Computed high-b-value high-resolution DWI improves solid lesion detection in IPMN of the pancreas

Objectives To examine the effect of high-b-value computed diffusion-weighted imaging (cDWI) on solid lesion detection and classification in pancreatic intraductal papillary mucinous neoplasm (IPMN), using endoscopic ultrasound (EUS) and histopathology as a standard of reference. Methods Eighty-two patients with known or suspected IPMN were retrospectively enrolled. Computed high-b-value images at b = 1000 s/mm2 were calculated from standard (b = 0, 50, 300, and 600 s/mm2) DWI images for conventional full field-of-view (fFOV, 3 × 3 × 4 mm3 voxel size) DWI. A subset of 39 patients received additional high-resolution reduced-field-of-view (rFOV, 2.5 × 2.5 × 3 mm3 voxel size) DWI. In this cohort, rFOV cDWI was compared against fFOV cDWI additionally. Two experienced radiologists evaluated (Likert scale 1–4) image quality (overall image quality, lesion detection and delineation, fluid suppression within the lesion). In addition, quantitative image parameters (apparent signal-to-noise ratio (aSNR), apparent contrast-to-noise ratio (aCNR), contrast ratio (CR)) were assessed. Diagnostic confidence regarding the presence/absence of diffusion-restricted solid nodules was assessed in an additional reader study. Results High-b-value cDWI at b = 1000 s/mm2 outperformed acquired DWI at b = 600 s/mm2 regarding lesion detection, fluid suppression, aCNR, CR, and lesion classification (p =  < .001–.002). Comparing cDWI from fFOV and rFOV revealed higher image quality in high-resolution rFOV-DWI compared to conventional fFOV-DWI (p ≤ .001–.018). High-b-value cDWI images were rated non-inferior to directly acquired high-b-value DWI images (p = .095–.655). Conclusions High-b-value cDWI may improve the detection and classification of solid lesions in IPMN. Combining high-resolution imaging and high-b-value cDWI may further increase diagnostic precision. Clinical relevance statement This study shows the potential of computed high-resolution high-sensitivity diffusion-weighted magnetic resonance imaging for solid lesion detection in pancreatic intraductal papillary mucinous neoplasia (IPMN). The technique may enable early cancer detection in patients under surveillance. Key Points • Computed high-b-value diffusion-weighted imaging (cDWI) may improve the detection and classification of intraductal papillary mucinous neoplasms (IPMN) of the pancreas. • cDWI calculated from high-resolution imaging increases diagnostic precision compared to cDWI calculated from conventional-resolution imaging. • cDWI has the potential to strengthen the role of MRI for screening and surveillance of IPMN, particularly in view of the rising incidence of IPMNs combined with now more conservative therapeutic approaches. Supplementary Information The online version contains supplementary material available at 10.1007/s00330-023-09661-6.


Introduction
Cross-sectional imaging has led to an increased discovery rate of pancreatic intraductal papillary mucinous neoplasms (IPMN) [1].IPMN represent precancerous lesions developing from the mucinous epithelium of the pancreatic duct [2].IPMN management ranges from surveillance to surgical resection, depending on the risk of a malignant transformation, which is estimated from imaging features, clinical symptoms, and laboratory values [3].Magnetic resonance imaging (MRI) has become a central diagnostic pillar in major international guidelines on the management of IPMN [3][4][5].
Whereas T2-weighted (T2w) imaging and magnetic resonance cholangiopancreaticography (MRCP) are highly sensitive for the detection and quantification of the fluid component, diffusion-weighted imaging (DWI) is highly sensitive for the detection of the solid component, which indicates malignant transformation [6,7].Previous studies showed promising results for high-b-value DWI in the identification of solid components in IPMN using b values up to 800 s/mm 2 [8].Similarly, b values of b ≥ 1000 s/mm 2 improve the detection rate in pancreatic ductal adenocarcinoma [9].However, the application of increased b values comes with longer acquisition times and exposition to motion artifacts [10].
Computed high-b-value DWI (cDWI) from at least two acquired lower b values based on a voxel-wise mono-exponential fit has been proposed to overcome the above-mentioned limitations [11][12][13][14].Moreover, highresolution DWI has been reported to increase image quality for both pancreatic DWI and cDWI [15,16].As diagnostic reliability in DWI stems from the signal intensity and image quality, integrating both cDWI and high-resolution DWI might strengthen the role of DWI in IPMNs.
Our primary aim was to compare the value of cDWI at b = 1000 s/mm 2 against acquired DWI at b = 600 s/mm 2 in IPMN.
We further investigated the value of high-resolution cDWI in pancreatic IPMN with regard to qualitative and quantitative parameters as well as diagnostic accuracy.Endoscopic ultrasound (EUS) and histopathology served as the standard of reference.

Study design
This study was designed as a single-center observational study.All patients undergoing the additional high-resolution DWI sequence were prospectively enrolled, and informed consent was given (IRB Protocol Nr. 102/21 S-EB).All other patients were retrospectively enrolled, and the requirement for informed consent was waived (IRB Protocol Nr. 180/17S).The study was conducted in accordance with the Declaration of Helsinki.
All patients were referred to the departments of surgery or internal medicine for work-up of suspected IPMN or followup of known IPMN.Patients who underwent an in-house MRI between February 2018 and June 2021 were included.Exclusion criteria were (1) IPMN size < 10 mm, (2) nondiagnostic image quality, (3) incomplete MRI datasets, and (4) entity other than IPMN.The patient inclusion flowchart can be found in the supplementary material (S.1).
The following clinical data were obtained: age at diagnosis, sex, IPMN type (main duct, side branch, mixed type) based on imaging criteria, and EUS as well as histopathology when available.EUS was performed by an experienced (> 7 years) gastroenterologist with specialization in EUS (Table 1).

Data acquisition and postprocessing
MRI datasets were acquired on a whole-body 3-T MRI system (Philips Ingenia Elition; Philips Medical Systems) using a combination of a 16-channel torso coil array and an inbuilt table posterior 12-channel coil array.
A T2-weighted (T2w) turbo spin echo (TSE) sequence and a conventional (i.e., full field of view, fFOV) 2D diffusion-weighted (DW) single-shot echo-planar imaging (ssEPI) sequence for whole upper abdomen coverage were acquired in all 88 patients.fFOV images were acquired at the following b values (averages) as routinely performed at our institution: 0 (1), 50 (1), 300 (2), 600 (5) s/mm 2 .In a subset of 39 patients, a high-resolution (i.e., reduced field of view, rFOV) 2D DW single-shot echo EPI sequence, covering the pancreas, was performed additionally, at the same b values as the fFOV sequence.Furthermore, in a subset of 21 patients with high-resolution DWI, an additional high b value at b = 1000 (11) s/mm 2 was obtained (r-aDWI1000).Figure 1 displays the image acquisition flow chart.Datasets were acquired in axial planes with respiratory triggering.Further sequence parameters are displayed in Table 2.
A dedicated software tool (Philips IntelliSpace Portal, Philips Medical Systems) was utilized to generate cDWI images.Therefore, b values of 0, 50, 300, and 600 s/mm 2 were used to calculate cDWI images at a b value of 1000 s/ mm 2 , based on a mono-exponential fit model, for both fFOV (f-cDWI1000) and rFOV (r-cDWI1000).

Quantitative analysis
A 5 mm measuring region of interest (ROI) was manually placed in the IPMN and the healthy-appearing tissue next to the lesion in the pancreatic head and tail.Nodules and prominent septations within the IPMN were omitted if possible.ROIs were placed in the b = 600 s/mm 2 images and then copied and pasted to the remaining images.The mean signal intensity (SI) and standard deviation (SD) inside the ROIs were recorded.
The apparent signal-to-noise ratio (aSNR) was calculated as follows:  The apparent contrast-to-noise ratio (aCNR) was calculated as follows: The contrast ratio between the lesion and the adjacent normal parenchyma was calculated as follows:

Reader study
We implemented a prospectively designed reader study to investigate whether alterations in qualitative and quantitative imaging parameters would affect radiological decisionmaking.Both readers were asked to qualitatively rate their diagnostic confidence regarding the presence or absence of diffusion-restricted nodules on a 4-point Likert scale (4 = definitely nodule, 3 = probably nodule, 2 = probably no nodule, 1 = no nodule) for acquired DWI and cDWI images.Both readers were blinded to the EUS result.A score of 3 or 4 implicated further work-up by EUS.A score of 1 or 2 would implicate no further EUS examination.
To provide consistency between the MRI and EUS findings, we searched our cohort for patients who underwent EUS for suspected IPMN after MRI or no longer than 3 months prior to the MRI.All eligible patients were included in the prospectively designed reader study.A total of 62 out of 88 patients in our cohort met the criteria for inclusion.In 5 patients, EUS and MRI were performed on the same day.In 24 patients, EUS was performed prior to the MRI scan (median: 25 days, IQR: 33 days), and in aSNR = SI normal parenchyma ∕SD normal parenchyma .aCNR = SI tumor − SI normal parenchyma ∕SD normal parenchyma .
Furthermore, three patients without EUS but with stable follow-up MRI as well as one patient with a stable follow-up PET-CT over the course of 2 years were included.One additional patient without EUS, but for whom pancreatectomy was performed within 3 months after the MRI, was also included.Finally, 67 patients were enrolled in the reader study.
In 18/67 (27%) patients, EUS revealed a solid nodule within the IPMN, indicating potential malignant transformation.Based on imaging findings and clinical indication, 20/67 patients were resected and histopathological correlation revealed malignancy in 13/20 patients.Histopathological examination was performed by experienced pathologists with a specialization in pancreatic pathologies.

Statistical analysis
Qualitative and quantitative metrics were analyzed using the Wilcoxon signed-rank test.Receiver operating curve (ROC) analysis was applied for diagnostic confidence.Sensitivity and specificity calculations were performed by binarising the Likert scale so that scores 1 and 2 were considered "negative" and 3 and 4, "positive."ROC curves were calculated by assigning a probability of 0.25 to Likert scale 1, 0.50 to 2, 0.75 to 3, and 1.0 to 4. The DeLong test for paired samples was used to compare areas under the ROC curves.

Patient characteristics
A total of 82 patients were finally enrolled.Patient characteristics can be found in Table 1.
Comparison between both cDWI images revealed a sensitivity, specificity, and ROC of 0.75, 0.93, and 0.83, respectively, for conventional images and 0.88, 0.96, and 0.91 for high-resolution images; p = 0.52.

Discussion
In our study, we examined the diagnostic performance of high-b-value cDWI and the added value of high-resolution high-b-value cDWI in pancreatic IPMN.Calculated high-bvalue cDWI outperformed acquired DWI at b = 600 s/mm 2 regarding lesion detection and diagnostic precision.Further diagnostic accuracy was achieved by combining cDWI and high-resolution DWI.Overall, the best image quality and highest diagnostic precision were found in high-resolution high-b-value cDWI.Of note, high-resolution cDWI was comparable to directly acquired high-resolution high-bvalue DWI.
The presence of mural nodules in imaging studies is a strong indicator of a malignant transformation of an IPMN [4,5].According to current European guidelines, the presence of a solid mass is considered an absolute indication for surgery in IPMN [4].
We found cDWI images to outperform DWI at b = 600 s/ mm 2 regarding lesion detection and delineation.In our sample, cDWI images revealed a significantly increased diagnostic accuracy in detecting solid nodules within IPMNs, reflected by higher sensitivity and specificity, compared to b = 600 s/mm 2 images.This subjective finding was confirmed by the significantly higher CR and CNR between the IPMN and the adjacent healthy pancreatic parenchyma and is attributable to the stronger fluid suppression in cDWI images.
Incomplete fluid suppression at b = 600 s/mm 2 not only obscures small diffusion-restricted nodules but also mimics solid components within IPMNs.Our results are in line with a previous study on cDWI in pancreatic ductal adenocarcinoma, reporting more accurate lesion detection b = 1000 s/ mm 2 [17].
Comparing conventional and high-resolution cDWI revealed increased lesion detection and higher ROC in favor of the high-resolution images.Improved lesion detection in rFOV-DWI compared to fFOV-DWI has been reported previously [16].In a recent meta-analysis, Xu et al reported that DWI slice thickness affected the diagnostic performance in discriminating benign from malignant IPMN [18].Hence, reduced T2 shine-through effect with higher b values and increased spatial resolution with thinner slices contribute to the improved diagnostic accuracy in high-b-value highresolution DWI as found herein.Fig. 6 displays the ROC curves for f-aDWI600 (0.54), f-cDWI1000 (0.86), and r-cDWI1000 (0.91) Our findings may be important in view of the paradigm shift in IPMN treatment, evolving from early surgical resection-with the risk of potential overtreatment-to close surveillance.Surveillance schemes vary among major guidelines, in part due to diverging results in favor of either MRI or EUS [19][20][21].However, the major advantage of MRI is its non-invasive nature.Hence, advances in MRI could strengthen the role of MRI in IPMN management.
Image quality decreased in cDWI compared to b = 600 s/ mm 2 images in our study.Reduced image quality has been recently reported in cDWI of the pancreas [13,22].We attributed our finding primarily to misregistration artifacts caused by misalignment due to through-plane motion artifacts in the acquired lower-b-value images.Advanced coregistration software could potentially increase image quality as proposed by Agkagi et al [11].
SNR significantly dropped in cDWI compared to b = 600 s/mm 2 images.Likewise, a study by Tamura et al reported decreased SNR in cDWI images in breast cancer patients [23].In a study by Gatidis et al, background intensity variation of cDWI increased monotonically with increasing b values [24].As the b values increase, signal intensity decreases, thus limiting the maximum b value applicable for cDWI [25].
Comparing qualitative image parameters between computed and directly acquired images revealed non-inferiority in cDWI images.Furthermore, aSNR, aCNR, and CR were significantly higher in cDWI images, which can be explained by the reduced acquisition time.These findings seem to be particularly interesting in light of the markedly reduced acquisition time for high-resolution cDWI compared to directly acquired high-resolution DWI in our cohort.However, it must be borne in mind that the rFOV-DWI sequence was employed as a study protocol.A scan time of 15 min as employed herein is not applicable to clinical routine.Hence, cDWI enables the generation of high-b-value images at a clinically justifiable examination time, supporting recent studies on the implementation of abbreviated MRI protocols in IPMNs [26].
Our study has limitations.First, cDWI images were primarily compared to standard DWI images at b = 600 s/ mm 2 as routinely performed at our institution.However, we are well aware that other institutions might use different b-values for pancreatic DWI.Hence, additional studies should elucidate if the potential benefits as shown in our study hold true for different b values.Second, we focused on a mono-exponential diffusion model.Other models, e.g., bi-tri-exponential diffusion models, could further add diagnostic value.Third, EUS was chosen as the standard of reference for most of our cases.However, as this is an operator-depended modality, a potential bias could not finally be ruled out.
Fourth, due to the above-mentioned limitations and particularly its retrospective and single-center nature, further prospective and multicenter cohorts need to ascertain the herein-found results.Fifth, we excluded all patients with IPMNs < 10 mm.However, patients with IPMNs > 10 mm were by far the most often referred group.Moreover, in our experience and also in view of previously published studies, the risk of malignant transformation/presence of solid nodules in cysts < 10 mm is extremely low [27,28].

Conclusion
Our study underlines the potential of high-b-value cDWI combined with high-resolution rFOV-DWI, in the imaging of pancreatic IPMN.These images provide better identification of solid nodules with no time penalty.Improvement of MRI might be important with the development of surveillance.
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Fig. 3
Fig. 3 In an 81-year-old female patient, T2w (A, B) and contrastenhanced T1w (C) images depict a BD-IPMN in the pancreatic head exhibiting focal contrast enhancement (arrow in C).The mural nodule is clearly better detectable in the high-b-value computed DWI

Fig. 4 AFig. 5
Fig. 4 A branch-duct IPMN in the pancreatic body is depicted in the axial T2w image (A).Better fluid suppression within the lesion is seen in the high-b-value computed image (C) compared to the standard b-value DWI (B)

Table 3
Ratings

Table 4
Corresponding p values for comparison of the qualitative image parameters

Table 5
The corresponding false-positive rate (FPR), false-negative rate (FNR), true-positive rate (TPR), and true-negative rate (TNR) obtained from the reader study based on EUS as the gold standard