Introduction

Pancreatic cancer is the seventh leading cause of global cancer deaths in industrialized countries [1] and the third leading cause of cancer-related deaths in the United States [2]. Pancreatic ductal adenocarcinoma (PDAC) is the most common type of pancreatic cancer, accounting for 90% of all pancreatic cancers [3]. Despite the advancement of diagnostic techniques, early diagnosis of PDAC is still challenging, and its incidence is estimated to continue to increase [4]. Autoimmune pancreatitis (AIP) is a rare autoimmune disorder that can cause similar symptoms to PDAC [5]. Diffuse enlargement of the pancreas (sausage-like) and low-attenuating rim-like capsule on contrast-enhanced computed tomography are well-known typical imaging findings of AIP [5]; however, 21.7–60.0% of AIP present as focal mass-forming pancreatitis [6,7,8,9]. Treatment options are completely different between AIP and PDAC; therefore, accurate differential diagnosis is required. Several reports have shown that magnetic resonance imaging (MRI) might be useful for distinguishing focal AIP from PDAC [6,7,8,9,10,11,12]; however, such differentiation remains challenging.

Diffusion-weighted imaging (DWI) has been used routinely in daily clinical practice owing to its excellent contrast resolution between lesions and the pancreatic parenchyma, without the use of contrast agents. Its usefulness in the detection and characterization of pancreatic diseases has been reported [13,14,15]. DWI with b-values of 800–1000 s/mm2 is widely used; however, higher b-values can be useful for the detection and characterization of PDAC [16] because diffusion-restricted tissues show relatively higher signal intensity (SI) than the normal pancreatic parenchyma with the increasing b-values. However, DWI with higher b-values has certain disadvantages, including the longer acquisition time and poorer image quality [16]. Computed DWI (cDWI) is a technique that can synthesize arbitrary target b-value DWI from a set of directly acquired b-value images by voxel-wise fitting [17]. cDWI can generate images with a higher diffusion effect than that achieved by clinical MRI units, as well as a higher signal-to-noise ratio in shorter acquisition time than with directly acquired DWI [17].

The usefulness of cDWI has been reported for several organs, such as the prostate [18], breast [19], liver [20], uterus [21], ovary [22], and middle ear [23]; however, only few reports are available on pancreatic cDWI [24, 25]. Moreover, several studies have reported the usefulness of DWI for AIP diagnosis [9,10,11], whereas the usefulness of cDWI in AIP has not been clarified. Thus, the purpose of this study was to assess the feasibility of cDWI in visualizing PDAC and focal AIP comparison with that of directly acquired DWI.

Materials and methods

Patients

This single-center, retrospective, cross-sectional study was approved by the relevant institutional review board, who waived the requirement for obtaining written informed patient consent due to the retrospective nature of the study. Patients with PDAC or AIP were consecutively enrolled between April 2012 and January 2017. The following inclusion criteria were used for PDAC: (i) pathologically diagnosed by fine needle aspiration or resection, and (ii) availability of 3.0-T MRI data within three months before fine needle aspiration or resection; for AIP, the criteria were (i) clinically diagnosed based on clinical diagnostic criteria in Japan (JPS2011) [26], (ii) availability of 3.0-T MRI data before steroid therapy, and iii) focal type.

Of the 181 patients enrolled for the study, 46 patients were excluded (Fig. 1). The final study cohort consisted of 135 patients (mean age, 68.2 ± 10.2 [range 40–88] years), including 111 patients with PDAC and 24 with AIP (Fig. 1).

Fig. 1
figure 1

Flowchart of patients’ enrollment. Of the 181 patients enrolled for the study, 46 patients were excluded. The final study cohort consisted of 135 patients, including 111 patients with pancreatic ductal adenocarcinoma and 24 with autoimmune pancreatitis. FNA fine needle aspiration, MRI magnetic resonance imaging

Full size image

DWI protocols

All MRI examinations were performed using a 3.0-T MR system (Discovery 750; GE Healthcare, Waukesha, WI, USA) with a 32-channel phased-array coil. The DWI data were acquired in the transverse plane by respiratory-triggered single-shot echo-planar imaging with water-selective excitation, using the respiratory triggering technique. Sections of 5 mm in thickness with no intersection gap were used to cover the pancreas. The following three b-values were used: 0, 500, and 1000 s/mm2, with three axes [x (RL), y (AP), and z (SI)] motion-probing gradient directions. The pulse sequence parameters were as follows: repetition time, 3000–10000 ms (based on the respiratory interval); echo time, 70 ms; flip angle, 90°; field of view, 36 × 36 cm; matrix, 128 × 192; number of excitations, 8; sensitivity encoding acceleration factor, 2; and acquisition time, 150–180 s. Then, DW images with b-values of 0 and 1000 s/mm2 were digitally transferred to dedicated post-processing software (SYNAPSE VINCENT; FUJIFILM Medical, Tokyo, Japan), and cDW images were generated with target b-values of 1500 (cDWI1500), 2000 s/mm2 (cDWI2000), and 3000 s/mm2 (cDWI3000) by fitting a mono-exponential model, to compare them with the directly acquired DW images with b-values of 500 (DWI500) and 1000 s/mm2 (DWI1000).

Qualitative image analysis

The directly acquired DW images (DWI500 and DWI1000) and the cDW images (cDWI1500, cDWI2000, and cDWI3000) were reviewed by two independent radiologists (with 11 and 3 years of clinical experience in abdominal MRI) who were blinded to the clinical data aside from the information that the patients had PDAC or AIP based on other MRI sequences. For each dataset, the two radiologists evaluated the image quality using a 4-point visual score (4, excellent = the whole pancreas is clearly shown without artifacts; 3, good = minor degradation is present but suitable for the evaluation of the whole pancreas; 2, fair = only part of the pancreas is visible; 1, poor = the pancreas is barely visible) (Fig. 2a) and classified the SIs of the lesions, as follows: type 1, clearly demarcated hyperintensity relative to the surrounding pancreas; type 2, hyperintensity, but with an unclear distal (tail sided) border because of hyperintense distal pancreatic parenchyma; and type 3, iso-intensity relative to the surrounding pancreas or no evidence of the lesions (invisible) [16, 24] (Fig. 2b).

Fig. 2
figure 2

Example images of image quality grading and signal intensity types of lesions. a Image quality was assessed using a 4-point visual score (4, excellent = the whole pancreas is clearly shown without artifacts; 3, good = minor degradation is present but suitable for the evaluation of the whole pancreas; 2, fair = only part of the pancreas is visible; 1, poor = the pancreas is barely visible). b Signal intensity types of lesions were classified as follows: type 1, clearly demarcated hyperintensity relative to the surrounding pancreas; type 2, hyperintensity, but with an unclear distal (tail sided) border of the lesions because of hyperintense distal pancreatic parenchyma; and type 3, iso-intensity relative to the surrounding pancreas or no evidence of the lesions (invisible)

Full size image

Quantitative image analysis

The same radiologists who performed qualitative image analysis also conducted quantitative measurements for the following: (a) the signal intensity ratio (SIR) of the lesions and proximal (head sided) or distal (tail sided) pancreatic parenchyma to spinal cord and (b) the contrast ratio (CR) of the lesions to the proximal or distal pancreatic parenchyma, using four manually defined, circular or oval regions of interest (ROIs) (proximal and distal pancreatic parenchyma, lesions, and spinal cord) for each DW image. The ROIs were first placed on DWI500, and then, the size, shape, and location of the ROIs were kept constant for all images of each patient by applying a copy-and-paste function on the monitor. The ROIs were carefully placed to avoid pancreatic ducts, cystic lesions, vessels, peripancreatic fat, or artifacts within the ROIs. If adequate areas were not available for measuring the proximal or distal pancreatic parenchyma due to the locations of the lesions, the sections were excluded from the evaluations. The SIR and CR were calculated using the following formulae [16, 24], using the average SI for the calculations:

$$ {\text{SIR}}\;{\text{of}}\;{\text{the}}\;{\text{proximal}}\;{\text{pancreatic}}\;{\text{parenchyma}}\;{\text{to}}\;{\text{the}}\;{\text{spinal}}\;{\text{cord}}\;\left( {{\text{SIR}}_{\text{proximal}} } \right) = \frac{{{\text{SI}}\;{\text{of}}\;{\text{the}}\;{\text{proximal}}\;{\text{pancreatic}}\;{\text{parenchyma}}}}{{{\text{SI}}\;{\text{of}}\;{\text{the}}\;{\text{spinal}}\;{\text{cord}}}} $$
$$ {\text{SIR}}\;{\text{of}}\;{\text{the}}\;{\text{distal}}\;{\text{pancreatic}}\;{\text{parenchyma}}\;{\text{to}}\;{\text{the}}\;{\text{spinal}}\;{\text{cord}}\;\left( {{\text{SIR}}_{\text{distal}} } \right) = \frac{{{\text{SI}}\;{\text{of}}\;{\text{the}}\;{\text{distal}}\;{\text{pancreatic}}\;{\text{parenchyma}}}}{{{\text{SI}}\;{\text{of}}\;{\text{the}}\;{\text{spinal}}\;{\text{cord}}}} $$
$$ {\text{SIR}}\;{\text{of}}\;{\text{the}}\;{\text{lesion}}\;{\text{to}}\;{\text{the}}\;{\text{spinal}}\;{\text{cord }}\left( {{\text{SIR}}_{\text{lesion}} } \right) = \frac{{{\text{SI}}\;{\text{of}}\;{\text{the}}\;{\text{lesion}}}}{{{\text{SI}}\;{\text{of}}\;{\text{the}}\;{\text{spinal}}\;{\text{cord}}}} $$
$$ {\text{CR}}\;{\text{of}}\;{\text{the}}\;{\text{lesion}}\;{\text{to}}\;{\text{the}}\;{\text{proximal}}\;{\text{pancreatic}}\;{\text{parenchyma}}\;\left( {{\text{CR}}_{\text{proximal}} } \right) = \frac{{\left( {{\text{SI}}\;{\text{of}}\;{\text{the}}\;{\text{lesion}} - {\text{SI}}\;{\text{of}}\;{\text{the}}\;{\text{proximal}}\;{\text{pancreatic}}\;{\text{parenchyma}}} \right)}}{{\left( {{\text{SI}}\;{\text{of}}\;{\text{the}}\;{\text{lesion}} + {\text{SI}}\;{\text{of}}\;{\text{the}}\;{\text{proximal}}\;{\text{pancreatic}}\;{\text{parenchyma}}} \right)}} $$
$$ {\text{CR}}\;{\text{of}}\;{\text{the}}\;{\text{lesion}}\;{\text{to}}\;{\text{the}}\;{\text{distal}}\;{\text{pancreatic}}\;{\text{parenchyma}}\;\left( {{\text{CR}}_{\text{distal}} } \right) = \frac{{\left( {{\text{SI}}\;{\text{of}}\;{\text{the}}\;{\text{lesion}} - {\text{SI}}\;{\text{of}}\;{\text{the}}\;{\text{distal}}\;{\text{pancreatic}}\;{\text{parenchyma}}} \right)}}{{\left( {{\text{SI}}\;{\text{of}}\;{\text{the}}\;{\text{lesion}} + {\text{SI}}\;{\text{of}}\;{\text{the}}\;{\text{distal}}\;{\text{pancreatic}}\;{\text{parenchyma}}} \right)}}. $$

Statistical analyses

Patient demographic data, SIR, and CR were compared between PDAC and AIP by Wilcoxon test and χ2 test. The size and location of the lesions were determined on MR images. Receiver operating characteristic (ROC) analyses were performed for SIRs and CRs that were significantly different between PDAC and AIP. The image quality was compared among the five DWI protocols by Friedman test, followed by Scheffe’s paired comparison. The SI types of lesions were compared among the five DWI protocols by χ2 test, followed by Wilcoxon signed-rank test. The SIRs and CRs were compared among the five DWI protocols by Friedman test. Cohen’s kappa values (κ) or intraclass correlation coefficients (r) were calculated to assess interobserver agreement. Agreement was considered excellent for κ or r > 0.8, good for 0.6 < κ or r ≤ 0.8, moderate for 0.4 < κ or r ≤ 0.6, fair for 0.2 < κ or r ≤ 0.4, and poor for κ or r ≤ 0.2. Coefficient of variation of SIRs and CRs was also calculated and compared between two readers by F test. Data from the first reader were used for the qualitative and quantitative analyses, while those from the second reader were used to calculate interobserver agreement.

All statistical analyses were performed using JMP software (version 14.2.0; SAS Institute Inc., Cary, NC, USA) and BellCurve for Excel (version 3.20; Social Survey Research Information Co., Ltd., Tokyo, Japan). P-values < 0.05 were considered statistically significant.

Results

Patients’ characteristics

The patients’ demographics and clinical characteristics are presented in Table 1. A significant difference was observed in the size of the lesions between PDAC and AIP (P = 0.001). Other factors including age, sex, body weight, and location of the lesions were not significantly different between groups (P = 0.059–0.817; Table 1).

Table 1 Patient characteristics
Full size table

Qualitative image analysis

The breakdown of image quality using the five DWI protocols is shown in Table 2 and Fig. 3a. There was a significant difference among the five DWI protocols (P < 0.001). In the paired comparison, no significant differences were observed between DWI500 and DWI1000 (P = 0.968), DWI500 and cDWI1500 (P = 0.183), and DWI1000 and cDWI1500 (P = 0.548). In all other combinations, DWI protocols with smaller b-values showed significantly higher median image quality than those with higher b-values (all P < 0.001).

Table 2 Results of qualitative and quantitative image analysis on each protocol
Full size table
Fig. 3
figure 3

The breakdown of image quality and signal intensity types of lesions. a Image quality. There was a significant difference among the five diffusion-weighted imaging (DWI) protocols (P < 0.001). b Signal intensity types of lesions. There was a significant difference among the five DWI protocols in both pancreatic ductal adenocarcinoma (P < 0.001) and autoimmune pancreatitis (P = 0.015). Image quality was analyzed by Friedman test, and signal intensity types of lesions were analyzed by χ2 test. cDWI computed diffusion-weight imaging

Full size image

The breakdown of SI types of lesions using the five DWI protocols is shown in Table 2 and Fig. 3b. In PDAC, there were significant differences among the five DWI protocols (P < 0.001). In the paired comparison, no significant difference was observed between cDWI1500 and cDWI3000 (P = 0.627). A higher incidence of type 1 lesions was found with cDWI2000 than with cDWI3000 (P = 0.002). In all other combinations, the incidence of type 1 lesions was significantly higher on DWI protocols with higher than with lower b-values (P < 0.001–0.002). In AIP, there were significant differences among the five DWI protocols (P = 0.015). In the paired comparison, the incidence of type 1 lesions was significantly lower with DWI500 and cDWI3000 than with other protocols (P = 0.003 for DWI500 vs DWI1000, P = 0.001 for DWI500 vs cDWI1500 and for DWI500 vs cDWI2000, and P = 0.022 for cDWI3000 vs cDWI1500 and for cDWI3000 vs cDWI2000). In all other combinations, no significant differences were observed (P < 0.100–0.328).

Quantitative image analysis

The mean size of ROIs was as follows: PDAC, 224.1 ± 303.6 mm2; AIP, 92.9 ± 75.1 mm2; proximal pancreatic parenchyma, 126.0 ± 49.3 mm2; distal pancreatic parenchyma, 133.3 ± 70.7 mm2; spinal cord, 28.3 ± 5.8 mm2.

The SIR and CR using the five DWI protocols are shown in Table 2 and Fig. 4. There were significant differences in both ratios among the five DWI protocols (P < 0.001) for both PDAC and AIP. SIRproximal, SIRlesion, and SIRdistal decreased, whereas CRproximal and CRdistal increased, as the b-value increased (Fig. 4). Comparison of PDAC and AIP showed significantly higher SIRproximal and SIRlesion on all cDWIs, and significantly higher SIRdistal on cDWI2000 and cDWI3000 in AIP than in PDAC (P < 0.001–0.031; Table 3). In contrast, there were no significant differences for CRproximal and CRdistal on all DWI protocols between PDAC and AIP (P = 0.194–0.961; Table 3). The area under the ROC curve (AUC) of SIRlesion was significantly lower on cDWI1500 than on cDWI2000 and cDWI3000 (P < 0.001), whereas there was no significant difference in the AUC of SIRlesion between cDWI2000 and cDWI3000 (P = 0.056). Moreover, the AUC of SIRdistal was significantly higher on cDWI3000 than on cDWI2000 (P = 0.001), while that of SIRproximal was not significantly different among cDWI1500, cDWI2000, and cDWI3000 (P = 0.514–1.000; Fig. 5).

Fig. 4
figure 4

Differences in signal intensity ratio and contrast ratio among the five diffusion-weighted imaging (DWI) protocols. SIRproximal, SIRlesion, and SIRdistal decreased, whereas CRproximal and CRdistal increased, as the b-value increased in both pancreatic ductal adenocarcinoma and autoimmune pancreatitis. cDWI computed diffusion-weight imaging, SIRproximal signal intensity ratio of the proximal pancreatic parenchyma to the spinal cord, SIRlesion signal intensity ratio of the proximal pancreas to the lesion, SIRdistal signal intensity ratio of the distal pancreatic parenchyma to the spinal cord, CRproximal contrast ratio of the lesion to the proximal pancreatic parenchyma, CRdistal contrast ratio of the lesion to the distal pancreatic parenchyma

Full size image
Table 3 Differences in signal intensity ratio and contrast ratio between pancreatic ductal adenocarcinoma and autoimmune pancreatitis
Full size table
Fig. 5
figure 5

Receiver operating characteristic analysis of signal intensity ratio. The AUC of SIRlesion was significantly lower on cDWI1500 than on cDWI2000 and cDWI3000 (P < 0.001); there was no significant difference between the AUCs of SIRlesion on cDWI2000 and cDWI3000 (P = 0.056). The AUC of SIRdistal was significantly higher on cDWI3000 than on cDWI2000 (P = 0.001). The AUC of SIRproximal was not significantly different among cDWI1500, cDWI2000, and cDWI3000 (P = 0.514–1.000). The AUCs were compared by χ2 test. Abbreviations: cDWI, computed diffusion-weight imaging; SIRproximal, signal intensity ratio of the proximal pancreatic parenchyma to the spinal cord; SIRlesion, signal intensity ratio of the proximal pancreas to the lesion; SIRdistal, signal intensity ratio of the distal pancreatic parenchyma to the spinal cord; AUC, area under the curve

Full size image

Interobserver agreement and coefficient of variation

Interobserver agreement was excellent for the SI types of lesions on DWI500, DWI1000, cDWI1500, and cDWI2000 (r = 0.817–0.848) and for CRproximal on DWI500 (κ = 0.814) and was good for other protocols (r or κ = 0.575–0.790; Table 4). There was no significant difference in all the coefficient of variation of SIR and CR between two readers (P = 0.106–0.995; Table 5). Case examples are shown in Figs. 6 and 7.

Table 4 Interobserver agreement between two radiologists
Full size table
Table 5 Coefficient of variation between two radiologists
Full size table
Fig. 6
figure 6

Representative images of pancreatic ductal adenocarcinoma in a 78-year-old woman. Arterial phase of gadoxetate disodium-enhanced 3D fat-saturated T1-weighted imaging (repetition time/echo time, 3.44/1.43; flip angle, 12°) shows a hypointense lesion measuring 22 mm in diameter (arrow) in the pancreatic tail. On diffusion-weighted imaging (DWI) with a b-value of 500 s/mm2 (DWI500), the lesion (arrow) shows hyperintensity with an unclear distal border (type 2). On DWI with a b-value of 1000 s/mm2 (DWI1000) and computed DWI with target b-value of 1500 (cDWI1500), 2000 (cDWI2000), and 3000 (cDWI3000) s/mm2, the lesion (arrow) shows clear hyperintensity relative to the distal pancreatic parenchyma (dotted arrow) (type 1); however, on cDWI3000, the distal pancreatic parenchyma is almost invisible

Full size image
Fig. 7
figure 7

Representative images of autoimmune pancreatitis in a 48-year-old man. Unenhanced 3D fat-saturated T1-weighted imaging (repetition time/echo time, 4.34/1.43; flip angle, 15°) shows a vague hypointense lesion measuring 18 mm in diameter (arrow) in the pancreatic tail. On diffusion-weighted imaging (DWI) with a b-value of 500 (DWI500) and 1000 (DWI1000) s/mm2, and computed DWI with target b-value of 1500 (cDWI1500), 2000 (cDWI2000), and 3000 (cDWI3000) s/mm2, the lesion (arrow) shows hyperintensity relative to the proximal (arrowhead) and distal pancreatic parenchyma (dotted arrow) (type 1); however, on DWI500, the border between the lesion and the pancreatic parenchyma is somewhat unclear

Full size image

Discussion

This retrospective study revealed that image quality was significantly higher with DWI500, DWI1000, and cDWI1500 than with cDWI2000 and cDWI3000. The incidence of clear hyperintense (type 1) PDAC was the highest on cDWI2000, followed by cDWI1500 and cDWI3000. The incidence of clear hyperintense (type 1) AIP was significantly higher on DWI1000, cDWI1500, and cDWI2000 than on DWI500 and cDWI3000. Interobserver agreement was good to excellent for all items. These results suggest that cDWI1500 or cDWI2000 are the most effective among the five DWI protocols, consistent with a previous report [24].

It is challenging to obtain directly acquired DW images at b-values of 1500 s/mm2 for the pancreas because the image quality becomes worse and the acquisition time becomes longer as the b-value increases. cDWI can produce DW images without decreasing the signal and in a shorter acquisition time than with directly acquired DWI. Thus, cDWI1500 generated from DW images with b-values of 0 and 1000 s/mm2 may be useful for pancreas imaging. Image contrast on DWI varies greatly with the b-value. At higher b-values, tissues with high water molecule path lengths, such as the pancreatic parenchyma, tend to lose signal rapidly, while tissues with restricted water diffusion, including PDAC, yield relatively higher signals [27, 28]. This explains why the incidence of clear hyperintense (type 1) PDAC on cDW images with b-values ≥1500 s/mm2 was higher than that on DWI500 and DWI1000. Several reports have shown a lower apparent diffusion coefficient value for AIP than for PDAC with b-values 500–1000 s/mm2 [9, 11, 12, 29], which might explain why the incidence of clear hyperintense (type 1) AIP on DWI1000 was equivalent to that on cDWI1500 and cDWI2000.

In our quantitative image analysis, all SIRs (SIRproximal, SIRlesion, and SIRdistal) decreased and all CRs (CRproximal and CRdistal) increased as the b-value increased. The result of SI decrease can be explained by the fact that higher b-values yield lower signal-to-noise ratio [27, 28]. The result of CRdistal is consistent with that of a previous study, while the result of CRproximal is not [24]. The authors reported no significant difference in PDAC to proximal pancreatic parenchymal CR among DWI1000, cDWI1500, and cDWI2000. This discrepancy may be caused by the different MRI scanners, scanning parameters, and post-processing software used. Further studies are needed to determine the optimal settings of cDWI for PDAC. When comparing PDCA and AIP, SIRproximal and SIRlesion on all cDWI protocols and SIRdistal on cDWI2000 and cDWI3000 were significantly higher in AIP than in PDAC. The results of SIRlesion are consistent to those of previous reports showing lower apparent diffusion coefficient values in AIP than in PDAC when using b-values 500–1000 s/mm2 [9, 11, 12, 29]. Increased cellularity due to dense infiltration of plasma cells and lymphocytes, chronic inflammatory changes with fibrosis, and edematous changes in AIP may be associated to the high signal intensity [5, 30]. It is not clear why SIRproximal and SIRdistal AIP were higher in AIP than in PDAC; however, the surrounding pancreatic parenchyma may also be infiltrated with plasma cells and lymphocytes, although this cannot be detected by imaging because of its autoimmune nature [31]. In this study, we found no significant differences in CRproximal and CRdistal on all DWI protocols between PDAC and AIP. Therefore, it may be difficult to distinguish AIP from PDAC by visual evaluation. The AUC of SIRlesion on cDWI1500 was significantly lower than that on cDWI2000 and cDWI3000, whereas there was no significant difference between the AUCs of SIRlesion on cDWI2000 and cDWI3000. The AUC of SIRdistal on cDWI3000 was significantly higher than that on cDWI2000 (P = 0.001). These results indicate that cDWI2000 and cDWI3000 are better for quantitative analysis than cDWI1500; however, lower image quality may be a problem in clinical practice. Validation studies are desired as a next step.

Our study has some limitations. First, AIP of various inflammatory activities was included in this study. The phase of inflammation can influence signal intensity on DWI because of differences in dense infiltration of plasma cells and lymphocytes or edematous changes. Second, significant differences in the size of lesions were observed between PDAC and AIP, possibly due to the retrospective study design. Degeneration or necrotic changes are observed more frequently in larger lesions, especially in PDAC, which can also influence the signal intensity of DWI. Third, the retrospective nature and relatively small number of AIP cases in this study were also limitations. Further prospective studies with a larger sample size are necessary.

In summary, cDWI1500 or cDWI2000 generated from DW images obtained with b-values of 0 and 1000 s/mm2 were found to be the most effective among the five tested DWI protocols (DWI500, DWI1000, cDWI1500, cDWI2000, and cDWI3000) for visualizing PDAC and focal AIP; however, the AUC of SIRlesion was significantly lower on cDWI1500 than on cDWI2000 and cDWI3000. Therefore, the combination of cDWI1500 and cDWI2000/cDWI3000 may be effective in diagnosing AIP.