| Part 1Objective:To assess the feasibility of hepatic fat fraction（HFF）obtained by whole-liver 3D segmentation in patients with different degrees of nonalcoholic fatty liver disease（NAFLD）.Materials and methods:A total of 4358 inpatients who underwent upper abdominal 1.5 T or 3.0 T MRI in our hospital from January 2017 to December 2019（including IDEAL-IQ fat fraction map and image quality meeting the measurement requirements）were retrospectively collected.Exclusion criteria:（1）Missing clinical data（medical history,physical examination and laboratory data within one month of undergoing magnetic resonance imaging,etc.）（n=999）;（2）Age<18 years old（n=1）;（3）History of alcohol abuse（the ethanol intake of men≥210 g/week and that of women≥140 g/week in the past ten years）[Calculation formula:ethanol intake（g）=volume（ml）×alcohol degree（%）×0.8]（n=87）;（4）Liver diseases that cause NAFLD（viral hepatitis,autoimmune hepatitis,drug-induced liver damage,hepatolenticular degeneration,α-1 antitrypsin deficiency,etc.）（n=183）;（5）Other liver diseases:including liver cirrhosis,liver malignant tumor,single space-occupying lesion in the liver>3 cm in diameter or the number of space-occupying lesions>3,post-hepatectomy,decompensated liver disease,infection,intrahepatic bile duct disease,etc.（n=1656）;（6）Drugs that may cause hepatic fat deposition:glucocorticoids,synthetic estrogen,tamoxifen,amiodarone,sodium valproate,olanzapine,etc.（n=19）;（7）History of radiotherapy and chemotherapy（n=926）;（8）Secondary hepatic fat deposition accompanied by systemic diseases:total parenteral nutrition,celiac disease,inflammatory bowel disease,hypopituitarism,hypothyroidism,lipoatrophy,hypogonadism,etc.（n=37）;（9）Body weight change of more than 5%within one month（n=81）;（10）Pregnancy status（n=1）.Finally,368 patients（172 males and 196 females）were included in the study.On the MRI IDEAL-IQ fat fraction map,the liver was divided into liver segment I,II,III,IV,V,VI,VII,and VIII according to the Couinaud liver segmentation method,and liver segment IV was divided into IVa and IVb subsections.Among them,liver segment I composed the caudate lobe,and liver segments II,III,IVa and IVb composed the left lobe of the liver,and liver segments V,VI,VII and VIII composed the right lobe of the liver.During the measurement,T2WI images were used as a reference,and three circular ROIs were placed on each of the nine segments of the liver（a total of 27 ROIs）.The intrahepatic bile ducts,blood vessels,hepatic fissures,liver margins and focal liver lesions were avoided during placing,and the area of total ROIs was guaranteed to be not less than 5 cm2.The HFF of each ROI was recorded,and the mean fat content in the left lobe of the liver,the mean fat content in the right lobe of the liver,and the mean fat content in the whole liver were obtained by averaging.According to the measurement result,all subjects were divided into four groups:non-NAFLD group（HFF<5%）,mild NAFLD group（5%≤HFF≤14%）,moderate NAFLD group（14%<HFF≤28%）,and severe NAFLD group（HFF>28%）.In this study,non-NAFLD 250 patients（67.93%）,mild NAFLD 97 patients（26.36%）and moderate NAFLD 21 patients（5.71%）were enrolled.At the ISP workstation,the whole liver was segmented on the MRI IDEAL-IQ fat fraction map using 3D semi-automatic segmentation software,and the HFF was calculated automatically.The intraclass correlation（ICC）was used to check the two observers’consistency.The differences of fat content among liver segments were analyzed by ANOVA for normally distributed data,or Fridman test for abnormally distributed data.The difference of fat content between the left lobe and the right lobe of liver was analyzed by paired sample t-test for normally distributed data,and Wilcoxon test for abnormally distributed data.Pearson or Spearman test was used to analyzed the correlation analysis of the results measured by the two methods.Bland-Altman test was used to analyze the consistency of the results measured by the two methods.A two-tailed P<0.05 was considered statistically significant.Results:All ICC values were greater than 0.9,and interobserver and intraobserver agreement were good.Except for the moderate NAFLD group,the HFFs of different liver segments in the other groups had significant differences（P<0.001）,and the HFF in the right lobe of the liver was significantly higher than that in the left lobe of the liver（P<0.001）.HFF（%）measured by whole liver 3D segmentation and traditional ROI sampling method in the non-NAFLD group were 3.10（2.50,3.80）and 2.58（2.03,3.33）,respectively,with a strong correlation（r=0.936,P<0.001）.HFF（%）measured by whole liver 3D segmentation and traditional ROI sampling method in the mild NAFLD group were 8.20（6.50,10.40）and 7.84（6.27,10.15）,respectively,with a strong correlation（r=0.991,P<0.001）.HFF（%）measured by whole liver 3D segmentation and traditional ROI sampling method in the moderate NAFLD group were 17.00（14.30,22.30）and 18.03（15.15,22.37）,respectively,with a strong correlation（r=0.990,P<0.001）.The results of Bland-Altman analysis showed that in the non-NAFLD group,the HFF was slightly overestimated by whole liver 3D segmentation.The difference（%）of HFF between whole liver 3D segmentation and traditional ROI sampling method was+0.47.The 95%limits of agreement ranged from+1.03 to-0.10,and 96.00%（240/250）was within the 95%limits of agreement.In the mild NAFLD group,the HFF was slightly overestimated by the whole liver 3D segmentation method.The difference（%）of HFF between the whole liver 3D segmentation method and the traditional ROI sampling method was+0.65.The95%limits of agreement ranged from+0.67 to-0.54,and 95.88%（93/97）was within the95%limits of agreement.However,in the moderate NAFLD group,the HFF was mild underestimated by whole-liver 3D segmentation.The difference（%）of HFF between whole-liver 3D segmentation and traditional ROI sampling method was-0.98.The 95%limits of agreement ranged from+0.56 to-2.52,and 90.48%（19/21）within the 95%limits of agreement.Conclusion:The hepatic FF can be slightly overestimated or underestimated in patients with NAFLD of different degrees using whole liver 3D segmentation with intrahepatic vasculature on MRI fat fraction map,but the overall consistency is good.The whole liver3D segmentation method based on MRI fat fraction map to evaluate liver fat content does not need to eliminate intrahepatic vasculature,which greatly simplifies the measurement workflow,and improves the image post-processing efficiency.It provides a preliminary basis for the simplified process of full-automatic 3D segmentation of whole liver in the future,and it has good clinical application value.Part 2Objective: To assess the relationship between ectopic fat deposition and various cardiovascular metabolic risk factors（including hypertension,type 2 diabetes mellitus（T2DM）,high triglycerides（TG）and low high-density lipoprotein cholesterol（HDL-C）based on the MRI fat fraction map.Materials and methods: A total of 4825 inpatients who underwent upper abdominal 1.5 T or 3.0 T MRI in our hospital from January 2017 to August 2020（including IDEAL-IQ fat fraction map with image quality meeting the measurement requirements）were retrospectively collected.Exclusion criteria:（1）Missing clinical data（medical history,physical examination and laboratory data within one month of undergoing magnetic resonance imaging）（n = 1038）;（2）Age < 18 years old（n = 1）;（3）History of alcohol abuse（the ethanol intake of men ≥ 210 g/week and the ethanol intake of women ≥ 140 g/week in the past ten years）[Calculation formula: ethanol intake（g）= volume（ml）× alcohol degree（%）× 0.8]（n = 87）;（4）Liver diseases that may cause NAFLD,liver cirrhosis,liver malignant tumors,single space-occupying lesion in the liver > 3 cm in diameter or the number of space-occupying lesions > 3,post-hepatectomy,decompensated liver diseases,infections,intrahepatic bile duct diseases,drugs that may cause hepatic fat deposition,secondary hepatic fat deposition accompanied by systemic diseases,etc.（n = 2074）;（5）History of pancreatic and bile duct diseases,acute or chronic pancreatitis,autoimmune pancreatitis,pancreatic tumor,pancreatic surgery,pancreatic trauma,dilation of biliary and pancreatic ducts,drugs that may cause pancreatic fat deposition,etc.（n = 173）;（6）Ascites,mesenteric surgery,abdominal giant mass,abdominal wall edema,abdominal wall fistulization surgery,bilateral paraspinal muscle disease history,etc.（n = 27）;（7）History of radiotherapy and chemotherapy（n = 1017）;（8）Pregnancy status（n = 1）;（9）Secondary hypertension and other types of diabetes besides T2DM（n = 8）;（10）Body weight change of more than 5% within one month（n = 79）.Finally,320 patients（148 males and 172 females）were included in the study.According to the previous literature and China Cardiovascular Disease Report 2017,the patients were divided into hypertension group（131 cases,70 males and 61 females）/non-hypertension group（189 cases,78 males and 111 females）,T2 DM group（66 cases,38 males and 28 females）/nonT2 DM group（254 cases,110 males and 144 females）,high TG group（73 cases,33 males and 40 females）/non-high TG group（247 cases,115 males and 132 females）,low HDLC group /（105 cases,64 males and 41 females）/non-low HDL-C group（215 cases,84 males and 131 females）.MRI fat fraction maps were used to assess different fat depots with double-blind method by two observers: including subcutaneous adipose tissue（SAT）area,SAT FF,visceral adipose tissue（VAT）area,VAT FF,preperitoneal adipose tissue（p PAT）area,p PAT FF,HFF,pancreatic fat fraction（PFF）,and skeletal muscle adipose tissue FF（SMAT FF）.The ICC was used to analyze interobserver and intraobserver consistency.Partial correlation analysis was used to analyze the correlation between the parameters of ectopic fat deposition in different parts.The differences of age,BMI,ectopic fat deposition parameters in different parts and blood biochemical indicators between the hypertension group/non-hypertension group,T2 DM group/non-T2 DM group,high TG group/non-high TG group,and low HDL-C blood group/non-low HDL-C blood group were analyzed using independent sample t test or Mann-Whitney U analysis.The differences of gender,menopausal status,family history of cardiovascular metabolic risk factors,smoking,drinking,and treatment of cardiovascular metabolic risk factors and other classified data between groups were analyzed using chi-square test.Multivariate Logistic regression analysis was used to calculate the odds ratio（OR）and 95% confidence interval（CI）of different ectopic fat deposition parameters to the risk of hypertension,T2 DM,high TG and low HDL-C.Two models were established in sequence（hereinafter referred to as Model 1 and Model 2,and Logistic regression was used for both models）.First,after correction by Model 1 for clinical data with statistical differences between groups,the ectopic fat deposition parameters independent of the clinical data were obtained.Second,Model 2 was used to further screen ectopic fat deposition that had an independent effect on hypertension,T2 DM,high TG,and low HDL-C.The receiver operating characteristic（ROC）curve was used to analyze the predictive values of ectopic fat deposition parameters independently and their combinations on hypertension,T2 DM,high TG and low HDL-C,and the area under the curve（AUC）was calculated,and the sensitivity and specificity were determined according to the Youden index.The Delong test was used to compare AUC values of ectopic fat deposition in different parts and evaluate whether the predictive abilities of each parameter for hypertension,T2 DM,high TG and low HDL-C were different.In addition,subgroup analyses were performed for gender stratification to explore gender differences in the predictive value of ectopic fat deposition in hypertension,T2 DM,high TG,and low HDL-C.A two-tailed P < 0.05 was considered statistically significant.Results: All ICC values were greater than 0.9,and interobserver and intraobserver agreement were good.The correlations among SAT FF,SAT area,VAT area,HFF,PFF and p PAT FF were significant（P < 0.05）,but the patterns of correlations were different.VAT area was an independent predictor of hypertension,T2 DM and low HDL-C,and AUC values were 0.769,0.725 and 0.720,respectively;however,SAT area was not associated with any cardiovascular metabolic risk factors.SAT FF is an independent predictor of hypertension,with the AUC value of 0.766;and after the combination of SAT FF and VAT area,the prediction efficiency was improved（AUC = 0.779,P > 0.05）.HFF was an independent predictor of high TG,with the AUC value of 0.726.PFF was an independent predictor of T2 DM,high TG and low HDL-C,and AUC values were 0.764,0.721 and 0.722,respectively.The predictive efficiency after the combination of PFF and VAT area for T2 DM was significantly higher than that of VAT area（AUC value: 0.769 vs.0.725,P = 0.006）.The predictive efficiency after the combination of PFF and HFF for high TG was significantly increased（AUC = 0.747,P > 0.05）.p PAT FF was an independent predictor of low HDL-C,with the AUC value of 0.706;and the predictive efficiency after the combination of p PAT FF,VAT area and PFF for low HDL-C was significantly higher than that of p PAT FF（AUC value: 0.740 vs.0.706,P = 0.018）.The results of gender stratified analysis showed that VAT area was an independent predictor of hypertension,T2 DM and low HDL-C in men,with the AUC values of 0.744,0.709 and 0.720,respectively;however,VAT FF was an independent predictor of hypertension,high TG and low HDL-C in women,with the AUC values of 0.798,0.734 and 0.692,respectively.In addition,p PAT area was an independent predictor of hypertension in men,with the AUC value of 0.745.Conclusion: VAT area,HFF,and PFF were significantly associated with various cardiovascular metabolic risk factors,while SAT area was not associated with each cardiovascular metabolic risk factor.However,these patterns of correlations varied by gender and risk factors.In addition,VAT has gender differences in the prediction of cardiovascular metabolic risk factors.VAT FF was only significantly associated with different cardiovascular metabolic risk factors in women.p PAT area was only significantly associated with various cardiovascular metabolic risk factors in men.These findings broaden the understanding of the association between ectopic fat deposition and cardiovascular metabolic risk factors,and further clarify the heterogeneity of obesity.Part 3Objective: To assess the relationship between the density of perivascular adipose tissue（PVAT）and descending thoracic aortic plaques based on CTA.Materials and methods: A total of 102 inpatients who underwent thoracic aorta CTA examination by Discovery CT 750 HD in our hospital from July 2015 to March 2021 and did not have arterial dissection,intramural hematoma,aneurysm,trauma and postoperative changes of thoracic aorta were retrospectively collected.Exclusion criteria:（1）Patients with esophageal disease,hiatal hernia,thoracic stomach and thoracic spine surgery（n = 11）;（2）Unconventional CTA scanning protocol（no 1.25 mm or 0.625 mm reconstruction,non-100 k Vp tube voltage,etc.）（n = 17）;（3）Patients with thoracic descending aortic malformation（n = 1）.Finally,73 patients were included in the study.The PVAT was segmented using ITK-SNAP and MATLAB software by threshold（threshold ranged from-195 HU to-45 HU）and the PVAT density（HU）was calculated.Assessment of atherosclerosis in the descending thoracic aorta was performed on the axial enhanced images.The descending thoracic aorta was anatomically divided based on the isthmus,the posterior intercostal arteries,and the diaphragm.The plaque with the largest thickness was selected in each segment,and the maximum wall thickness was measured perpendicular to the center of the aorta.Classification of descending thoracic aorta atherosclerosis was performed according to the maximum thickness of aortic plaque,ulceration,and mural thrombus: irregular thickening of the aortic wall ≥ 2mm was defined as aortic plaque;plaque thickness ≥ 4 mm or ulcer or mural thrombus was defined as aortic complex plaque;maximum defect depth on the plaque surface ≥ 2 mm was defined as ulcer plaque.All plaques were classified into four grades: 1）no plaque;2)plaque with thickness between 2 mm and 4 mm;3)plaque thicker than 4 mm;4)ulcer plaque.Grade 1 and 2 are divided into low-risk plaque group,while grade 3 and 4 are divided into highrisk plaque group,and this study included 31 patients in the low-risk plaque group and 42 patients in the high-risk plaque group.In addition,the mean plaque burden score（MPBS）was defined by combining the thickness and circumference of plaques in each segment of the descending thoracic aorta.Plaque thickness（PT）scoring criteria were defined as: 0 = no plaque,1 = mild（PT < 3 mm）,2 = moderate（3 mm ≤ PT ≤ 5 mm）,and 3 = severe（PT > 5 mm）.The Plaque circumference（PC）scoring criteria were defined as class: 0 = no plaque,1 = mild（PC < 1/3 of the vessel circumference）,2 = moderate（1/3 ≤ PC ≤ 2/3 of the vessel circumference）,and 3 = severe（PC > 2/3 of the vessel circumference）.MPBS was calculated as follows: MPBS =（（PT1 + PT2 +…+ PTn）+（PC1 + PC2 +…+ PCn））/n（n refers to the number of segments）.Independent sample t-test or Mann-Whitney U analysis was used to analyze the differences of measurement data between two groups.The difference between two groups was analyzed with Chi-square test for classified data.Correlations between PVAT density and MPBS were analyzed using Pearson or Spearman’s test.The confounders were corrected by linear regression analysis to calculate the β coefficient and 95%CI of PVAT density and MPBS.Multivariate Logistic regression was used to correct the confounders and the OR value and 95%CI of PVAT density and high-risk plaque were calculated.The predictive value of PVAT density for high-risk plaque was evaluated,and the AUC value was calculated using ROC curve analysis.The sensitivity and specificity were determined according to the Youden index.A two-tailed P < 0.05 was considered statistically significant.Results: In this study,73 patients（56 males and 17 females）were finally included,with the median age of 69 years old,average BMI of 24.50 ± 3.55 kg/m2,average PVAT density of-91.57 ± 4.65 HU,and median MPBS of 2.11.For clinical data,the age,hypertension and smoking status in the high-risk plaque group were significantly higher than those in the low-risk plaque group（P < 0.05）.For imaging parameters,the PVAT density and MBPS in the high-risk plaque group were significantly higher than those in the low-risk plaque group（P < 0.05）.All ICC values were greater than 0.9,and interobserver and intraobserver agreement were good.There was a significant correlation between PVAT density and MPBS（r = 0.377,P = 0.001）.After adjustment for age,hypertension,and smoking status,linear regression analysis showed a significant correlation between PVAT density and MPBS with a β coefficient of 0.073,a 95%CI of 0.002-0.114（P = 0.044）.After adjustment for age,hypertension,and smoking status,logistic regression analysis showed a significant correlation between PVAT density and high-risk plaque,with an OR value of 1.263,a 95%CI of 1.093-1.459（P = 0.002）.Analysis of ROC showed that PVAT density was an independent predictor of high-risk plaque with AUC value of 0.814,95%CI of 0.717-0.912,specificity of 0.645,and sensitivity of 0.881（P < 0.001）.Conclusion: The increased PVAT density was associated with high-risk plaque and increased plaque burden of descending thoracic aorta atherosclerosis,and PVAT density can be used as an imaging marker of the severity of descending thoracic aorta atherosclerosis. |
PDF Full Text Request