Background

Pancreatic cancer (PC) is a highly lethal ailment that causes over 460,000 fatalities each year [1], with a 5-year survival rate ranging from 2 to 12% [2, 3]. As the annual mortality rate increases, PC-related fatalities may be the second leading cause of death in the United States by 2030 [4]. A significant proportion of patients are identified with distant metastasis or local tumor progression rather than timely surgical resection due to the absence of an efficacious screening method for PC and the asymptomatic or atypical symptoms observed during early phases of PC. Only 10–15% of patients meet the criteria to undergo curative resection [5]. Furthermore, the development of novel treatment strategies for PC is imperative, alongside its early detection.

A recently developed treatment strategy for PC [6,7,8,9,10,11], neoadjuvant therapy (NAT) decreases tumor stage, postoperative recurrence, and metastasis. This improves the number of patients who are able to undergo surgery for locally advanced and borderline resectable PC, as well as increases the percentage of tumors that are removed via R0 resection. Both CT and MRI yield comparable results for PC staging and diagnosis [12]. In clinical applications, quantitative MRI has emerged as an essential tool, not only aiding radiologists in their clinical diagnoses but also assisting in the elimination of diagnostic subjectivity and furnishing standardized, calibrated parameters to account for system variations [13]. Clinical practice frequently employs apparent diffusion coefficient (ADC) derived from diffusion-weighted imaging (DWI) [14, 15] techniques to quantify the diffusion of water molecules through tissues. Standard criteria for the collection and analysis of DWI have been suggested [16,17,18,19,20] in order to expand the potential of ADC to be utilized in a variety of organs. The characteristics of certain tissues, including cellularity, vascularity, and the Ki-67 index, have been identified to correlate with variations in the ADC values of pancreatic tumors [21,22,23]. ADC has been used to evaluate the effects of NAT on PC patients [24]. There is ongoing debate regarding the effects of ADC measurements in PC patients undergoing NAT therapy [25,26,27,28,29,30]. ADC values are substantially impacted by alterations in the tissue components of PC following NAT, including necrosis, edema, inflammation, fibrosis, and the elimination of tumor cells and adjacent tissue. In light of the limited sample size observed in previous investigations [25,26,27,28,29,30], it is necessary to clarify the values of ADC on PC patients with NAT, so as to augment the accuracy of quantitative DWI for assessing treatment response. The objective of this study was to assess the utility of ADC in predicting prognostic parameters such as R0 resectability or tumor regression grade (TRG) levels among patients with PC, and to evaluate the value of ADC for the treatment response evaluation of PC patients with NAT.

Methods

Patients

The retrospective was granted approval by the Biomedical Research Ethics Committee of our institution. Between April 2019 and December 2021, 1752 patients underwent contrast-enhanced MRI or CT and received pancreatic tumor surgery. A total of 125 patients (72 men, 53 women; mean age, 60.6 ± 8.5 years) who underwent surgical excision of malignancies and were given NAT. Among these patients, 109 patients underwent contrast-enhanced MRI examinations prior to NAT, and 115 individuals underwent contrast-enhanced MRI examinations after the completion of NAT. Finally, this study included 103 individuals who underwent MRIs in both the pre-NAT and post-NAT groups (Fig. 1). Out of the 103 cases, 39 patients underwent twice MRI examinations with the same scanner, one before and one after the NAT was concluded.

Fig. 1
figure 1

Flow chart of the patient selection process

Full size image

MRI

All patients underwent MRI examinations utilizing pancreatic protocols on 1.5-T or 3-T MRI systems (Signa HDxt and Discovery MR750, GE Healthcare, Milwaukee, USA; MAGNETOM Skyra and Avanto, Siemens, Erlangen, Germany), which include T1-weighted imaging (T1WI) with two echoes and water-fat Dixon reconstruction, contrast-enhanced T1WI (contrast agent, 0.1–0.15 mmol/kg; rate, 2.0–3.0 ml/s), DWI and magnetic resonance cholangiopancreatography (MRCP). DWI employed a single-shot free-breathing echo-planar-imaging sequence with three orthogonal diffusion gradients. Table 1 provides a summary of the DWI parameters utilized by MRI scanners at present. Tumor diameters were measured using the response evaluation criteria in solid tumors (RECIST) 1.1 criteria in conjunction with MRI.

Table 1 The diffusion weighted imaging parameters for the MRI scanners used in this study
Full size table

ADC measurements

The ADC maps were generated by utilizing a monoexponential model DWI (b-values: 50 and 800 s/mm2). By delineating one elliptical or circular region of interest (ROI) on the solid portion of the tumor for each patient (Figs. 2 and 3), with caution to exclude cystic lesions and dilated pancreatic duct by the collaboration of two measurers, who calculated the mean, minimum, and maximum ADC values for every patient both before and after NAT. A radiologist (12 years’ experience in abdominal MRI) and an MRI physicist (13 years’ experience in abdominal MRI) determined the ADCs utilizing Image J (v1.8.0; National Institutes of Health, Bethesda, USA) and RadiAnt DICOM viewer (v2021.1; Medixant, Poznan, Poland).

Fig. 2
figure 2

Representative MRI images and apparent diffusion coefficient (ADC) measurements of pancreatic cancer with tumor regression grade 0 before and after neoadjuvant therapy (NAT). MRI scans taken pre- and post-NAT are shown in (a-h) and (i-p). Figures (a, i) and (b, j) are diffusion-weighted images with b value of 50 and 800 s/mm2, respectively; (c, k) ADC maps; (d, i) T2-weighted images; (e, m) Pre-contrast water images of the DIXON VIBE; (f, n) Arterial phase T1WI; (g, o) Venous-phase T1WI; (h, p) Delayed-phases T1-weighted images; (q) H&E, magnification×40 and (r) H&E, magnification ×200. On the ADC maps, a rounded or oval-shaped region of interest (ROI) is drawn within the solid tumor areas for ADC measurements. The tumor’s mean, minimum, and maximum ADC values can be determined via an ROI (Pre-NAT: ROI, 280.8 mm2; mean: 1.036 × 10− 3 mm2/s; minimum: 0.761 × 10− 3 mm2/s; maximum: 1.287 × 10− 3 mm2/s; Post-NAT: ROI, 64.5 mm2; mean: 2.009 × 10− 3 mm2/s; minimum: 1.758 × 10− 3 mm2/s; maximum: 2.319 × 10− 3 mm2/s). (Yellow ROI examples in figures c and k)

Full size image
Fig. 3
figure 3

Representative MRI images and apparent diffusion coefficient (ADC) measurements of pancreatic cancer with tumor regression grade 3 pre- and post-neoadjuvant therapy (NAT). MRI scans taken pre- and post-NAT are shown in (a-h) and (i-p). Figures (a, i) and (b, j) are diffusion-weighted images with b value of 50 and 800 s/mm2, respectively; (c, k) ADC maps; (d, i) T2-weighted images; (e, m) Pre-contrast water images of the DIXON VIBE; (f, n) Arterial phase T1WI; (g, o) Venous-phase T1WI; (h, p) Delayed-phases T1-weighted images; (q) H&E, magnification×40 and (r) H&E, magnification ×200. On the ADC maps, a rounded or oval-shaped region of interest (ROI) is drawn within the solid tumor areas for ADC measurements. The tumor’s mean, minimum, and maximum ADC values can be determined via an ROI (Pre-NAT: ROI, 581.7 mm2; mean: 1.371 × 10− 3 mm2/s; minimum: 0.894 × 10− 3 mm2/s; maximum: 1.905 × 10− 3 mm2/s; Post-NAT: ROI, 202.7 mm2; mean: 1.740 × 10− 3 mm2/s; minimum: 1.443 × 10− 3 mm2/s; maximum: 2.164 × 10− 3 mm2/s). (Yellow ROI examples in figures c and k)

Full size image

Pathological analysis

The macroscopic description, specimen specifications, morphological description, margins and neighbors, lymph node metastasis, diagnosis, TRG, and other pertinent information were documented in the structured pathological reports for NAT of PC at our hospital. TRG is classified into four phases (Grades 0 to 3) by the College of American Pathologists (CAP) grading system, which is determined by the proportion of residual tumor cells to stroma. Grade 0 indicates complete response to PC with no surviving tumor cells, Grade 1 indicates the presence of a solitary or small number of focal residual tumor cells; Grade 2 indicates the presence of focal residual tumor cells accompanied by interstitial fibrosis; and Grade 3 indicates that the NAT was ineffective and a considerable number of tumor cells persisted. R0 and R1 were ascertained in the absence of tumor cell infiltration within a 1 mm margin of the resection. Prior to the issuance of final pathology reports, pathological images were analyzed by a minimum of two pathologists. The Ki-67 labeling index represents the percentage of nuclei positively stained among the whole tumor cells.

Statistical analysis

MedCalc (version 13.0.0.0, MedCalc Software, Ostend, Belgium) was utilized for statistical analysis. Normality assumptions of ADCs were assessed using the Kolmogorov-Smirnov test. Normal distribution data are expressed as mean ± standard deviation (range), and non-normally distributed data are expressed as median (range). Wilcoxon tests or T-tests were used to examine potential differences in the measured ADC values (mean, minimum, and maximum) between samples collected before and after NAT. NAT methods (chemotherapy, chemoradiation), tumor locations (head/neck, body/tail), TRG levels (0–2, 3), N stages (N0, N1/N2), and tumor resection margin status (R0, R1) were considered when comparing the mean, minimum, and maximum ADCs between the two groups before and after NAT using Mann-Whitney or independent samples T-tests. The gross, pre-NAT, and post-NAT MRI tumor sizes were compared using the Friedman test with post hoc analysis. The correction between the pre- and post-NAT ADCs, the rate of ADC changes post-NAT, and the rate of tumor size changes post-NAT were examined using Spearman correlation analyses. Furthermore, we also carried out correlation analyses to examine the correlations between ADC values and Ki-67 or the tumor grade. The threshold for statistical significance was established as P < 0.05.

Results

Patient characteristics

Table 2 provides a summary of the clinical characteristics of the 103 patients. The mean age of the 61 males and 42 females was 59.7 ± 8.6 years (range: 33–76). Of the total pancreatic lesions detected, 52.4% (54 of 103) were located in the head/neck area. Histological analyses identified lymph node metastases in 50% (51 of 102) of patients, excluding one individual who did not undergo peripancreatic lymphadenectomy. At the time of tumor resection, 74.8% (77 of 103) of the cases included in this study had R0 margin status. The median tumor size as determined by pre-NAT, preoperative MRI, and pathological measurements was 30.0 mm (range, 8.2 to 89 mm), 20.0 mm (range, 0 to 56.6 mm), and 28.0 mm (range, 5 to 75 mm), respectively. Among the three groups, PC had the smallest preoperative median MRI size. Ki-67 were recorded in 97 patients, with a median of 30% (range, 0 to 90%).

Table 2 Characteristics of the enrolled 103 patients with pancreatic cancer
Full size table

ADC measurements pre- and post-NAT

Subsequent to NAT, the mean ADC of tumors increased by 23.3%, the minimum ADC increased by 25.4%, and the maximum ADC increased by approximately 23.4% (Table 3). Pre- or post-NAT, the difference between the subgroups of all the investigated factors was not significant (P = 0.085 to 0.999) for the ADC measurements of PC on both the two groups in factors such as NAT methods, tumor locations, TRG levels, N stages, and resection margin status. In particular, pre- or post-NAT, the difference among the groups of TRG (G0/G1, G2, G3), was also not significant for the ADCmean, ADCmax and ADCmin with P values ranging from 0.070 to 0.719.

Table 3 Comparisons of ADC measurements of pancreatic cancer (n = 103) before and after neoadjuvant therapy (NAT)
Full size table

Further analysis revealed significant fluctuations in the ADC values, encompassing the minimum, mean, and maximum, when longitudinal comparisons of ADC measurements were conducted between the subgroups of all parameters examined before and after NAT (all P < 0.01). There was no correlation observed between the rate of change in tumor size and the pre-NAT ADC values or the rate of change in any of the three categories (mean, minimum, and maximum) following NAT completion (P = 0.237 to 0.967).

Furthermore, the analyses of ADC values before and after NAT treatment for the 39 patients who underwent MRI scans on the identical MRI scanners prior to and following NAT treatment yielded comparable outcomes to those observed in the 103 cases.

After NAT, the Ki-67 of patients with TRG levels of 0–2 (median, 20%; range, 0 to 90%) was significantly lower than that of patients with TRG level TRG 3 (median, 30%; range, 3–90%) with P value of 0.018. There was no correlation between ADC values (mean, minimum, and maximum) and Ki-67 (P = 0.102 to 0.789) or TRG levels (P = 0.090 to 0.973) pre- or post-NAT (Table 4).

Table 4 Correlations between ADC values and the Ki67 index, as well as tumor grade before and after neoadjuvant therapy (NAT)
Full size table

ADC measurements of borderline resectable PC

In this study, 47 patients with PC that was borderline resectable participated. Following NAT, the mean, minimum, and maximum ADC values of tumors increased by roughly 21.8%, 22.7%, and 21.6%, respectively. Before or after NAT, there was no significant difference (P = 0.077 to 0.956) between the subgroups of all tested parameters for the ADC measurements of PC on TRG levels (0–2, 3) and resection margin status (R0, R1). There was a correlation between the rate of tumor size variations following NAT and the mean ADC values prior to NAT (Spearman’s coefficient: 0.288, P = 0.049).

Discussion

To reduce the variability of ADC data, this study applies the identical combination (50 and 800 s/mm2) for ADC calculation across to all MRI systems. A multi-b-value DWI study utilizing six models suggests that DWI can be employed to examine changes in PC both before and after NAT [31]. The performance of the quantitative parameter of ADC changes before and after NAT was superior to that of the parameter’s own repeatability. Following NAT, the minimum, maximum, and mean ADCs for PC patients increased significantly, as shown in the present study; however, there was no significant difference in the percentage increases for R0 resectability or TRG levels. The mean tumor ADC of the 103 patients prior to NAT was 1.31 × 10− 3 mm2/s, which was within the range of ADCs associated with PC [32]. The average ADC values of the tumors from 103 patients increased by about 25.4% to 1.626 × 10− 3 mm2/s subsequent to NAT. The elimination of PC cells subsequent to NAT may facilitate water molecule diffusion and contribute to the increased ADC values.

Prior investigations concerning the impact of ADC on the prediction of NAT’s effectiveness for PC have produced controversial findings. An association between the pathological response of the tumor and the baseline mean ADC value was identified by Cuneo et al. [24] through the analysis of DWI images from seven patients who had resectable PC prior to NAT. In contrast to the non-responding patient group, the responding patient group exhibited a higher pre-NAT ADC baseline value. Chan et al. discovered, using a sample of thirteen cases, that ADC increased in 54% of instances after NAT, and in 95% of those instances, it increased by more than 10% [25]. Using genetically modified rodents as models for PC NAT, Trajkovic-Arsic et al. [33] discovered that ADC can be used to predict the early NAT response of PC. The study also compared the mean ADC before and after NAT in six PC patients and discovered that the cases with the greatest decrease in tumor size had the greatest ADC increases after NAT. The study conducted by Zimmermann et al. [28] examined 25 patients and found that the average ADC values were 1.32 × 10− 3 mm2/s prior to NAT completion and 1.43 × 10− 3 mm2/s subsequent to NAT completion [28]. The distinction between the two categories was not readily apparent. Multiple studies [25, 27, 30, 34] indicate that the average ADC for PC is considerably greater after NAT than it was prior to NAT. The findings corroborate our assertions. Following NAT, there is a possibility of decreased tumor cell density and stromal reorganization, both of which could impact the ability of water molecules to diffuse across the tissue and facilitate the utilization of DWI to identify alterations in ADC values. Our findings suggest that ADC effectively captures the changes that occur in PC following NAT, despite the non-uniformity of the NAT techniques employed in the patients. Further research will be guided by these noteworthy findings.

Studies on a variety of tumors have shown that ADC values can be used to predict pathological features [35,36,37]. Ki-67 is an indicator of cell proliferation and could be used for the evaluation of tumors. In the liver metastases from both breast and colorectal cancers, a significant correlation was observed between ADC values and Ki-67 indices. Conversely, in the case of pancreatic cancer metastases, no substantial correlations were detected [38]. Ki-67 demonstrates a significant correlation with ADC values in primary central nervous system lymphoma [39]. In meningioma, the authors observed an inverse correlation between Ki-67 and both ADCmean and ADCmax, while no significant correlation was found with ADCmin [40]. Surov et al. noted that ADC values are not predictive of molecular subtypes or lymph node metastases in invasive breast cancer. However, they did find a weak correlation between ADC and the expression levels of Ki-67 [41, 42]. For the pancreatic tumors, a negative correlation between pathological features such as tumor cellularity and Ki-67 and ADC values of pancreatic neuroendocrine tumors [23, 43]. Xie et al. found the ADCs of PC was not significantly correlated with Ki-67 [22]. In the current study, we also found that the ADCs of PC before and after NAT were not correlated with Ki-67 and TRG levels. Consequently, ADC values are not suitable as a proliferation biomarker for PC.

Two critical clinical indicators of NAT are the TRG levels of PC and the state of the resection margin. In this study, we analyzed the variations in ADC among PC patients with NAT, stratified by resection margin status and TRG levels, including the mean, minimum, and maximum ADC values. Before or after NAT, there was no discernible difference in any of the examined parameters between the subgroups. Pre- and post-NAT ADC measurements were unable to predict tumor treatment response, including margin status and TRG levels of PC, according to our findings. kada et al. [26] identified the pre-NAT ADC value as a predictor of R0 curability in patients with borderline resectable PC in a retrospective study. However, the authors of another prospective study found that the post-treatment ADC value may serve as a predictor of R0 resectability in patients with borderline resectable PC [27]. Contradictory findings may be attributable to the varied case sources and tiny sample sizes utilized in the studies. The grading of TRGs is predominantly determined by the existence of interstitial fibrosis induced by treatment and damaged tumor cells. Pancreatic cancer cells are extensively distributed and proliferate haphazardly in this investigation. In general, tumor cells are distributed across an extensive array of proliferative stroma, with a particular concentration in the vicinity of the tumor. Interstitial fibrosis induced by NAT is comparable in appearance to fibroproliferative stroma linked to PC and fibrosis resulting from chronic pancreatitis. Consequently, the grading of TRG may be imprecise and subjective. Conversely, clinical pathology is subject to certain constraints on account of the utilization of minuscule tissue samples. Additional analysis of 47 patients with equivocal resectable PC in this study revealed no differences in pre-treatment or post-treatment ADC measurements or TRG levels (0–2, 3) between the R0 and R1 groups.

The preoperative MRI tumor sizes subsequent to NAT were notably reduced in comparison to the pre-NAT MRI tumor sizes and the pathological size of PC (both P < 0.001), with a median underestimation of tumor size was approximately 10 mm and 8 mm, respectively. In prior publications, we delineated various rationales for the incongruity between the pathology findings and the imaging sizes of the tumors [44,45,46]. The present study examined the correlation between the pre-NAT mean ADC values and the rate of variations in tumor size for the borderline resectable PC. Hence, the mean ADCs prior to NAT could potentially serve as a metric for assessing the effectiveness of NAT in the treatment of pancreatic cancer [30], although its value requires additional verification.

There were several limitations to this investigation. First, there were variations in the neoadjuvant techniques utilized in this study. For example, albumin-bound paclitaxel, gemcitabine, FOLFIRINOX, and mFOLFIRINOX have all been utilized in chemotherapy. Due to the imperfection of the post-NAT staging system and the evolution of NAT protocols and treatment recommendations in recent years, it is difficult to manage PC patients in clinical practice [47]. Second, although previous research has shown that employing multiple b values can lead to more precise ADC value measurements [48], in our clinical context of pancreatic MRI, ADC maps were constructed using DWI with only two b values in order to balance scanning time. Third, PC tumor cells are often distributed across a diverse array of stroma, particularly in the immediate vicinity of the tumor. It is challenging to describe the contour of the tumor, particularly after NAT, and imaging evaluation typically underestimates the number of tumor cells dispersed throughout the tumor [10]. In this investigation, the ADC values were therefore measured using the solid sample ROI method [49, 50]. Prior research has extensively investigated the impact of ROI approaches, including whole-volume, single-slice, and solid tumor samples, on the measurements of ADC values [49, 50]. However, we refrained from evaluating this particular influence.

Conclusions

Increasing percentage values did not predict the R0 resectability and TRG levels in PC patients, despite the fact that the minimum, maximum, and mean ADC values all increased substantially after NAT for the examinations done in the routine praxis.