| Objectives:Pancreatic fibrosis（PF）is a pathological change of many pancreatic diseases such as chronic pancreatitis and pancreatic cancer,which is closely related to the pathogenesis,disease progression and treatment response of chronic pancreatitis and pancreatic cancer.Early detection and grading of PF is an important and challenging clinical goal.Based on multi-modal functional magnetic resonance imaging,this paper discusses the early diagnosis and grading of PF evaluated by image quantitative features.Methods:Prospective recruited adult patients（n=144）scheduled for pancreatic surgical procedures at Shengjing Hospital of China Medical University between Dec 2018 and May2020.Each had undergone abdominal MRI,IVIM-DWI,and T1 imaging preoperatively.Final diagnoses were based on postoperative histologic examinations of surgical specimens.According to the stage of PF after operation,PF grades were distributed as follows:F0,82;F1,22;F2,22;and F3,18.An in-house MATLAB script（v2018a;Math Works,Natick,MA,USA）was availed to draw Regions of interests（ROIs）which at the largest slice of proximal pancreatic parenchyma next to lesions.It is automatically converted from amplitude image to waveform diagram and elasticity diagram to obtain elasticity value in k Pa.We selected the slice having the largest sampling of proximal pancreatic parenchyma next to lesions,meantime,a direct inversion algorithm（Mayo Clinic）for automated on-scanner processing of wave images facilitated T1 mapping acquisition.For IVIM-DWI,we also selected the slice having the largest sampling of proximal pancreatic parenchyma next to lesions,while also choosing the slice with the fewest voxels in all b-values（to ensure that all voxels in ROI range were included in all b value slices）.ROIs were then drawn on proximal pancreatic parenchyma,adjacent to lesions.The post-processing software automatically copies ROIs to other b-value images,and all b-values are imported into a voxel-based nonlinear double exponential model,based on the following formula:Sb/S0=（1-f）×exp（-b×D）+f×exp（-b×（D+D*）.This application also automatically generates corresponding true diffusion coefficient（D）,pseudo diffusion coefficient（D*）,perfusion fraction（f）,and an IVIM-DWI signal attenuation curve.MRE images were closely matched with T1 and IVIM-DWI images to produce similar ROIs on matched slices.The ROIs were drawn to maximize pancreatic parenchymal areas.Two physicians with more than 5 years experiences in abdominal imaging diagnosis evaluated the imaging morphological characteristics of tumors and delineated ROI.Sections were then stained using Masson’s trichrome method,Hematoxylin-eosin and Sirius red stains to assess fibrous elements of pancreatic stump.ROC analysis served to evaluate stiffness,T1relaxation time,and IVIM-DWI parameters as determinants of PF.A multivariate logistic regression model helped explore associations between imaging variables（including T1relaxation time,stiffness,IVIM-DWI parameters,pancreatic thickness,and MPD diameter）and different PF grades.Results:Both pancreatic stiffness（r=0.754;P<0.001）and T1 relaxation time（r=0.433;P<0.001）correlated significantly with PF（%）.To determine PF grades≥F1,a combined model（area under the curve[AUC]=0.906）performed significantly better than pancreatic stiffness（AUC=0.855;P<0.001）or T1 relaxation time（AUC=0.754;P<0.001）alone.For PF grades≥F2 or grade F3,both the combined model（≥F2:AUC=0.910;F3:AUC=0.939）and pancreatic stiffness≥F2:AUC=0.906;F3:AUC=0.929）outperformed T1 relaxation time（≥F2:AUC=0.768[P=0.005 and P=0.004,respectively];F3:AUC=0.816[both P<0.005]）.All IVIM-DWI parameters generated AUC values<0.700.Conclusions:multimodality functional MRI can as a means of detecting and characterizing PF,particularly at an early stage.MRE and T1 mapping perform significantly better in combination（vs alone）in grading PF;and MRE separately surpassed T1 mapping.In multivariate analysis,both pancreatic stiffness and T1 relaxation time emerged as independent influencing factors of PF grade,with the combined model（pancreatic stiffness+T1 relaxation time）performing at high level... |
