Correlation Study Of Gating Technology In Overcoming The Effects Of Breathing And Cardiac Motions On Hepatic Diffusion-weighted Magnetic Resonance Imaging

[Background]Diffusion-weighted magnetic resonance imaging (DW-MRI) is a noninvasive imaging method that uses MRI to observe the diffusion movements of water molecules in living tissue. Currently, DW-MRI has been used for diagnosing and evaluating neurological diseases. Through recent advances in MRI hardware, software, and technologies, the clinical applications of DW-MRI have been expanded. Now, it is increasingly being used in various body systems, including the abdominal system. Many scholars believe that DW-MRI is valuable in detecting small liver lesions, differentially diagnosing liver lesions, evaluating tumor efficacies, and qualitatively characterizing various organs. Therefore, the image quality of hepatic DW-MRI is critical to the diagnosis and treatment of liver diseases.Owing to technical reasons, the use of DW-MRI in the treatment of hepatic disorders is still a hot issue. This is primarily because DW-MRI is highly sensitive to physiological movements, such as breathing, the beating of the heart, and abdominal peristalsis, which can all affect image quality. Compounding the problem is that the imaging time of conventional spin-echo-DW-MRI is long, thus the clinical applications of DW-MRI are limited. The emergence of ultra-fast imaging technologies, such as echo-planar imaging (EPI) and parallel imaging, offer new capabilities for hepatic DW-MRI by reducing and restraining the motion artifacts and thereby improving the image quality. Thus, the application of DW-MRI to the liver is now possible. The single-shot spin-echo echo planar imaging (SS-SE-EPI) is the most commonly used hepatic DW-MRI method, but it is still impacted potentially by heartbeat.Though hepatic DW-MRI has improved continuously in recent years, three are still a lot of problems, such as artifacts, low signal-to-noise ratio, and high apparent diffusion coefficient (ADC) errors have contributed, but the primary limitation remains physiological movement. Currently, attention has been focused on breathing. There are four main respiratory compensation techniques:Free-breath (FB), Respiratory-triggered (RT), Breathhold (BH), and Navigator-triggered (NT). Each method has its advantages and disadvantages, and some have helped in solving the breathing-induced image motion artifacts, but none has provided consistent study results.The problems caused by cardiac motion are less of a concern. The hepatic diffusion imaging usually needed to collect a series data with different b values. If not controlled, the data would usually be collected at different phases of the cardiac cycle, while there was no stable trigger time (phase). Thus, each DW-MRI image would be different because of the impact of cardiac motion. Cardiac motion, in turn, affects the b value, which is a measure of the duration and strength of the diffusion gradient; the greater the b value, the greater the weighted diffusion. The single exponential model calculation of the ADC value requires at least two different DW-MRI data with different b values, and can be obtained according to the equation: ADC=ln(S2/S1)/(b1-b2), in which b1 and b2 are different b values, and Sb2 and Sbl are their corresponding DW-MRI images. Therefore, data from different cardiac phases, which causes changes of voxel location, would affect the calculation accuracy of ADC. Thus, most scholars believe that cardiac motion may be affects the accuracy of calculation, and can cause hepatic DW-MRI signals to be missed, especially towards the left liver lobe. Therefore, one of the major technical difficulties of hepatic DW-MRI was how to overcome the effects of cardiac motion, thus avoiding signal loss.In this study, multiple acquisition techniques (RT, FB, BH, NT, and electrocardiogram-triggered (ECGT)) were combined during the acquisition of hepatic diffusion imaging. Our goal was to identify the program that could minimize the impacts of breathing and heart beating on DW-MRI. Meanwhile, the effectiveness and reproducibility of different acquisition programs were assessed to determine the optimal phase and ECGT time, so that the optimal hepatic DW-MRI scanning program could be identified to achieve the best image quality and the most accurate and stable ADC measurement. Together, we hope to provide reliable and objective data for the clinical diagnosis and treatment assessment of liver diseases.1. Study objectivesWe had four primary objectives. First, we explored the combination of various DW-MRI acquisition techniques, including respiratory compensation techniques (FB, RT, NT, and BH) and ECG compensation technique (ECGT), on the quality and reproducibility of ADC measurements in normal liver parenchyma.Second, we identified the effects of ECGT at different phase trigger times on the quality and repeatability of ADC measurements on normal liver parenchyma.Third, we assessed the effects of different hepatic anatomical positions (anterior; median and posterior; superior; median and inferior) on the quality and reproducibility of ADC measurements on normal liver parenchyma.Fourth, we explored the optimal DW-MRI scanning program that would provide improved image quality and clinical value of hepatic DW-MRI for further clinical use.2. Materials and Methods2.1. Study subjectsThis study was approved by the Ethics Committee of Panyu Center Hospital, Guangzhou. A total of 36 young healthy volunteers were selected from January 2013 to June 2014. Each participant underwent hepatic DW-MRI twice, and collection methods and techniques (FB, RT, NT, BH, and ECGT) were applied during the scanning process. All volunteers signed the informed consent before scanning.All enrolled patients were 20-40 years old with normal liver function (with negative viral serological markers), no history of liver disease (including mild fatty liver), and no prior liver surgery.Patients were excluded if they had any of the aforementioned liver involvements, heart disease (or abnormal ECG or cardiac function), took medications within 6 months of study enrollment, or had a history of alcohol abuse. Patients were excluded if they were claustrophobic or had any metal implanted. Patients were required to hold their breath for a minimum of 20 seconds. Data were excluded if the image quality was of poor quality. 2.2 Magnetic resonance scanning sequenceAll volunteers received the complete hepatic MRI scanning twice. Heart rate, pulse, and respiratory rate were measured before the MRI. The MRI special pads were pasted to the left sternal 2-5 intercostal space and the apical impulse site, and conventional breathing exercises were performed before scanning. Then, the volunteer was positioned in the supine position, ECGT and body coils were placed, and the scanning was performed from head-to-toe.The conventional sequences included the positioning phase coronal T2WI, axial T1 WI, and 15 coronal DW-MRI sequences, FB, FB with ECGT, BH, BH with ECGT, NT (100 mm rectangular excitation pulse was placed at the highest point of the right diaphragm to detect the position of right), and NT with ECGT (a total of six scanning programs). When ECGT was used, the trigger delay times were selected as 0 ms or 25 ms,200 ms,400 ms, and 600 ms. Due to the physical constraints of the MRI machine, RT and ECGT could not be used simultaneously, so there was no program of RT with ECGT. During the experiments, the RT-acquired images were analyzed, and no statistically significant difference was found between the RT-and NT-collected data within the right and left liver lobes (left liver:P= 0.212; right liver:P= 0.677). The minimal trigger time was 25 ms when NT and ECGT were applied simultaneously. The DW-MRI scanning mode used single-shot echo planar imaging (SS-EPI), and the b values used were 0 and 500 s/mm2.The volunteers received another hepatic MRI scan with the same parameters within 48 h of previous full scan. We tried to position the patient for the second scan as close as possible to the first scan position.After scanning, the images were transmitted to the workstation for processing and analysis.2.3 Image AnalysisThe original DW-MRI images were analyzed, and the ADC of region of interest (ROI), located on superior, median and inferior, was measured. The data were statistically analyzed by limit of agreement..2.31 Generation and accuracy assessment of the ADC imageThe average ADC image was determined by linear regression analysis with a single exponential linear function model, In (SIb/SIO)= b ×-ADC (SIb and SIO were the signal strengths when b was 500 and 0 s/mm2, respectively), the least squares method was used to fit one optimal line towards its minimal residual sum of squares. 2.32 Measurement of ADCThe ADC value was determined by a doctor with more than 5 years of experience in diagnosing hepatic DW-MRI. Two methods were used. Method two was performed 10 days after all data from method one were analyzed. The first method we used Image J software (National Institutes of Health, Bethesda, MD) to measure the three median levels of each sequence. The second method:first, the ADC image data (dicom format) in the DW-MRI sequence of each volunteer were converted to nii format by the MRIcroN data conversion tool. Then, we used the spm8 registration tool of Matlab 8.1 to register the 15 ADC data for each volunteer. We note that the data correction might exhibit split-level and non-matching of data because of positioning and motion.The data registration was divided into two steps. First, we used a rigid registration (realign tools), with the 15 ADC data-calculated average data as the reference image, to re-register the 15 ADC data and obtain the average data template. Seconds, with the template as the reference image, we used the affine transformation (normalize tool) to finely register the original 15 non-registered ADC values. Finally, one superimposed registration image of 15 sequences was obtained. The MRIcroN tool was then used to draw the ROI of three median levels of the registered template, and the acquisition method was the same as method one. Matlab 8.1 was used to export all ROI measurement data, which were assessed for consistency of the two methods.2.33 Placement of region of interestAmong the five-layer ADC images acquired by each technique, three median images were selected and divided into anterior, median, and posterior levels. The coronal planes of the left and right liver were divided into superior, median, and inferior levels, with one ROI placed in each level. The areas were kept constant (40 mm2). When setting ROI, we tried to avoid the eye-visible blood vessels and bile ducts and to keep the ROI locations of the two methods consistent by measuring at least 5 mm away from the liver edge.The left and right liver lobe of each volunteer was fully scanned twice, and each level of each liver lobe had three ROI yielding a total of 3-layer lobes (anterior, median and posterior),15 sequences, and 540 ADC values.2.4 Statistical analysis2.4.1 Expression method of ADC valueThe ADC values were expressed as mean ± standard deviation, and the Levene test was used to test for the homogeneity of variance. The Kolmogorov-Smirnov test was used to assess normality.2.4.2 Consistency of measurements by the two methodsThe consistency evaluation of these two methods used the intra-and interclass correlation coefficients (ICCs). The left and right liver lobe ICCs were calculated, respectively, as the mean ADCs of the superior, median, and inferior ROI of the anterior, median, and posterior levels. The ICCs of the left and right liver lobes measured by the two methods were calculated, and when the value exceeded 0.75, it was considered as having good consistency.2.4.3 Comparison of the repeatability of the ADC values based on method and anatomical locationThe Bland-Altman method was used to assess the reproducibility of the ADC values. The average absolute differences and 95% limits of agreement (LOAs) of these average differences obtained by the two scanning procedures were compared. The smaller the average absolute differences and 95% LOAs, the better the repeatability.2.4.4 Comparison of ADC valuesAn analysis of variance was used to compare the average ADCs of 9 ROIs of the left and right hepatic lobes, with P< 0.05 considered as a significant statistical difference.A two-way analysis of variance was used on the data obtained with the same delay triggering time on the different respiratory compensation techniques, with P< 0.05 considered as a significant statistical difference. The Bonferroni method was used for pairwise comparisons when there were differences. The same analysis was used for the different-delay triggering time experiments.The ADC values of the 9 ROI from each lobe compared with a random-grouped factor analysis, with P< 0.05 considered as a significant statistical difference.2.4.5 Statistical softwareSPSS 13.0 statistical software was used for the statistical analyses.3. Results3.1 Population demographicsThe MRI images of a total of 36 healthy volunteers were included in this study. Five of them did not finish the tests because of bad image quality due to their bulk movement and no breathold well (n=2) or incomplete acquisition of DWI sequences due to long scanning time (n=3).31 volunteers successfully finished the scans (15 men and 16 women). The average age was 23.8±2.6years old and was similar between men and women (men:mean age of 23.9±3.5 years, range:21-36 years; women:mean age 23.7±1.6 years, range:21-28 years). The average heart rate before the first check was 71.5±7.5 bpm, (range:55-82 bpm). The average breathing rate before the first check was 17.9±2.4 breaths/min (range14-22 breaths/min). The average heart rate before the second check was 70.6±6.4 bpm (range:60-80 bpm). The average breathing rate before the second check was 17.8±2.4 breaths/min (range: 13-22 breaths/min).3.2Consistency of the two measurement methodsThe analysis of two data revealed the consistency of the ICC range:the left hepatic lobe was 0.776-0.966, the lowest ICC came from NT with ECG delay time at 25 ms (0.776); the highest ICC came from NT with ECG delay time at 400 ms (0.966). The right lobe was 0.870-0.967, the lowest ICC came from BH (0.870); the highest ICC came from FB with ECG delay time at 200 ms (0.967). In general, the variance of the ICC range of right lobe measurements was wider than that of left lobe.3.3 Repeatability of ADC values of the right and left hepatic lobes using three techniques(FB、BH、NT) with different ECGT delay timesThe repeatability of ADC measurement of the left hepatic lobe was worse than the right lobe. The range of ADC values of the left hepatic lobe after NO ECG scanning was:FB (-0.36-0.47) × 10-3 mm2/s, BH (-0.31-0.46) × 10-3mm2/s, NT (.0.44-0.61) × 10-3 mm2/s; the right hepatic lobe:FB (-0.26-0.35) × 10-3mm2/sec, BH (-0.19-0.23) × 10-3mm2/s, NT (-0.29-0.36) × 10-3 mm2/s.The non-ECGT comparison showed that the repeatability of BH [left hepatic lobe:(-0.31-0.46) × 10-3 mm2/s; right lobe:(-0.19-0.23) × 10-3 mm2/s] was the best.When ECGT and respiratory compensation were simultaneously applied, the comparison of 12 ECGT and respiratory compensation sequences showed that the left hepatic lobe NT with ECG at 400 ms exhibited the best repeatability, and the smallest difference range was (-0.27-0.46) × 10-3 mm2/s. The right lobe NT with ECG at 600 ms exhibited the best repeatability, the minimal difference range was (-0.25-0.32) × 10-3 mm2/s.The comparison of 4 different delay times between the left and right lobes showed that, in general, the ECG at 600 ms showed better repeatability than 0 ms, 200 ms, and 400 ms. The consistency range of ECG at 600 ms of the left hepatic lobes after scanning was:FB (-0.44-0.54) × 10-3 mm2/s, BH (-0.34-0.55) × 10-3 mm2/s, NT (-0.44-0.6) × 10-3 mm2/s. The consistency range of ECG at 600 ms of the right hepatic lobes after scanning was:FB (-0.29-0.34) × 10-3 mm2/s, BH (-0.31-0.39) × 10-3 mm2/s, NT (-0.25-0.32) × 10-3 mm2/s.3.4 ADC measurements and repeatability of different hepatic coronal anatomical locationsAmong the three technologies(FB、BH、NT) accompanied with different ECGT delay times, the ADC measurements showed a decreasing trend from superior to inferior, and the left and right lobes both exhibited statistical significance (P< 0.001). From anterior to posterior, the ADC measurements among the three had no increasing or decreasing rule, while the differences were significant (P< 0.001). The pairwise comparisons (anterior and median, median and posterior, median and posterior) between the lobes were not statistically different (left lobe P= 0.058-0.943; right lobe P= 0.055-0.838).Using different respiratory compensation techniques accompanied with different ECGT delay times, the repeatability of the different hepatic coronal anatomical locations was not uniform, and the changes were large. Notably, the points with better repeatability of ROI obtained by FB and NT in the right lobe were more uniform, and the repeatability of inferior points of the anterior layer was good.3.5 Comparison of mean ADC values of left and right hepatic lobesThe measurement results of the mean ADC values and standard deviations from the left and right lobes showed that the ADC value of the left lobe was significantly greater than the right lobe (P< 0.001). When the delay time was shorter (0 ms-200 ms), the values of the right or left hepatic lobe obtained by NT with ECGT were higher than BH and FB with ECGT, and the differences were statistically significant.When the delay time was longer (400 ms and 600 ms), the difference of the mean ADC values of the left hepatic lobe using the three methods was not statistically significant. However, those of the right lobe at 400 ms were significant using these three methods, but the values at 600 ms obtained by FB and NT were greater than BH, and the differences were statistically significant. In the left hepatic lobe, compared with the minimal triggering time, NT with ECG at 25 ms (left hepatic lobe:1.812 × 10-3mm2/s) showed reduced ADC values at the longer delay time (NT with ECG at 400 ms, left hepatic lobe:1.700×10-3 mm2/s). In BH, with the delay time increasing, the ADC values were slightly increased, BH with ECG at 0 ms (left hepatic lobe: 1.637 × 10-3 mm2/s), BH with ECG at 600 ms (left hepatic lobe:1.715 × 10-3 mm2/s), and the difference was statistically significant (P= 0.008). In the right lobe, the BH ADC values were also significantly increased with increasing delay times (P< 0.001), but the ADC values obtained by FB with ECG and NT with ECG at different delay times did not change significantly (P= 0.063-0.667; P= 0.131).4. Conclusions1) When not accompanied with ECGT, BH could produce better repeatability of the left and right hepatic lobes.2) The comparison of 15 coronal sequences in the left hepatic lobe found that NT with ECGT at 400 ms showed better repeatability in the left hepatic lobe, which showed us ECGT could overcome the influence on left hepatic lobe caused by heart-beat. While ECGT in the right hepatic lobe was not effective. The integration of data of the right and left hepatic lobes showed that the data were reproducible and stable at 600 ms.