| Objective:To explore the correlation between T2DM and pancreatic fat deposition by using magnetic resonance IP/OP imaging technique.And analyze the difference of texture features between T2DM pancreas and normal pancreas on MRI.Methods:Part Ⅰ:Quantitative correlation between pancreatic fat deposition and T2DM by magnetic resonance IP/OP technique.The imaging data,clinical basic features and laboratory indexes related to fat metabolism of 66 patients with T2DM and 84 patients with normal blood glucose who underwent 3.0T MRI upper abdomen examination in the affiliated Hospital of North Sichuan Medical College from September 2017 to September 2019 were retrospectively analyzed.Pancreatic fat fraction was quantified by IP/OP technique,and pancreatic fat deposition was quantified by pancreatic fat fraction.Part Ⅱ:Analysis of Pancreatic texture in patients with T2DM based on 3.0T Magnetic Resonance Imaging.The imaging data and clinical basic features of 30 T2DM patients and 30 patients with normal blood glucose who received 3.0T MRI upper abdominal examination in the affiliated hospital of north Sichuan medical college from September 2017 to June 2020 were analyzed retrospectively.the texture features were extracted and analyzed by using the built-in radiomics plate of 3D Slicer.Results:Part Ⅰ:1.In T2DM group,the mean PFF(3.7-10.56%),pancreatic head PFF(2.34-11.08%),pancreatic body PFF(2.25-11.33%)and pancreatic tail PFF(2.34-13.46%)And pancreatic head PFF(1.38-7.88%),pancreatic body PFF(0.77-8.59%)and pancreatic tail PFF(1.38-9.58%),with statistical significance(p<0.05).2.The lipid metabolism indexes of triglyceride(0.66-3.22mmol/L)and very low density lipoprotein cholesterol(0.34-1.76mmol/L)in T2DM group were higher than those of normal blood glucose group(0.53-2.37mmol/L).However,the level of high density lipoprotein cholesterol in T2DM group(0.7-1.46mmol/L)was lower than that in normal blood glucose group(0.87-1.61mmol/L),and the difference of the above fat metabolism indexes between the two groups was statistically significant(p<0.05).The PFF in T2DM group was positively correlated with age(r=0.366,p<0.05)and body mass index(r=0.303,P<0.05).3.The AUC of mean PFF to diagnose T2DM(0.703)is larger than that of pancreatic head PFF(0.655),pancreatic body PFF(0.664)and pancreatic tail PFF(0.640).Part Ⅱ:A total of 107 texture parameters were obtained,and the results showed that the differences of 33 texture feature parameters were statistically significant.The features with statistical differences are further screened out by using lasso regression model,The results were 10Percentile and Small Area Low Gray Level Emphasis.The 10Percentile characteristic case group(0.68-1.44)was higher than the control group(0.50-1.08),and the difference was statistically significant(p<0.05).The quartile of Small Area Low Gray Level Emphasis in the case group(0.46,0.56,0.73)was lower than that in the control group(0.54,0.73,1.36),and the difference was statistically significant(p<0.05).The AUC of 10Percentile(0.712)to diagnosis T2DM is larger than that of small area low gray level emphasis(0.672).Conclusion:1.The local and mean PFF in T2DM group is higher than that in normal blood glucose group,which indicates that pancreatic fat deposition is closely related to T2DM and may be a potential risk factor for T2DM.Mean PFF is superior to local pancreatic PFF in diagnosing T2DM,which indicates uneven pancreatic fat deposition.2.Many texture features in T2DM group are significantly different from those in normal blood glucose group,which suggests that pancreatic texture analysis is of great value in the diagnosis of T2DM and may provide important reference for the development of accurate diagnosis of artificial intelligence images in the future. |
PDF Full Text Request