Rapid Quantification Of Liver Fat Content In Nonalcoholic Fatty Liver Disease Based On MDixon Method

Part one:Comparison of Breath-Hold and Respiratory-Triggered Proton MR Spectroscopy in Quantification of Liver Fat ContentObjective1. To examine the reproducibility of multiple breath-hold (BH) 1H-MRS and respiratory-triggered (RT)1H-MRS on the quantification of liver fat content(LFC) in human subjects.2. To explore the feasibility of multiple BH 1H-MRS and RT1H-MRS on the quantification of LFC in human subjects.3. To evaluate the consistency and correlation of multiple BH1H-MRS and RT 1H-MRS for quantification of LFC.Materials and Methods1. Study PopulationA total of 60 subjects, were enrolled in this study, including 30 nonalcoholic fatty liver disease (NAFLD) patients and 30 patients with no obvious liver disease as a healthy control group. All the patients underwent upper abdominal MR imaging in Guangdong No.2 Provincial People’s Hospital from February 2013 to December 2013.The diagnosis of NAFLD were consistency with diagnostic criteria of 《Guidelines for management of nonalcoholic fatty liver disease:an updated and revised edition (2010))) by The Chinese National Workshop on Fatty Liver and Alcoholic Liver Disease for the Chinese Liver Disease Association. Inclusion criteria of NAFLD were as follows:(1) Patient aged over 16 years. (2) Evidence of absent or minimal alcohol consumption (i.e.,<70 g of alcohol per week for women and<140 g of alcohol per week for men). (3) Absence of confounding disease that can lead to fatty liver disease, including acute and/or chronic viral hepatitis, drug-induced liver disease, total parenteral nutrition, Wilson’s disease, autoimmune liver disease. (4) Diagnosis of fatty liver were made with ultrasonography (US) or computed tomography(CT) at this time. Inclusion criteria of healthy control group were as follows:(1) No history of any liver diseases. (2) Normal liver echogenicity (on US) or liver attenuation(on CT) were seen at this time, with no obvious liver disease. Exclusion criteria were as follows:(1) Patients with contraindications to MR imaging (i.e. patients with claustrophobia, pacemaker or metallic implant) and pregnancy was also an exclusion criterion. (2)Patients with diseases that affect iron deposition in liver, such as patients with hemochromatosis, thalassemia or patients with a history of blood transfusion. (3) Patients with uninterpretable spectrum due to respiratory artifacts were excluded.2. Instruments and EquipmentA 1.5T MR scanner (Achieva, Nova-Dual Philips Medical Systems, the Netherlands) and 16-channel torso phase-array coil was used.3. Acquisition Method of 1H-MRSAll patients underwent MR examination in the morning and were instructed to fast for 8 hours before the MR examination.1H-MRS as one part of the upper abdominal MR examination was performed by a same trained technician of MRI. Patients were examined in the supine position and were instructed to cooperate their breath. They were told to breathe calmly and regularly, and try best to keep the same amplitude. They were instructed to hold breath during a breath-hold sequence in the end of expiratory in order to maintain consistency.1H-MRS were performed after all the routine sequences of MR. The routine sequences of MRI included as follows:three-plane localizing images, mDixon sequence, T2WI, SPAIR, DWI. In order to make a accurate localizing, reconstructed images of mDixon were used as localizing images.Single voxel point-resolved spectroscopy sequence (PRESS) was used. A 20mm×20mm×20mm volume of interest (VOI) was placed in a right hepatic segment (Couinaud segment VII) by a same trained radiologist. VOI should be 10 mm far away from the liver edge, avoiding vascular and biliary structures in the liver and the adjacent tissues, organs and fat. As VOI was placed, automatically optimized gradient shimming performed. As full width half max of water<10Hz, data acquisition began. Water suppression was not performed.RT 1H-MRS was performed before BH 1H-MRS. We used the Screen Capture Function to copy the localizing image with VOI placed and sent it to the post-processing workstation, so as to record the exact position of the VOI, and VOI of the subsequent BH 1H-MRS can be placed in the same position of the liver. Acquisition parameters are as follows:TR:2000ms, TE:41ms, NSA:32。Multiple BH1H-MRS was performed subsequently. VOI was placed at the same position of that in RT. Patients were instructed to hold breath in the end of expiratory. The total acquisition time was 80s, and was divided into four sections of BH according to patient’s capacity of BH. Each section of BH was 20s, and after one section of BH, patients were asked to breathe freely for 8-10s and a next BH sections was followed. Acquisition parameters are as follows:TR:2000ms, TE:41ms, NSA: 80In order to evaluate the reproducibility of RT 1-MRS and BH1H-MRS, we repeated the acquisition of RT 1H-MRS and BH1H-MRS, with the same size and position of VOI, and the same acquisition parameters.4.1H-MRS Data Analysis1H-MRS data were sent to the post-processing workstation of Philips and were analyzed by software of Spectro View. Spectrum was fitted. The peak areas of water (4.7 ppm) and methylene (1.3 ppm) obtained from RT 1H-MRS and BH 1H-MRS were recorded respectively and LFC was calculated as:LFCMRS=(Alipid/ Alipid+Awater)×100%. The average of the two acquisitions results was used as the final result of the same acquisition method.5. Statistical AnalysisData were analyzed by using SPSS 17.0 software packages. All results were expressed as mean ± standard deviation (SD). The reproducibility of RT’H-MRS and BH 1H-MRS were assessed by paired-sample T test. LFCMRS-RT calculated with RT 1H-MRS and LFCMRS-BH calculated with multiple BH 1H-MRS were analyzed using linear regression analysis. Pearson correlation coefficients, Bland-Altman 95% limit of agreement, and Lin concordance correlation coefficients were calculated. P<0.05 was considered to indicate significance.ResultsOf the 60 patients included in our study,40 were men, and 20 were women. The overall mean age was 42.7years±14.1 (standard deviation, SD) (range,16-77 years). The overall mean body mass index (BMI) was 25.96 kg/m2±4.54 (range,17.93-38.06 kg/m2)。All patients were able to cooperate 1H-MRS examination well and all the data obtained met the requirements for analysis. There was no significant difference between the two acquisition of RT’H-MRS and BH1H-MRS (both P<0.0001), indicating a good reproducibility.The mean LFCMRS-RT calculated with RT 1H-MRS was (8.6±8.7)% (range, 0.2%-38.5%) and the mean LFCMRS-BH calculated with multiple BH 1H-MRS was (9.4±9.3)%(range,0.1%-35.4%). There was no significant difference between LFCMRS-RT and LFCMRS-BH when assessed by using paired-sample T test (P< 0.0001). LFCMRS-RT and LFCMRS-BH was significant correlated, and the Pearson correlation coefficient was 0.973 (P<0.0001). And with Bland-Altman method, they appeared highly correlated with 91.7% points within the 95% limit of agreement. Finally, the Lin concordance coefficient was 0.951(P<0.0001).Conclusion1. The single voxel PRESS RT 1H-MRS and BH 1H-MRS demonstrate a good reproducibility in human liver.2. Both RT’H-MRS and multiple BH 1H-MRS can be used to quantify LFC in patients with NAFLD and healthy subjects as normal control. These two methods have a strong correlation and perfect consistency. The results of LFC calculated with these two methods have no significant difference.3. RT1 H-MRS and multiple BH’H-MRS are alternative tools for intrahepatic lipid quantification. These two methods could be chosen to perform with flexibility according to patients’ cooperation and operators’ skills in clinical application.Part two:Rapid Quantification of Liver Fat Content in Nonalcoholic Fatty Liver Disease based on mDixon methodObjective1. To assess to the feasibility of liver fat content(LFC) quantification by mDixon method in patients with nonalcoholic fatty liver disease (NAFLD) with images of liver obtained by mDixon method post-processed by software Image J (National Institutes of Health, USA).2. To evaluate the accuracy of LFC quantification by mDixon method in patients with NAFLD, with single voxel PRESS respiratory-triggered (RT) hydrogen (1H) MR spectroscopy as reference standard.Materials and Methods1. Study PopulationA total of 70 patients with NAFLD were included in this study. The 70 patients underwent 1.5T MR scanner in Guangdong No.2 Provincial People’s Hospital from February 2013 to December 2014. The diagnosis of NAFLD were consistency with diagnostic criteria of 《Guidelines for management of nonalcoholic fatty liver disease: an updated and revised edition (2010)》 by The Chinese National Workshop on Fatty Liver and Alcoholic Liver Disease for the Chinese Liver Disease Association. Inclusion criteria of NAFLD were as follows:(1) Patient aged over 16 years. (2) Evidence of absent or minimal alcohol consumption (i.e.,<70 g of alcohol per week for women and<140 g of alcohol per week for men). (3) Absence of confounding disease that can lead to fatty liver disease, including acute and/or chronic viral hepatitis, drug-induced liver disease, total parenteral nutrition, Wilson’s disease, autoimmune liver disease. (4) Diagnosis of fatty liver were made with ultrasonography (US) or computed tomography(CT) at this time. Exclusion criteria were as follows:(1) Patients with contraindications to MR imaging (i.e., patients with claustrophobia, pacemaker or metallic implant) and pregnancy was also an exclusion criterion. (2) Patients with diseases that affected iron deposition in liver, such as patients with hemochromatosis, thalassemia or patients with a history of blood transfusion. (3) Patients with uninterpretation spectrum due to respiratory artifacts were excluded.2. Instruments and EquipmentsA 1.5 T MR scanner (Achieva, Nova-Dual Philips Medical Systems, the Netherlands) and 16-channel torso phase-array coil were used.3. MR examinationsAll patients underwent MR examination in the morning and were instructed to fast for 8 hours before MR examination. All the sequences of MR was performed by a same trained technician. Patients were examined in the supine position. Patients were instructed to cooperate their breath by the same trained technician. They were told to breathe calmly and regularly, and try best to keep the same amplitude. They were instructed to hold breath during a breath-hold sequence in the end of expiration in order to maintain consistency.Three dimensional(3D) spoiled gradient echo mDixon sequence and 1H-MRS were performed in the same MR examination. The volume scanning of the whole liver mDixon sequence were performed before 1H-MRS obtained. By using three-plane localizing images, the scan range of mDixon sequence was made, including the whole liver. Acquisition parameters were as follows:TR:5.6ms, TE1: 1.8ms, TE2:4.0ms, flip angle:15°, matrix:256×160-256, rectangular FOV:adjusted according to patient’s size, section thickness:6mm, intersection gap:-3mm. mDixon sequence acquisition was obtained within a single breath-hold (with breath-hold time from 13.8s to 17.7s). Four different group images were created by mDixon method: water-only images(W), in phase images(IP), out phase images(OP) and fat-only images(F).1H-MRS acquisition:In order to make a accurate localization, reconstructed water-only images of mDixon were used as localizing images.Single voxel respiratory-triggered(RT) point-resolved (PRESS)1H-MRS was used. Acquisition parameters were as follows:TR:2000 ms, TE:41ms, NSA:32, A 20mmx20mmx20mm volume of interest(VOI) was placed in a right hepatic segment (Couinaud segment Ⅶ) by a same trained radiologist. VOI should be 10 mm away from the liver edge, avoiding vascular and biliary structures in the liver and the adjacent tissues, organs and fat. As VOI was placed correctly, we used the Screen Capture Function to copy the localizing image with VOI placed and sent it to the post-processing workstation, so as to record the exact position of the VOI, and subsequently matched the VOI in images obtained by mDixon. Automatically optimized gradient shimming performed after VOI placed correctly. As full width half max of water<10Hz, spectrum data acquisition began. Water suppression was not performed.4. MR images processing(1) We used free image-processing software Image J (National Institutes of Health, USA) to process images obtained by mDixon methods and measured the LFC. The original water-only images (W) and fat-only images (F) were sent to Image J respectively, and "Image-to-stack" was used to turn multiple images into a "stack" image, and the "stack" images named as "W" or "F" respectively were saved. "Image calculator" was used to add "W" and "F" stack image and a stack image of "W+F" was created, and then "Image calculator" was used to divide "F" stack image by "W+F" stack image and a stack image of fat fraction (FF) (FF=F/W+F) was created. Then we can use "area select tool" in the toolbar to place in any area of liver, click the "measure" in the toolbar, and LFC of this area would be calculated by the software. In order to compare the results to that of the 1H-MRS, a rectangular VOI of 20mm×20mm (matching the VOI of’H-MRS) was placed in the same position of VOI in 1H-MRS using the copy localizing image sent to workstation as reference. LFC was measured on three continuous sections and the average of three results was used as the final result of LFCmDixon·(2)1H-MRS Data Analysis:1H-MRS Data were sent to the post-processing workstation of Philips and were analyzed by software of Spectro View. Spectrum was fitted. The peak areas of water (4.7 ppm) and methylene (1.3 ppm) obtained from RT 1 H-MRS were recorded respectively and LFCMRS was calculated as:LFCMRS=(Alipid/ Aiipid+Awater)×100%.5. Statistical AnalysisData were analyzed by using SPSS 17.0 software packages. All results are expressed as mean±standard deviation (SD). Agreement between LFCmDixon and LFCMRS was assessed in three steps. First, we drew a scatter-plot and performed linear regression to assess the relationship between LFCmDixon and LFCMRS. The Pearson correlation coefficient was also calculated. Second, we estimated agreement by Bland-Altman method, in which the difference between LFCmDixon and LFCMRS was plotted against their means, with the 95% confidence intervals. Third, Lin concordance correlation coefficients were calculated. P<0.05 was considered to indicate significance.ResultsOf the 70 patients with NAFLD included in our study,40 were men, and 30 were women. The overall mean age was 42.7years±15.4 (standard deviation, SD) (range,18-81 years). The overall mean body mass index (BMI) was 28.73 kg/m2±4.74 (range,19.10-55.36kg/m2)。The total acquisition time of 1H-MRS was three minutes and twenty-five seconds as the acquisition time of mDixon was 15.9 seconds. On the images of LFC(colored) created by Image J from mDixon images, tissues appears different color(dark blue to red) depending on the fat content of the tissues. Liver appears different blue to yellow-green as the LFC increasing.The mean LFCmDixon measured from FF images by Image J was (16.9±8.7)% (range,6.6%-37.0%). And the mean LFCMRS was (16.0±8.8)%, (range, 5.7%-41.4%). LFC in all patients was greater than 5.56% and all patients in our group met the diagnosis criteria of fatty liver in 1H-MRS.LFCmDixon and LFCMRS was significant correlated, and the Pearson correlation coefficient was 0.920(P<0.0001). And with Bland-Altman method,95.8% points are within the 95% limit of agreement. Finally, the Lin concordance coefficient was 0.968(P<0.0001).Conclusion1. mDixon images processed by Image J software could quantify LFC in a rapid and easy way without any manual calculation.2. LFCmDixon measured with Image J from mDixon images have excellent correlation and concordance with LFCMRS.3. Since mDixon method is widely available and its acquisition time is extremely short, and Image J software is easy to operate, this method can replace 1H-MRS in clinical study to quantify LFC rapidly.4. Further research should be carried to explore whether mDixon method is suitable for patients with cirrhosis and iron deposition in liver.