T2* mapping in magnetic resonance imaging for stratifying the hypoxic microenvironment in pancreatic ductal adenocarcinoma: a preliminary study

Abstract

Background

T2* mapping provides a noninvasive approach for quantifying tissue oxygenation via the detection of the paramagnetic effects associated with deoxyhemoglobin, with validated correlations to hypoxia in other cancers. However, its application in stratifying hypoxic microenvironments in pancreatic ductal adenocarcinoma (PDAC) remains limited. This preliminary study aimed to evaluate T2* mapping as a quantitative imaging biomarker for hypoxia levels in PDAC in order to offer a potential tool for the noninvasive characterization of the tumor microenvironment and the development of personalized treatment strategies.

Methods

This retrospective study enrolled 50 patients with PDAC pathologically confirmed between June 2022 and July 2023. Hypoxia-inducible factor-1α (HIF-1α) expression was used to evaluate PDAC hypoxia levels, and patients were stratified into low and high hypoxia groups. T2* values and other clinicoradiological indicators were compared between the two groups via binary logistic regression analysis. A logistic regression model was built with independent factors related to hypoxia levels. The predictive performance of the model was evaluated with area under the curve (AUC) and leave-one-out cross-validation (LOOCV). The correlation between the PDAC HIF-1α expression and T2* values was assessed via the Spearman rank correlation coefficient, and hypoxia levels were compared with pathological findings.

Results

T2* values differed significantly between the low and high hypoxia groups (62.63±3.33 vs. 59.16±3.97 ms; P=0.002), as did rim enhancement (P<0.001). Multivariate analysis identified the independent predictors of hypoxia to be T2* values [odds ratio (OR) =0.819; 95% confidence interval (CI): 0.674–0.994; P=0.044] and rim enhancement (OR =6.261; 95% CI: 1.532–25.581; P=0.011). The logistic regression model achieved an initial AUC of 0.822, retaining diagnostic performance after LOOCV (AUC =0.785). T2* values exhibited a significant inverse correlation with HIF-1α expression (r=−0.463; P<0.001). High-hypoxia tumors were associated with poorer differentiation (P<0.001).

Conclusions

T2* mapping may serve as a noninvasive imaging biomarker for stratifying the hypoxic microenvironment in PDAC.

Introduction

Pancreatic cancer is a highly aggressive malignancy within the digestive system, characterized by significantly high mortality rates in relation to its incidence (1). In China, pancreatic cancer ranks as the sixth leading cause of cancer-related mortality (2). Pancreatic ductal adenocarcinoma (PDAC) constitutes over 90% of pancreatic malignancies, with current clinical outcomes remaining poor, as evidenced by a 5-year survival rate below 12% (3). Although the survival outcomes in recent years have improved for certain patients due to a combination of surgical resection and chemotherapy, the overall therapeutic efficacy is limited by the tumor microenvironments’ high heterogeneity and immunosuppressive properties (4,5). Within this context, hypoxia has emerged as a defining hallmark of PDAC progression (6), driving metastasis, chemoresistance, and poor prognosis (7,8).

The hypoxic microenvironment in PDAC arises from abundant stromal deposition, extracellular matrix expansion, and low vascular density, collectively impairing intratumoral perfusion (9). This hypoxia fuels critical pathophysiological processes, including metabolic reprogramming, treatment resistance, and disease aggressiveness (10). Consequently, conventional therapies (e.g., gemcitabine-albumin-paclitaxel) exhibit restrained efficacy (11), contributing to the persistently poor survival of patients with PDAC (12). Stratifying hypoxia levels is thus essential for developing targeted therapies and personalized treatment strategies (13).

Current hypoxia assessment remains clinically challenging. Although hypoxia-inducible factor-1α (HIF-1α) serves as a key molecular mediator, orchestrating adaptive responses under hypoxia and demonstrating constitutive upregulation in PDAC (14), its measurement requires an invasive biopsy. HIF-1α further promotes treatment resistance via transcriptional regulation of DNA repair and apoptotic pathways (15) and thus exerts a dual function as a biomarker and therapeutic target (16,17).

T2* mapping magnetic resonance imaging (MRI) offers a noninvasive alternative for hypoxia quantification. This technique measures paramagnetic susceptibility shifts from tissue deoxyhemoglobin, providing reproducible, whole-tumor oxygenation data that indirectly reflect partial oxygen pressure (pO₂) dynamics (18). Clinically validated in breast cancer, glioblastoma, and sarcomas, T2*-derived parameters correlate strongly with histopathological hypoxia and can predict therapeutic outcomes (12,19,20). However, its utility in PDAC remains unclear.

Therefore, this study aimed to preliminarily investigate whether T2* mapping can serve as a quantitative imaging biomarker for the degree of hypoxia in PDAC in order to develop a tool for the noninvasive assessment of the hypoxic microenvironments and to inform the development of personalized treatment strategies. We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-2020/rc).

Methods

Patient selection

This retrospective, observational, clinical study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments and was approved by the Medical Ethics Committee of Changzhou First People’s Hospital (approval No. 2023012). The requirement for informed consent was waived by the Medical Ethics Committee due to the retrospective nature of the analysis. Between June 2022 and July 2023, 133 patients who underwent T2* mapping scanning were included in this study. The inclusion criteria were as follows: (I) MRI findings highly suggestive of PDAC; (II) completion of pancreatic MRI examination with imaging data within 1 month before surgery; and (III) availability of complete clinical history, laboratory findings, and imaging documentation. Meanwhile, the exclusion criteria were as follows: (I) the confirmation of non-PDAC on the basis of postoperative biopsy; (II) poor quality MRI images caused by severe motion artifacts or metallic stent artifacts; (III) inadequate tissue fixation or unsatisfactory staining of pathological specimens precluding histological grading; (IV) other pancreatic space-occupying lesions; and (V) administration of neoadjuvant chemotherapy and radiation therapy. Finally, 50 patients with PDAC were enrolled in this study (Figure 1).

Figure 1.

Detail of patient exclusion. MRI, magnetic resonance imaging; PDAC, pancreatic ductal adenocarcinoma.

MRI procedure

All MRI examinations were conducted with a 3.0-T MRI scanner (MAGNETOM Vida, Siemens Healthineers AG, Erlangen, Germany) equipped with an 18-channel phased-array body coil. The positioning center was aligned with the coil center and the xiphoid process. The scan range extended from the diaphragmatic dome to the inferior margin of the pancreas. The MRI sequences encompassed axial T1-weighted imaging (T1WI), axial T2-weighted imaging (T2WI), and axial dynamic contrast-enhanced MRI (DCE-MRI). The T2* mapping sequence was performed with a two-dimensional multiecho spoiled gradient-echo (fast low-angle shot-based) sequence with multiple echoes. All echoes were acquired within a single breath-hold of 12 seconds, eliminating the need for motion correction. In total, each patient performed three breath-holds, and the overall scan time was 58 s. Subsequently, T2* values were derived by fitting a mono-exponential decay model with a constant offset to the multiecho signal intensity data. The signal S(t) at echo time, t, can be described by the following:

$$\text{S}\left(\text{t}\right)=\text{A}+{\text{S}}_0^{\text{*}}\cdot \text{exp}\left(-\text{t}/\text{T}2\text{*}\right)$$ [1]

where A is a constant baseline offset, S₀ is the initial signal intensity, and T2* is the effective transverse relaxation time (21). The detailed acquisition parameters of all MR sequences are shown in Table 1.

Table 1. Imaging protocol parameters.

Parameters	T1WI	T2WI	T2* mapping	DCE-MRI
TR (ms)	4.0	1,400	203	3.8
TE1/ΔTE (ms)	1.29/–	94/–	2.46/2.46	1.25/–
Thickness (mm)	3.0	5.0	3.0	3.0
FOV (mm)	380×308	360×292	380×285	360×270
Matrix	256×320	311×384	205×256	216×288
FA (°)	9	160	60	12
Bandwidth (Hz/Px)	1,040	685	470	600

DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; FA, flip angle; FOV, field of view; MRI, magnetic resonance imaging; T1WI, T1-weighted imaging; T2WI, T2-weighted imaging; TE, echo time; TR, repetition time.

Image analysis

The initial imaging data were transferred to the syngo.via postprocessing workstation (Siemens Healthineers). Two abdominal radiologists with 5 and 7 years of experience in pancreatic MRI independently identified a region of interest (ROI) on T2* mapping sequences with guidance from axial T2WI and T1-weighted DCE-MRI. They were blinded to the other patient information and to each other’s analyses. To standardize the lesion analysis as much as possible, the ROI was marked on the slice where the cancer exhibited the largest diameter, with areas of necrosis, cystic degeneration, or significant artifacts being avoided and as much tumor tissue as possible being included (Figures 2,3). Additionally, the radiologists evaluated several imaging features for each PDAC sample, including tumor signal intensity on each unenhanced T1WI and T2WI, tumor size (the maximum cross-sectional diameter), tumor location, distal pancreatic atrophy (≤7 mm at the pancreatic body) (22), pancreatic main duct dilatation (main pancreatic duct dilatation ≥3 mm) (23), and rim enhancement (irregular ringlike enhancement on contrast-enhanced phases) (24). In cases of discrepancy between the results of the two reviewers, a senior radiologist (with more than 15 years of experience) reviewed the cases to reach a consensus through discussion. To assess the interobserver reproducibility of imaging features and T2* values, both radiologists independently reviewed the imaging studies and annotated ROIs. Intraobserver reproducibility was evaluated by having one radiologist repeat the measurements after a 1-month interval.

Figure 2.

Images showing poorly differentiated PDAC in the tail of the pancreas in a 78-year-old woman. MR images demonstrated that the tumor diameter was 3.6 cm on axial (A) T1-weighted DCE-MRI and (B) T2* mapping. (C) Strong nuclear staining for HIF-1α in a proportion of tumor cells (IHC staining, ×200). The red circle indicates the PDAC tumor ROI. DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; HIF-1α, hypoxia-inducible factor-1α; IHC, immunohistochemistry; MR, magnetic resonance; PDAC, pancreatic ductal adenocarcinoma.

Figure 3.

Images showing highly differentiated PDAC in the head of the pancreas in a 59-year-old woman. MR images demonstrated that the tumor diameter was 2.2 cm on axial (A) T1-weighted DCE-MRI and (B) T2* mapping. (C) Nuclear staining for HIF-1α in a proportion of tumor cells (IHC staining, ×200). The red circle indicates the PDAC tumor ROI. DCE-MRI, dynamic contrast-enhanced magnetic resonance imaging; HIF-1α, hypoxia-inducible factor-1α; IHC, immunohistochemistry; MR, magnetic resonance; PDAC, pancreatic ductal adenocarcinoma.

Pathological analysis

The pathological specimens were processed via the streptavidin-peroxidase (SP) immunohistochemical staining method. Tissue sections were fixed in 10% formalin and embedded in paraffin. Subsequently, 5-µm sections were cut from formaldehyde-fixed tissue, dewaxed, and counterstained. The primary antibody—mouse anti-human HIF-1α monoclonal antibody and horseradish peroxidase-labeled secondary antibody—were applied and incubated. Antigen-antibody complexes were visualized with a freshly prepared DAB chromogenic solution. Nuclear counterstaining was performed with hematoxylin. Further details on the staining procedures are provided in Appendix 1.

A pathologist, blinded to clinical parameters and tumor grading, evaluated the expression of HIF-1α. HIF-1α expression was classified based on an established semiquantitative scoring system (25). Five representative high-power fields (HPFs; ×200 magnification and a diameter of 5.5 mm) from sections enriched with cancer cells were examined. Positive immunoreactivity was characterized by cytoplasmic or nuclear yellowish-brown deposits in tumor cells, primarily in the cytoplasm with occasional nuclear staining (Figures 2,3). The staining intensity score was classified as follows: 0, none; 1, weak; 2, moderate; and 3, strong. In addition, the positive cell percentage of HIF-1α was categorized as follows: 0, <5%; 1, 5% to 25%; 2, 26% to 50%; 3, 51% to 75%; and 4, >75%. HIF-1α expression was the product of the integrated positive cell percentage and staining intensity. Cases were stratified according to the study’s mean index value of 4.6 as the threshold, as follows: low hypoxia group, HIF-1α expression ≤4 (scores 0, 1, 2, 3, and 4); and high hypoxia group, HIF-1α expression ≥6 (scores 6, 8, 9, and 12).

Statistical analysis

Statistical analyses were conducted with R version 3.6.3 (The R Foundation for Statistical Computing, Vienna, Austria) and SPSS version 27.0 software (IBM Corp., Armonk, NY, USA). Statistical significance difference was set at P<0.05 for two-tailed tests. The intraclass correlation coefficient (ICC) was used to evaluate intra- and interobserver agreement in the measurements of T2* values (26). Interobserver agreement for the presence of imaging features was assessed via the Cohen’s kappa coefficient. The normality of continuous variables was determined via the Shapiro-Wilk test. Patient demographic information, laboratory data, and pathological features are presented as counts and percentages. Meanwhile, continuous variables are presented as the mean ± standard deviation and were compared via the Student t-test, one-way analysis of variance, and the Fisher exact test. Multiple comparisons were corrected with the Bonferroni method. Nonnormally distributed data are expressed as the median and interquartile range, while categorical variables are presented as frequency counts (n) and were analyzed with the Chi-squared test. We performed univariate logistic regression analysis to evaluate the relationship between clinicoradiological features, MRI parameters, and hypoxia levels. Multivariate logistic regression analysis included all variables with a P value less than 0.10 in the univariate analysis. Receiver operating characteristic (ROC) curve analysis was performed to assess the ability to diagnose hypoxia levels in PDAC. Additionally, the area under the curve (AUC) was calculated to assess the diagnostic performance of the logistic regression model, which consisted of independent associated factors. Due to the limited sample size, the data were not divided into separate training and test sets. Instead, leave-one-out cross-validation (LOOCV) was used to assess model robustness. In each iteration, one patient was held out as the test sample, while the multivariate logistic regression model was refitted on the remaining patients. Predicted probabilities from all iterations were then aggregated to compute cross-validated performance metrics, including the AUC. To evaluate the correlation between T2* values and HIF-1α expression, the Spearman correlation analysis was employed (27). The magnitude of Spearman rank correlation coefficient (ρ) was interpreted as follows: very weak, 0.00–0.19; weak, 0.20–0.39; moderate, 0.40–0.59; strong, 0.60–0.79; and very strong, 0.80–1.00.

Results

Clinicopathological features of patients

The study ultimately enrolled 50 patients (29 males and 21 females) with an age range between 45 and 81 years (mean 68.7±8.6 years). The HIF-1α expression score had the following distribution: 0, 1 case; 1, 4 cases; 2, 11 cases; 3, 9 cases; 4, 5 cases; 6, 10 cases; 8, 4 cases; 9, 2 cases; and 12, 4 cases. Based on the predefined criteria, cases were categorized into low (n=30, 60%) and high hypoxia groups (n=20, 40%). Differences in clinicoradiological features between the patients with resected PDAC in the low and high hypoxia levels are presented in Table 2.

Table 2. Demographic and clinicoradiological characteristics.

Characteristics	Number	Low hypoxia (n=30)	High hypoxia (n=20)	P
Demographics
Age (years)	50	67.50±7.29	70.55±10.29	0.225
Sex	50			0.349
Male		19 (63.3)	10 (50.0)
Female		11 (36.7)	10 (50.0)
BMI (kg/m2)	50	22.86±2.36	23.91±3.36	0.205
Radiological findings
Location	50			0.485
Head	28	18 (60.0)	10 (50.0)
Body/tail	22	12 (40.0)	10 (50.0)
Tumor size (mm)	50	34.03±9.80	33.90±7.69	0.959
Distal pancreatic atrophy				0.355
Present	26	14 (46.7)	12 (60.0)
Absent	24	16 (53.3)	8 (40.0)
Pancreatic main duct dilatation				0.722
Present	44	26 (86.7)	18 (90.0)
Absent	6	4 (13.3)	2 (10.0)
Rim enhancement				<0.001
Present	18	5 (16.7)	13 (65.0)
Absent	32	25 (83.3)	7 (35.0)
Unenhanced T1WI				0.136
Isointensity	4	1 (3.3)	3 (15.0)
Hypointensity	46	29 (96.7)	17 (85.0)
Unenhanced T2WI				0.780
Isointensity	11	7 (23.3)	4 (20.0)
Hyperintensity	39	23 (76.7)	16 (80.0)
T2* values (ms)	50	62.63±3.33	59.16±3.97	0.002
Serum marker
Albumin (g/dL)				0.670
≥3.5	46	28 (93.3)	18 (90.0)
<3.5	4	2 (6.7)	2 (10.0)
CA19-9 (U/mL)				0.131
≥200	19	15 (51.7)	6 (30.0)
<200	24	14 (48.3)	14 (70.0)
CA125 (U/mL)				0.658
≥35	22	4 (14.3)	18 (90.0)
<35	26	24 (85.7)	2 (10.0)
CEA (U/mL)				0.141
≥5	14	6 (20.7)	8 (40.0)
<5	35	23 (79.3)	12 (60.0)

Data are presented as mean ± standard deviation or number (%), unless otherwise stated. BMI, body mass index; CA125, carbohydrate antigen 125; CA19-9, carbohydrate antigen 19-9; CEA, carcinoembryonic antigen; T1WI, T1-weighted imaging; T2WI, T2-weighted imaging.

Intra- and interobserver variability of T2* mapping and imaging features in PDAC

The agreement between intra- and interobserver reproducibility was excellent for the assessment of T2* mapping in PDAC [intraobserver ICC =0.915, 95% confidence interval (CI): 0.856–0.951; interobserver ICC =0.870, 95% CI: 0.782–0.924]. The interobserver agreement for imaging features was good or excellent (κ=0.72–0.84).

Association between clinicoradiological features, T2* values, and the PDAC hypoxia groups

Since the aim of this study was to preoperatively predict PDAC hypoxia levels, clinical and imaging features were incorporated into both univariate and multivariate logistic regression analyses, with pathological features being excluded. In the univariate logistic regression analysis, rim enhancement [odds ratio (OR) =9.286; P=0.001] and T2* values (OR =0.772; P=0.005) were included. Subsequently, multivariate logistic regression analysis indicated the independent predictive features of PDAC hypoxia levels to be rim enhancement (OR =6.261; P=0.011) and T2* values (OR =0.819; P=0.044) (Table 3). A logistic regression model combining T2* values and rim enhancement was constructed to predict PDAC hypoxia levels, which achieved an initial AUC of 0.822 (95% CI: 0.699–0.944), with a sensitivity of 0.750 and a specificity of 0.867 (Figure 4). After LOOCV, the logistic regression model yielded a more conservative cross-validated AUC of 0.785, with a cross-validated accuracy of 0.720, a sensitivity of 0.950, and a specificity of 0.567 (Figure 5). The slight decrease from the initial AUC (0.822) to the cross-validated AUC (0.785) can be expected in small datasets and indicates mild overfitting, but the model retained acceptable discriminative ability.

Table 3. Univariable and multivariate logistic regression of variables for the low and high hypoxia groups.

Characteristics	Univariate analysis				Multivariate analysis†
	OR	95% CI	P		OR	95% CI	P
Demographics
Age (years)	0.957	0.892–1.027	0.223
Sex
Male	Reference
Female	1.727	0.548–5.448	0.351
BMI (kg/m2)	0.872	0.705–1.078	0.204
Radiological findings
Location
Head	Reference
Body/tail	1.500	0.479–4.695	0.486
Tumor size (mm)	1.002	0.940–1.068	0.958
Distal pancreatic atrophy
Absent	Reference
Present	1.714	0.545–5.396	0.357
Pancreatic duct dilatation
Absent	Reference
Present	1.358	0.229–8.382	0.723
Rim enhancement
Absent	Reference				Reference
Present	9.286	2.458–35.075	0.001		6.261	1.532–25.581	0.011
T2* values (ms)	0.772	0.644–0.925	0.005		0.819	0.674–0.994	0.044
Serum marker
Albumin (g/dL)	1.556	0.201–12.053	0.672
CA19-9 (U/mL)
<200	Reference
≥200	2.500	0.751–8.318	0.135
CA125 (U/mL)
<35	Reference
≥35	1.500	0.247–9.111	0.660
CEA (U/mL)
<5	Reference
≥5	0.391	0.110–1.390	0.147

^†, variables with P<0.10 in the univariate analysis were included in the multivariate analysis. BMI, body mass index; CA125, carbohydrate antigen 125; CA19-9, carbohydrate antigen 19-9; CEA, carcinoembryonic antigen; CI, confidence interval; OR, odds ratio.

Figure 4.

The initial ROC curve of the logistic regression model (T2* values and rim enhancement) for distinguishing PDAC hypoxia levels. The prediction model combining T2* values and rim enhancement: logit(P) = 11.068 − 0.200 × X1 + 1.834 × X2 (X1, T2* values; X2, rim enhancement). AUC, area under the curve; CI, confidence interval; PDAC, pancreatic ductal adenocarcinoma; ROC, receiver operating characteristic.

Figure 5.

Internal validation of the logistic regression model via LOOCV. (A) The ROC curve was generated from the internal validation. The cross-validated AUC was slightly lower than the initial AUC of the logistic regression model but still demonstrated acceptable discriminatory performance. (B) The confusion matrix of the logistic regression model after LOOCV. AUC, area under the curve; CI, confidence interval; LOOCV, leave-one-out cross-validation; ROC, receiver operating characteristic.

Correlation between HIF-1α expression and T2* values

A moderate correlation was observed between T2* values and HIF-1α expression (r=−0.463; P<0.001) (Figure 6), indicating that T2* values decreased as HIF-1α expression increased.

Figure 6.

Correlation analysis between HIF-1α expression and T2* values. HIF-1α, hypoxia-inducible factor-1α.

Association between hypoxia levels and pathological characteristics

After multiple comparison adjustment was conducted via the Bonferroni method, statistical analysis indicated significant differences in hypoxia levels based on PDAC differentiation degree, with a progressive increase in expression as the differentiation grade decreased. However, there was no statistically significant association between hypoxia levels and perineural invasion (P=0.641), lymph node metastasis (P=0.201), peripancreatic invasion (P=0.265), or lymphovascular invasion (P=0.355) (Table 4).

Table 4. Relationship between pathological characteristics and the low and high hypoxia groups.

Characteristics	Number	Low hypoxia (n=30)	High hypoxia (n=20)	P
Lymph node metastasis				0.201
Present	22	11 (36.7)	11 (55.0)
Absent	28	19 (63.3)	9 (45.0)
Perineural invasion				0.641
Present	46	27 (90.0)	19 (95.00)
Absent	4	3 (10.0)	1 (5.00)
Peripancreatic invasion				0.265
Present	47	27 (90.0)	20 (100.0)
Absent	3	3 (10.0)	0 (0.00)
Lymphovascular invasion				0.355
Present	24	16 (53.3)	8 (40.0)
Absent	26	14 (46.7)	12 (60.0)
Differentiation degree				<0.001
Well	2	2 (6.7)	0 (0.0)
Moderate	25	21 (70.0)	4 (20.0)
Poor	23	7 (23.3)	16 (80.0)

Data are presented as number (%), unless otherwise stated.

Discussion

Our study investigated the value of T2* mapping in assessing the levels of hypoxia in PDAC and its correlation with pathological characteristics. The results showed that in PDAC, increased hypoxia levels were associated with lower T2* values and a higher incidence of rim enhancement. Rim enhancement and T2* values were independent predictors in multivariate analysis for differentiating hypoxia levels in PDAC. The logistic regression model of these factors achieved robust discrimination performance in predicting HIF-1α expression status (AUC =0.822), retaining acceptable validity after internal validation via LOOCV (AUC =0.785). Importantly, T2* values were found to have an inverse correlation with HIF-1α expression in PDAC. Furthermore, hypoxia levels varied across pathological features characterized by a differentiated degree.

In our study, both T2* values and rim enhancement emerged as independent predictors of hypoxia in PDAC. T2* mapping is primarily influenced by the concentration of paramagnetic substances, and deoxyhemoglobin concentration is one of the most significant factors. This mechanism has been supported by Xu et al. (28), who demonstrated a significant positive correlation between T2* values and hemoglobin oxygen saturation. Accordingly, T2* values can serve as a sensitive indicator of local oxygen content within tumors. In our study, T2* values of patients with PDAC in the high hypoxia group were significantly lower than those in the low hypoxia group (59.16±3.97 vs. 62.63±3.33 ms). This finding may be attributed to a greater severity of local hypoxia, reduced blood oxygen content, and the accumulation of paramagnetic substances, such as deoxyhemoglobin and ferric hemoflavin, within tumors in the high hypoxia group, which collectively accelerate spin dephasing and consequently reduce T2* values (18). Notably, the T2* values in our PDAC cohort were higher than those reported by Klaassen et al. (approximately 37 ms for PDAC) (29). This discrepancy is primarily attributable to differences in MRI acquisition parameters across studies. Another important distinction is that we employed a fat-suppressed sequence, which reduces interference from adipose tissue and provides a more accurate quantification of T2* values specific to the tumor parenchyma. Additionally, potential variations in patient characteristics regarding tumor stage, prior therapy, or comorbidities may also influence tissue susceptibility and the resultant measurements. This discussion further highlights the importance of standardizing acquisition protocols and postprocessing workflows in future multicenter trials to enhance the comparability and reproducibility of quantitative MRI biomarkers.

HIF-1α expression serves as the reference biomarker for tumor hypoxia, and moderate correlations between T2* values and HIF-1α expression have been established in previous clinical studies. Zhang et al. reported that R2* (1/T2*) was correlated with HIF-1α in cervical carcinoma (r=0.491) (30), Wang et al. observed similar associations in gliomas (r=0.43) (31), and Klaassen et al. (32) confirmed this relationship in PDAC (r=0.56). Consistent with these findings, our results showed a moderate inverse correlation between T2* values and HIF-1α expression (r=−0.463; P<0.001). In contrast, a stronger positive correlation between R2* and HIF-1α (r=0.791) was reported in preclinical mouse xenograft models of human hepatocellular carcinoma (33). There are two reasons for this discrepancy: First, we only analyzed the HIF-1α expression in a single pathological section, without analyzing the entire tumor. Thus, it was impossible to make a one-to-one correspondence with the T2* values. Second, apart from hypoxia, there were other factors, including iron deposition, fibrosis, and inflammatory infiltration, that could modulate T2* values. Fibrosis, as a hallmark of PDAC, can restrict water proton mobility and shorten T2* relaxation times independently of oxygenation status. Similarly, intratumoral iron deposition, often resulting from hemorrhage or macrophage activity, increases local magnetic susceptibility, thereby accelerating T2* decay. Inflammatory infiltration may also contribute through associated edema, cellular density, or paramagnetic immune cell components. These nonhypoxic influences likely attenuated the strength of the correlation between T2* and HIF-1α expression observed in our study.

Similarly, rim enhancement can predict hypoxia via its established association with aggressive tumor biology. This phenomenon can be explained by a hypoxic and necrotic tumor core surrounded by a hypervascular peripheral rim with leaky neovessels and dense fibrotic stroma, which creates a barrier that traps contrast agents at the margins while preventing perfusion centrally. A previous study (34) has indicated that in invasive cancers with rim enhancement, the positive ratios of lymph node metastasis and blood vessel invasion, along with the negative ratio of hormone receptors, are higher than are those in invasive cancers without rim enhancement. Lee et al. found that rim-enhanced PDAC on dynamic contrast-enhanced MRI exhibited a higher incidence of poor differentiation and that patients with increased intratumoral necrosis had a significantly worse disease-free survival and overall survival compared to those with non-rim-enhancing PDACs (24). Yamaguchi et al. reported rim enhancement on contrast-enhanced computed tomography to be significantly associated with poorer survival outcomes in patients with PDAC (35). Our findings are in line with these results. In future work, we will further investigate the relationship between rim enhancement and patient survival outcomes.

Notably, in our study, the logistic regression model combining T2* values and rim enhancement achieved good diagnostic efficacy (AUC =0.822) on the full dataset of 50 patients. Due to the modest sample size, LOOCV was performed to provide a less biased estimate of generalizability. Importantly, even after this internal validation, the cross-validated AUC of 0.785 remained higher than the AUC of 0.700 reported by Wang et al. (36), who used tumor size to predict HIF-1α expression in PDAC. In our preliminary study, we incorporated quantitative imaging parameters and conventional radiological enhancement features into the model, which demonstrated potential utility as a noninvasive tool for the preoperative stratification of hypoxia levels in patients with PDAC. Our findings indicate that the model may serve as a practical decision-support aid for radiologists and oncologists in refining therapeutic strategies.

Furthermore, significant variations in hypoxia levels were observed across PDAC differentiation grades. Poorly differentiated tumors had a higher HIF-1α expression and lower T2* values compared to well-/moderately differentiated tumors. Theoretically, poorly differentiated tumors manifest as pronounced desmoplastic reactions with extensive fibrosis (37), and the fibrotic stroma is rich in collagen and extracellular matrix components, with elevated interstitial pressure and reduced vascularity-induced hypoxia (38). However, our study did not identify statistically significant differences between hypoxia levels or T2* values. The absence of significant associations may be attributed to the small sample size. In future work, we intend to expand the cohort sizes and implement spatial mapping techniques to account for these discrepancies.

This study involved certain limitations that should be acknowledged. First, we employed a single-center design with a relatively small sample size. Further efforts are needed to expand the number of participating centers and increase the sample size to validate these results. Second, the pathological specimens were obtained solely from localized tumor regions, which may not fully spatially align with radiological features reflecting the entire tumor. Third, all MRI data acquisitions were acquired with a single scanner under fixed parameters. The repeatability of T2* measurements across different vendors, field strengths, and acquisition parameters remains to be established, necessitating technical standardization before widespread clinical adoption. Additionally, our study only focused on the relationship between T2* mapping and HIF-1α. Finally, we did not evaluate other potentially relevant parameters, such as diffusion-weighted imaging or DCE-MRI. In subsequent research, we intend to integrate multiparametric MRI to provide a more comprehensive characterization of the hypoxic tumor microenvironments.

Conclusions

The conventional rim enhancement feature in quantitative T2* mapping may serve as a noninvasive biomarker for evaluating hypoxia-related microenvironmental changes in PDAC. Our findings provide preliminary evidence indicating that this approach could potentially aid in the preoperative identification of more aggressive hypoxic phenotypes, particularly in poorly differentiated tumors, and may offer supportive information for personalized therapeutic decision-making, including the selection of candidates for neoadjuvant therapy. However, technical standardization, larger-scale multicenter validation, and multimodal studies are needed to confirm these observations and establish their role in optimizing PDAC management.

Supplementary

The article’s supplementary files as

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. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Our study was approved by the Medical Ethics Committee of Changzhou First People’s Hospital (No. 2023012). Informed consent was waived by the Medical Ethics Committee owing to the retrospective nature of the study.

Data Sharing Statement

Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-2020/dss

qims-16-04-291-dss.pdf ^{(99.7KB, pdf)}

DOI: 10.21037/qims-2025-2020