border contour (regular/irregular), and shape (round or non-round); and chemical shift artifact (present/absent). The most suspicious lymph node for each patient was defined by each reader taking into account a combination of size and morphological characteristics; the most suspicious was the one with more suspicious features, and if no suspicious features were present or more than one lymph node had similar number of suspicious features, the largest one was considered for assessment. The Dutch criteria [22] for baseline assessment of lymph node involvement were also used to determine the presence of suspicious lymph nodes, which are based on the combination of short axis measurement and morphologic criteria (with heterogeneous signal intensity, irregular borders, and round shape considered suspicious features). A lymph node was considered suspicious if it was < 5 mm in the short axis with all three suspicious morphologic criteria present; 5–9 mm in the short axis with the presence of at least two suspicious morphologic criteria; or > 9 mm in the short axis regardless of morphology. Chemical shift artifact was characterized by the presence of dark and light bands along the frequency axis at fat/water interfaces.

Reference Standard

Histopathology of resected lymph nodes following TME served as the reference standard for lymph node involvement, using the pN staging system to classify nodal metastasis (pN+ indicates the presence of lymph node involvement while pN− indicates the absence of lymph node involvement).

Statistical Analysis

DW ADC values (ADCmean, ADCmin, and ADCmax) and T2WI criteria were compared at the patient level between patients with pN+ status and patients with pN− status using the Wilcoxon rank sum test or the Fisher’s exact test as appropriate. P values < 0.05 were considered significant.

Intraclass correlation coefficient (ICC) and kappa (κ) statistics were used to assess reader agreement on continuous and categorical imaging features, respectively. ICC was interpreted as follows: ICC < 0.5 indicates poor agreement, ICC between 0.5 and 0.75 indicates moderate agreement, ICC between 0.75 and 0.9 indicates good agreement, and ICC > 0.9 indicates excellent agreement. Kappa (κ) values were interpreted as follows: κ < 0 indicates no agreement, κ between 0 and 0.2 indicates slight agreement, κ between 0.21 and 0.4 indicates fair agreement, κ between 0.41 and 0.6 indicates moderate agreement, κ between 0.61 and 0.8 indicates substantial agreement, and κ between 0.81 and 1.0 indicates almost perfect agreement. The statistical guarantor for the statistical analysis in this study is Junting Zheng, MS. All statistical analyses were performed using the software R (The R Foundation for Statistical Computing) version 4.3.1 including packages gtsummary and irr.

Results

Patient Characteristics

The study sample comprised 69 patients, 37 (53.6%) men and 32 (46.4%) women, with a median age of 55 years (interquartile range [IQR]: 48–66 years). The median interval between MRI and TME was 24 days (IQR: 12–32 days). On surgical pathology, 28 patients (40.6%) had T1 tumors, 21 patients (30.4%) had T2 tumors, and 20 patients had T3 tumors (29.0%). Twenty-three patients (33.3%) showed the presence of lymph node involvement on pathology (pN+), including 6/28 of patients with T1 tumors (21.4%), 6/21 of patients with T2 tumors (28.6%), and 11/20 of patients with T3 tumors (55.0%). Patient characteristics are summarized in Table 1. The flow of patient inclusion is given in Figure 1.

Table 1.

Patient characteristics (n = 69).

CLINICAL DATA
Age, median (IQR) 55 years (48–66 years)
Female 32 (46%)
Overall (n= 69) pN− (n=46) pN+ (n= 23)
mrT stage
 T1 1 (1.4%) 1 (2.2%) 0 (0%)
 T1/T2 51 (73.9%) 34 (73.9%) 17 (73.9%)
 T2 17 (24.6%) 11 (23.9%) 6 (26.1%)
pT stage
 T1 28 (40.6%) 22 (47.8%) 6 (26.1%)
 T2 21 (30.4%) 15 (32.6%) 6 (26.1%)
 T3 20 (29.0%) 9 (19.6%) 11 (47.8%)
Distance from anal verge (cm)*, median (IQR) 8.60 (6.40, 9.80) 8.00 (6.03, 9.75) 8.80 (7.25, 10.95)
Open in a new tab

QR = Interquartile range.

*

Based on senior reader measurements

Figure 1.

Figure 1.

Open in a new tab

Flow of patient inclusion into the study.

Relationship Between Preoperative DWI ADC and Lymph Node Involvement Based on Postoperative Histopathology

The results of preoperative DWI assessment are summarized in Table 2.

Table 2.

Lymph node features on preoperative MRI and lymph node involvement on postoperative histopathology.

Junior reader Senior reader
Characteristic pN− (n=46) pN+ (n = 23) p-value pN− (n=46) pN+ (n = 23) p-value
Short axis, cm, Median (IQR) 0.40 (0.30, 0.50) 0.40 (0.40, 0.60) 0.21 0.50 (0.40, 0.50) 0.50 (0.40, 0.60) 0.08
Shape 0.05 0.02
 Ovoid 36 (78.3%) 12 (52.2%) 32 (69.6%) 9 (39.1%)
 Round 10 (21.7%) 11 (47.8%) 14 (30.4%) 14 (60.9%)
Heterogeneity 0.005 <0.001
 Absent 43 (93.5%) 15 (65.2%) 41 (89.1%) 9 (39.1%)
 Present 3 (6.5%) 8 (34.8%) 5 (10.9%) 14 (60.9%)
Contour 0.001 0.002
 Irregular 1 (2.2%) 7 (30.4%) 2 (4.3%) 8 (34.8%)
 Regular 45 (97.8%) 16 (69.6%) 44 (95.7%) 15 (65.2%)
Chemical shift 0.07 <0.001
 Absent 13 (28.3%) 12 (52.2%) 14 (30.4%) 18 (78.3%)
 Present 33 (71.7%) 11 (47.8%) 32 (69.6%) 5 (21.7%)
Mean ADC, in × 10−3 mm2/s, median (IQR) 926 (743, 1,365) 1,087 (872, 1,425) 0.31 1,086 (886, 1,566) 1,178 (757, 1,445) 0.89
Dutch criteria 0.07 <0.001
 Negative 42 (91.3%) 17 73.9%) 42 (91.3%) 10 (43.5%)
 Positive 4 (8.7%) 6 (26.1%) 4 (8.7%) 13 (56.5%)
Open in a new tab

There was no significant difference in ADCmean, ADCmin, and ADCmax (× 10−3 mm2/s) across all segmented lymph nodes between the two groups of patients (1.00, 0.36, and 1.70 for patients with pN+ status, and 0.91, 0.30, and 1.49 for patients with pN− status; p = 0.28–0.68). Figure 2 shows a plot graphic of ADCmin and ADCmean in pN− and patients with pN+ status.

Figure 2.

Figure 2.

Open in a new tab

Plot graphic of minimum apparent diffusion coefficient (ADCmin) and mean ADC (ADCmean) in patients with pathological lymph node positive (pN+) result and patients with pathological lymph node negative (pN−) result.

For the junior reader, the ADC mean (in 10−3 mm2/s) of the most suspicious lymph node was 1,087 (IQR: 872–1,425) for patients with pN+ status while it was 926 (IQR: 743–1,365) for patients with pN− status. For the senior reader, the ADC mean was 1,178 (IQR: 757–1,445) for pN+ patients while it was 1,086 (IQR: 886–1,566) for pN− patients. There was no significant difference in ADCmean between the two groups of patients for both readers (p = 0.31 and 0.89, respectively). Figure 3 shows the images of two patients with morphologically similar lymph nodes on MRI that had different pathological outcomes: the patient with pN+ status had a higher ADC value than the patient with pN− status.

Figure 3.

Figure 3.

Open in a new tab

Images of two different patients with morphologically similar lymph nodes on MRI but different results on pathology. Apparent diffusion coefficient (ADC) value was higher in the patient with the pathological node positive (pN+) result. T2-weighted image (a), diffusion-weighted image (b), and ADC map (c) of a 64-year-old male patient with clinical T1/T2 stage mid/upper rectal tumor. The largest lymph node was characterized as homogenous, with regular contour and a short axis of 0.4 cm, consistent with negative lymph node per the Dutch criteria. ADC measurement was 647. Surgical pathology showed negative lymph node. T2-weighted image (d), diffusion-weighted image (e), and ADC map (f) of a 50-year-old male patient with a clinical T1/T2 stage mid rectal tumor. The largest lymph node was characterized as homogenous, with regular contour and a short axis of 0.4 cm, consistent with negative lymph node per the Dutch criteria. ADC measurement was 2,776. Surgical pathology showed positive lymph node.

Relationship Between Preoperative T2WI Features and Lymph Node Involvement Based on Postoperative Histopathology

The results of preoperative T2WI assessment are also summarized in Table 2.

Regarding the short axis diameter on T2WI, for reader 1, it was 0.4 cm (IQR: 0.3–0.5) for patients with pN+ status while it was 0.4 cm (IQR: 0.4–0.6) for patients with pN− status. For reader 2, the median short axis diameter was 0.5 cm (IQR: 0.4–0.5) for patients with pN+ status while it was 0.5 cm (IQR: 0.4–0.6) for patients with pN− status. There was no significant difference in terms of the median short axis diameter between the two groups of patients for both readers (p = 0.21 and 0.08, respectively). Lymph node size was stratified by short-axis diameter as follows: 42 (60.9%) and 29 (42.0%) measured < 5 mm, 26 (37.7%) and 39 (56.6%) measured 5–9 mm, and 1 (1.4%) and 1 (1.4%) were ≥ 9 mm for reader 1 and reader 2, respectively.

Regarding T2WI morphological features, round shape, heterogenous signal, and irregular contour were more common in patients with pN+ status for both readers (all p ≤ 0.05).

The presence of a chemical shift artifact was more common in patients with pN− status for the senior reader (p < 0.001) but not for the junior reader (p = 0.05). Figure 4 shows two cases where chemical shift artifact would be helpful to differentiate nodal metastasis from benign lymph nodes.

Figure 4.

Figure 4.

Open in a new tab

T2-weighted images of two different patients. Small superior rectal node (0.6 × 0.4 cm) with negative chemical shift artifact (a). Pathology showed positive nodal metastases. Mesorectal node (1.3 × 1.0 cm) with chemical shift artifact (b). Pathology was negative for nodal metastases.

The Dutch criteria were able to differentiate patients with pN+ status from patients with pN− status for the senior reader (p < 0.001) but not for the junior reader (p = 0.073). The diagnostic performance of the Dutch criteria is summarized in Table 3. For the senior reader, the sensitivity of the Dutch criteria was 56.5% while it was 26.1% for the junior reader. Notably, the Dutch Criteria had a high specificity of 91.3% for both readers. The accuracy of the Dutch criteria was 79.7% (95% CI: 68.3, 88.4%) for the senior reader and 69.6% (95% CI: 57.3, 80.1%) for the junior reader. The false negative rate and the false positive rate were 28.8% (95% confidence interval [CI]: 17.8%, 42.1%) and 40.0% (95% CI: 12.2%, 73.8%) for the junior reader, and 19.2% (95% CI: 9.6%, 32.5%) and 23.5% (95% CI: 6.8%, 49.9%) for the senior reader.

Table 3.

Diagnostic performance of the Dutch criteria to predict lymph node involvement on postoperative histopathology.

Junior reader (95% CI) Senior reader (95% CI)
Accuracy 69.6 (57.3, 80.1) 79.7 (68.3, 88.4)
Sensitivity 26.1 (10.2, 48.4) 56.5 (34.5, 76.8)
Specificity 91.3 (79.2, 97.6) 91.3 (79.2, 97.6)
PPV 60.0 (26.2, 87.8) 76.5 (50.1, 93.2)
NPV 71.2 (57.9, 82.2) 80.8 (67.5, 90.4)
Open in a new tab

Values are percentages.

The diagnostic performance of the Dutch criteria when combined with the absence of chemical shift artifact is summarized in Table 4. This combination resulted in a significant increase in sensitivity for both the junior and senior readers (p = 0.04 for both), indicating improved ability to detect metastatic lymph nodes. Specifically, sensitivity increased from 52.2% to 82.6% in the senior reader. However, this gain in sensitivity was accompanied by a significant reduction in specificity (p = 0.003 for the junior reader and p = 0.001 for the senior reader). The overall accuracy was 62.3% for the junior reader and 71.0% for the senior reader. While positive predictive values (PPV) remained modest (44.4% and 54.3%, respectively), the negative predictive values (NPV) were relatively high, particularly for the senior reader (88.2%).

Table 4.

Diagnostic performance of the Dutch criteria combined with the absence of chemical shift to predict lymph node involvement on postoperative histopathology.

Junior reader (95% CI) Senior reader (95% CI)
Accuracy 62.3 (50.5, 72.8) 71.0 (59.4, 80.4)
Sensitivity 52.2 (30.6, 73.2) 82.6 (61.2, 95.0)
Specificity 67.4 (52.0, 80.5) 65.2 (49.8, 78.6)
PPV 44.4 (25.5, 64.7) 54.3 (36.6, 71.2)
NPV 73.8 (58.0, 86.1) 88.2 (72.5, 96.7)
Open in a new tab

Values are percentages.

Inter-reader Agreement

Inter-reader agreement was good for short axis diameter (ICC = 0.76, [95% CI: 0.54, 0.87]), moderate for internal signal characteristic (κ = 0.42 [95% CI: 0.15, 0.65]), substantial for border contour (κ = 0.75 [95% CI: 0.36, 0.93]), and fair for shape, chemical shift artifact, and the Dutch criteria (κ = 0.34 [95% CI: 0.11, 0.55], 0.38 [95% CI: 0.12, 0.57], and 0.32 [95% CI: 0.08, 0.58], respectively).

Discussion

Our study focused on clinical ERC determined preoperatively on MRI, as distinguished from studies that have focused on locally advanced rectal cancer, as we believe that data on MRI in ERC represents a knowledge gap in the literature. Our results demonstrate that quantitative analysis of mesorectal lymph nodes on preoperative DWI, by determining the ADC of the most suspicious lymph node or by determining the ADC across all mesorectal lymph nodes, could not differentiate metastatic from benign lymph nodes. Specifically, there was no significant difference in the ADCmean of the most suspicious lymph node nor in the ADC (ADCmean, ADCmin, ADCmax) across all mesorectal lymph nodes between patients with and without nodal metastases.

Based on current literature, lymph node metastases occur in approximately 10% to 15% of T1 rectal tumors and about 18% to 24% of T2 rectal tumors [23–29]. In our cohort, we found higher occurrence of nodal metastases (21.4% and 28.6%). Several factors may account for this finding. Variations in surgical technique or pathological assessment – such as the number of lymph nodes retrieved and examined – can influence the detection rate of nodal metastases. It is possible that our center achieved a higher nodal yield or applied more meticulous pathological evaluation, thereby increasing the likelihood of detecting metastases. Additionally, the relatively small size of our cohort may contribute to statistical variability; in smaller samples, a few additional cases of nodal positivity can markedly influence the overall percentage.

DWI obtains its contrast from variations in cellular density between tissues. Lymphoid tissue, which has a relatively high cellular density, induces diffusion restriction, leading to elevated signal intensity on DWI and low signal on ADC [30]. Since the presence of malignant cells may alter the cellular structure within lymph nodes, some studies have investigated the potential of DWI to distinguish between benign and metastatic lymph nodes, mostly in head and neck [31–34] and gynecological [35; 36] cancers, with only a few studies having been done in rectal cancer. Of the studies that have been done in rectal cancer, Lambregts et al. [18] demonstrated that ADC was higher for malignant nodes in patients with locally advanced rectal cancer after chemoradiation; however, the addition of DWI to T2WI did not improve the accuracy of T2WI alone. Other attempts to establish ADC thresholds in rectal cancer exhibited variable sensitivities and specificities from 67–88% and 60–97%, respectively [18–21].

We hypothesize that the utility of ADC in our study may have been limited by the small size of both malignant and benign lymph nodes combined with the suboptimal resolution of DW images and ADC maps leading to image distortion. The median short axis diameter was 0.4 cm for the junior reader and 0.5 cm for the senior reader in both metastatic and benign nodes. This finding is corroborated by other studies in the literature. Brown et al. [17] reported that the diameter of malignant and benign was similar in their study sample. Moreover, approximately 30–50% of metastatic lymph nodes have been reported in the literature to be equal to or smaller than 5 mm in size [37–39]. Additional technical factors could also be contributory to the limited utility of ADC in our study, such as different scanner field strengths and different scanner technology between the scans [40; 41].

Regarding the secondary aim in our study, which was to assess preoperative T2WI lymph node size and morphologic criteria in predicting lymph node metastases in ERC, we found that the Dutch criteria, which relies on the combination of short axis measurement and morphologic criteria, significantly differentiated pN+ from pN− patients for the senior reader but not for the junior reader. The junior reader reported the presence of suspicious morphologic criteria less frequently than the senior reader, leading to a low sensitivity and to the underdiagnoses of metastatic lymph nodes. Additionally, we found that the sensitivities of the Dutch criteria in our study (56.5% for the senior reader and 26.1% for the junior reader) were lower than previously reported; for instance, Brown et al. [17] reported a sensitivity of 85%. This discrepancy may, at least in part, be attributed to the early-stage nature of disease in our study sample, in which metastatic lymph nodes tend to be smaller – the median short axis diameter was 0.4 cm for the junior reader and 0.5 cm for the senior reader in both metastatic and benign nodes – which makes the criteria to mainly rely on the assessment of T2W morphologic features, what can be also limited in small nodes.

Previously, a comprehensive meta-analysis, limited to studies with node-by-node comparison of MRI and pathology, affirmed that incorporating morphological features (such as irregular margin and mixed signal intensity) alongside size criteria enhances the sensitivity and specificity of MRI [16]. Nonetheless, the modest diagnostic enhancement was attributed to the subjective assessment of these findings by the radiologists. In the future, enhancing the utilization of objective features and incorporating AI technologies has the potential to alleviate the influence of reader experience on the diagnostic performance of MRI. Indeed, a radiomics study published recently showed better performance of the senior radiologist compared with the junior radiologist in the diagnosis of metastatic lymph node in patients with rectal cancer and proposed a weakly supervised deep learning model that improved the performance of the junior and senior radiologists [42].

Of note, despite the superior performance of the senior radiologist in using the Dutch criteria, the sensitivity of the Dutch criteria was only 56.5% (95% CI: 34.5%, 76.8%). This implies that if these criteria were employed to assess the feasibility of organ preservation approaches, a significant proportion of positive lymph nodes would remain undetected. Alternatively, the presence of such ERC tumors in the middle or upper rectum usually presents few obstacles to low anterior resection with sphincter preservation, and a patient can then receive adjuvant chemotherapy in the event that involved nodes had not been predicted. A potential area of further investigation would be the development of unique nodal criteria or combinations therein for ERC, to supplement the Dutch Criteria which may be more suited to locally advanced rectal cancer. On the other hand, the specificity of the Dutch criteria was high for both readers (> 90%), and therefore it could still be helpful to avoid false positive nodal diagnosis and counteract overtreatment with neoadjuvant treatment.

In addition to the morphological criteria taken into consideration in the Dutch criteria, we assessed the presence of chemical shift artifact on T2WI. The chemical shift artifact is characterized by the presence of dark and light bands along the frequency axis at fat/water interfaces. Since the presence of fatty hilum in the lymph nodes correlates with benignity, some authors have suggested that the chemical shift artifact could potentially disappear in malignant lymph nodes [43; 44]. In our study, the presence of chemical shift was associated with benign lymph node status for the senior reader and might represent an additional potential imaging feature that could add value in the imaging assessment of lymph nodes. Moreover, we found that combining the absence of chemical shift artifact with the Dutch criteria significantly improved sensitivity, although at the expense of reduced specificity for both readers. Further research is warranted to explore this issue in greater depth.

Our study had some limitations. First, this study had a small sample size and was retrospective in nature. Second, we performed patient-level analysis as we were not able to perform node-by-node evaluation correlating MRI and pathology. Third, most lymph nodes were small, which may have limited the characterization of morphologic features on T2WI and segmentation on DWI. Fourth, the scans were performed on different MRI machines, which could have impacted the ADC value. Lastly, the fair to moderate inter-reader agreement underscores the subjective nature of certain imaging features and the need for improved standardization.

In conclusion, the DWI quantitative ADC parameter was not helpful in differentiating metastatic from benign lymph nodes in ERC. Size and morphologic criteria assessed on T2WI remain the better MRI criteria to assess lymph nodes; however, their accuracy may be suboptimal if a patient wishes to pursue organ-conserving treatment. Finally, the Dutch criteria had a high specificity for both junior and senior readers, and in the hands of an experienced reader, the Dutch Criteria had a reasonable accuracy and sensitivity with the presence of involved nodes, suggesting the presence of a learning curve for radiologists. Accurate nodal staging in ERC is crucial for decision-making in an era that emphasizes organ-conserving treatment strategies and to avoid overtreatment with neoadjuvant therapy and its potential toxicity. Our study reinforces the limitations of MRI in nodal assessment and highlights the need for additional research and exploration of alternative and more objective methods for lymph node characterization. Future studies could investigate the value of more advanced MRI sequences, such as intravoxel incoherent motion (IVIM)-DWI, fat-suppressed T2-weighted or gradient echo sequences, and chemical shift imaging, as well as radiomics and texture analysis, to improve the morphological and functional assessment of lymph nodes in rectal cancer.

Supplementary Material

Supplementary Figure 1

Key Points:

Question

The performance of MRI for nodal staging in clinical early rectal cancer in particular is a knowledge gap in the literature.

Finding

The Dutch criteria based on T2WI size and morphologic parameters performed better in differentiating metastatic from benign lymph nodes than the quantitative DWI ADC parameter.

Clinical Relevance

Accurate nodal staging in early rectal cancer is crucial for treatment decision-making. Our study highlights the need for additional research of alternative and more objective methods for lymph node characterization.

Abbreviations:

ADC

apparent diffusion coefficient

ERC

early rectal cancer

ICC

intraclass correlation coefficient

ROI

region of interest

T2WI

T2-weighted imaging

TME

total mesorectal excision

References

  • 1.Sung H, Ferlay J, Siegel RL et al. (2021) Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 71:209–249 [DOI] [PubMed] [Google Scholar]
  • 2.Siegel RL, Wagle NS, Cercek A, Smith RA, Jemal A (2023) Colorectal cancer statistics, 2023. CA Cancer J Clin 73:233–254 [DOI] [PubMed] [Google Scholar]
  • 3.Atkin WS, Cook CF, Cuzick J, Edwards R, Northover JM, Wardle J (2002) Single flexible sigmoidoscopy screening to prevent colorectal cancer: baseline findings of a UK multicentre randomised trial. Lancet 359:1291–1300 [DOI] [PubMed] [Google Scholar]
  • 4.Loud JT, Murphy J (2017) Cancer Screening and Early Detection in the 21(st) Century. Semin Oncol Nurs 33:121–128 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Morino M, Risio M, Bach S et al. (2015) Early rectal cancer: the European Association for Endoscopic Surgery (EAES) clinical consensus conference. Surg Endosc 29:755–773 [DOI] [PubMed] [Google Scholar]
  • 6.Mendis S, To YH, Tie J (2022) Biomarkers in Locally Advanced Rectal Cancer: A Review. Clin Colorectal Cancer 21:36–44 [DOI] [PubMed] [Google Scholar]
  • 7.Paty PB, Nash GM, Baron P et al. (2002) Long-term results of local excision for rectal cancer. Ann Surg 236:522–529; discussion 529–530 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Chen YY, Liu ZH, Zhu K, Shi PD, Yin L (2013) Transanal endoscopic microsurgery versus laparoscopic lower anterior resection for the treatment of T1–2 rectal cancers. Hepatogastroenterology 60:727–732 [DOI] [PubMed] [Google Scholar]
  • 9.Peeters KC, Marijnen CA, Nagtegaal ID et al. (2007) The TME trial after a median follow-up of 6 years: increased local control but no survival benefit in irradiated patients with resectable rectal carcinoma. Ann Surg 246:693–701 [DOI] [PubMed] [Google Scholar]
  • 10.Juul T, Ahlberg M, Biondo S et al. (2014) Low anterior resection syndrome and quality of life: an international multicenter study. Dis Colon Rectum 57:585–591 [DOI] [PubMed] [Google Scholar]
  • 11.Goffredo P, Smith JJ (2022) The Paradox of Early Stage Rectal Cancer: More ReSeARCh in the Right Direction. Ann Surg Oncol 29:1513–1515 [DOI] [PubMed] [Google Scholar]
  • 12.Gunderson LL, Callister M, Marschke R, Young-Fadok T, Heppell J, Efron J (2008) Stratification of rectal cancer stage for selection of postoperative chemoradiotherapy: current status. Gastrointest Cancer Res 2:25–33 [Google Scholar]
  • 13.Glynne-Jones R, Wyrwicz L, Tiret E et al. (2017) Rectal cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol 28:iv22–iv40 [DOI] [PubMed] [Google Scholar]
  • 14.Al-Sukhni E, Milot L, Fruitman M et al. (2012) Diagnostic accuracy of MRI for assessment of T category, lymph node metastases, and circumferential resection margin involvement in patients with rectal cancer: a systematic review and meta-analysis. Ann Surg Oncol 19:2212–2223 [DOI] [PubMed] [Google Scholar]
  • 15.Park JS, Jang YJ, Choi GS et al. (2014) Accuracy of preoperative MRI in predicting pathology stage in rectal cancers: node-for-node matched histopathology validation of MRI features. Dis Colon Rectum 57:32–38 [DOI] [PubMed] [Google Scholar]
  • 16.Zhuang Z, Zhang Y, Wei M, Yang X, Wang Z (2021) Magnetic Resonance Imaging Evaluation of the Accuracy of Various Lymph Node Staging Criteria in Rectal Cancer: A Systematic Review and Meta-Analysis. Front Oncol 11:709070. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Brown G, Richards CJ, Bourne MW et al. (2003) Morphologic predictors of lymph node status in rectal cancer with use of high-spatial-resolution MR imaging with histopathologic comparison. Radiology 227:371–377 [DOI] [PubMed] [Google Scholar]
  • 18.Lambregts DM, Maas M, Riedl RG et al. (2011) Value of ADC measurements for nodal staging after chemoradiation in locally advanced rectal cancer-a per lesion validation study. Eur Radiol 21:265–273 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Kim SH, Ryu KH, Yoon JH et al. (2015) Apparent diffusion coefficient for lymph node characterization after chemoradiation therapy for locally advanced rectal cancer. Acta Radiol 56:1446–1453 [DOI] [PubMed] [Google Scholar]
  • 20.Cho EY, Kim SH, Yoon JH et al. (2013) Apparent diffusion coefficient for discriminating metastatic from non-metastatic lymph nodes in primary rectal cancer. Eur J Radiol 82:e662–668 [DOI] [PubMed] [Google Scholar]
  • 21.Bipat S, Glas AS, Slors FJ, Zwinderman AH, Bossuyt PM, Stoker J (2004) Rectal cancer: local staging and assessment of lymph node involvement with endoluminal US, CT, and MR imaging--a meta-analysis. Radiology 232:773–783 [DOI] [PubMed] [Google Scholar]
  • 22.Kusters M, Marijnen CA, van de Velde CJ et al. (2010) Patterns of local recurrence in rectal cancer; a study of the Dutch TME trial. Eur J Surg Oncol 36:470–476 [DOI] [PubMed] [Google Scholar]
  • 23.Kim EM, Son IT, Kim BC, Park JH, Kang BM, Kim JW (2023) A Retrospective Multicenter Study of Risk Factors, Stratification, and Prognosis of Lymph Node Metastasis in T1 and T2 Colorectal Cancer. J Clin Med 12 [Google Scholar]
  • 24.Fields AC, Lu P, Hu F et al. (2021) Lymph Node Positivity in T1/T2 Rectal Cancer: a Word of Caution in an Era of Increased Incidence and Changing Biology for Rectal Cancer. J Gastrointest Surg 25:1029–1035 [DOI] [PubMed] [Google Scholar]
  • 25.Okabe S, Shia J, Nash G et al. (2004) Lymph node metastasis in T1 adenocarcinoma of the colon and rectum. J Gastrointest Surg 8:1032–1039; discussion 1039–1040 [DOI] [PubMed] [Google Scholar]
  • 26.Chang HC, Huang SC, Chen JS et al. (2012) Risk factors for lymph node metastasis in pT1 and pT2 rectal cancer: a single-institute experience in 943 patients and literature review. Ann Surg Oncol 19:2477–2484 [DOI] [PubMed] [Google Scholar]
  • 27.Zong Z, Li H, Hu CG et al. (2021) Predictors of lymph-node metastasis in surgically resected T1 colorectal cancer in Western populations. Gastroenterol Rep (Oxf) 9:470–474 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Deng J, Zhou S, Wang Z, Huang G, Zeng J, Li X (2022) Comparison of Prognosis and Lymph Node Metastasis in T1-Stage Colonic and Rectal Carcinoma: A Retrospective Study. Int J Gen Med 15:3651–3662 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Morita Y, Kudo SE, Takashina Y et al. (2025) Lymph node metastasis risk factors in T2 colorectal cancer. DEN Open 5:e70040. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Kwee TC, Takahara T, Ochiai R, Nievelstein RA, Luijten PR (2008) Diffusion-weighted whole-body imaging with background body signal suppression (DWIBS): features and potential applications in oncology. Eur Radiol 18:1937–1952 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.de Bondt RB, Hoeberigs MC, Nelemans PJ et al. (2009) Diagnostic accuracy and additional value of diffusion-weighted imaging for discrimination of malignant cervical lymph nodes in head and neck squamous cell carcinoma. Neuroradiology 51:183–192 [DOI] [PubMed] [Google Scholar]
  • 32.Vandecaveye V, De Keyzer F, Vander Poorten V et al. (2009) Head and neck squamous cell carcinoma: value of diffusion-weighted MR imaging for nodal staging. Radiology 251:134–146 [DOI] [PubMed] [Google Scholar]
  • 33.Sumi M, Sakihama N, Sumi T et al. (2003) Discrimination of metastatic cervical lymph nodes with diffusion-weighted MR imaging in patients with head and neck cancer. AJNR Am J Neuroradiol 24:1627–1634 [PMC free article] [PubMed] [Google Scholar]
  • 34.King AD, Ahuja AT, Yeung DK et al. (2007) Malignant cervical lymphadenopathy: diagnostic accuracy of diffusion-weighted MR imaging. Radiology 245:806–813 [DOI] [PubMed] [Google Scholar]
  • 35.Kim JK, Kim KA, Park BW, Kim N, Cho KS (2008) Feasibility of diffusion-weighted imaging in the differentiation of metastatic from nonmetastatic lymph nodes: early experience. J Magn Reson Imaging 28:714–719 [DOI] [PubMed] [Google Scholar]
  • 36.Holzapfel K, Duetsch S, Fauser C, Eiber M, Rummeny EJ, Gaa J (2009) Value of diffusion-weighted MR imaging in the differentiation between benign and malignant cervical lymph nodes. Eur J Radiol 72:381–387 [DOI] [PubMed] [Google Scholar]
  • 37.Kotanagi H, Fukuoka T, Shibata Y et al. (1993) The size of regional lymph nodes does not correlate with the presence or absence of metastasis in lymph nodes in rectal cancer. J Surg Oncol 54:252–254 [DOI] [PubMed] [Google Scholar]
  • 38.Yao YF, Wang L, Liu YQ, Li JY, Gu J (2011) Lymph node distribution and pattern of metastases in the mesorectum following total mesorectal excision using the modified fat clearing technique. J Clin Pathol 64:1073–1077 [DOI] [PubMed] [Google Scholar]
  • 39.Langman G, Patel A, Bowley DM (2015) Size and distribution of lymph nodes in rectal cancer resection specimens. Dis Colon Rectum 58:406–414 [DOI] [PubMed] [Google Scholar]
  • 40.Kwee TC, Takahara T, Luijten PR, Nievelstein RA (2010) ADC measurements of lymph nodes: inter- and intra-observer reproducibility study and an overview of the literature. Eur J Radiol 75:215–220 [DOI] [PubMed] [Google Scholar]
  • 41.Surov A, Meyer HJ, Pech M, Powerski M, Omari J, Wienke A (2021) Apparent diffusion coefficient cannot discriminate metastatic and non-metastatic lymph nodes in rectal cancer: a meta-analysis. Int J Colorectal Dis 36:2189–2197 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Xia W, Li D, He W et al. (2024) Multicenter Evaluation of a Weakly Supervised Deep Learning Model for Lymph Node Diagnosis in Rectal Cancer at MRI. Radiol Artif Intell 6:e230152. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Farshchian N, Tamari S, Farshchian N, Madani H, Rezaie M, Mohammadi-Motlagh HR (2011) Diagnostic value of chemical shift artifact in distinguishing benign lymphadenopathy. Eur J Radiol 80:594–597 [DOI] [PubMed] [Google Scholar]
  • 44.Zhang H, Zhang C, Zheng Z et al. (2017) Chemical shift effect predicting lymph node status in rectal cancer using high-resolution MR imaging with node-for-node matched histopathological validation. Eur Radiol 27:3845–3855 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Figure 1
NIHMS2117815-supplement-Supplementary_Figure_1.pptx (93.1KB, pptx)

RESOURCES