| BackgroundAt present, magnetic resonance diffusion weighted imaging (MR DWI) is the only noninvasive imaging technique in detecting water molecular diffusional movement and reflecting the physiological function in vivo. It has been evolved into a mature functional MR imaging technique for central neural system disorders. With recent developments in MR hardware and software, especially introduce of echo planar imaging (EPI), DWI has been expanded its use in abdominal, particularly in the liver. The applications of DWI in liver include liver lesion detection, lesion characterization and evaluation of tumor response to treatment and as well as staging in liver diffuse disease. DWI is a performed without use of gadolinium-based contrast media to assess the in vivo diffusion movement qualitatively and quantitatively, which is very useful in patients with severe renal dysfunction at the risk of nephrogenic systemic fibrosis.The apparent diffusion coefficient (ADC) derived from DWI represents mobility of water molecules within tissue and is therefore believed to reflect the changes in lesion cellularity and the development of microscopic tumor necrosis, which can happen before these changes become visible on conventional anatomic images. Many studies have demonstrated that ADC measurements are helpful in differentiating benign from malignant hepatic lesions and can be used as imaging biomarkers for the assessment of treatment response.However, liver DWI is still in its infancy, one of the critical problems to make sure consistent and widespread application of quantitative ADC measurements for disease characterization and tumor response assessment is the knowledge of ADC measurement reproducibility. They are influenced by many factors, such as imager, scanning sequence, parameters (b values), field of view (FOV), matrix, slice thick, etc. The Radiological Society of North America(RSNA) Quantitative Imaging Biomarkers Alliance(QIBA) suggested that the error caused by the influencing factors mentioned above can be resolved by establishing standardization work process. Even so, the effect of breathing movement, gastrointestinal peristalsis and independent body movement during data collection is difficult to eliminate. Many studies found that movement related artifacts is the most important factors that affect the result accuracy, among which breathing movement is the most one.Liver DWI acquisition technology includes free breathing (FB), multi-breathhold (MBH), respiratory triggered (RT), navigator-triggered (RT), etc. FB DWI allows multiple acquisition to improve the signal-to-noise ratio (SNR); at the same time, it can obtain thin images; Because of FB DWI without time constraints, it can obtain data for multi-b-value. But its disadvantages are mainly image fuzzy caused by respiratory movement and making the characterization of focal liver lesions inaccurate because of heterogeneity signal caused by volume effect. MBH DWI can minimize movement, but it has low signal-to-noise ratio, thicker thickness. And many patients tend to have small respiratory movement in the late breathless and can’t keep the same level at the end of each breathless, therefore it also has some movement artifact. Respiration triggered and navigator DWI technology both aquire data during the period of the minimal movement of diaphragm and to some extent can reduce the motion artifact. There are still movement of liver and diaphragm for the two methods within the time window of acquisition of data. Some researches show that the size of the various methods of the ADC values and its reproducibility of different acquisition technology is different. But what kind of technology with the best ADC values reproducibility is unclear. So we need to further explore the ADC values and its reproducibility obtained with four acquisition techniques.Besides imaging techniques, liver anatomical locations were demonstrated to affect ADCs as well. Studies showed that liver parenchyma in segment II had higher ADCs than those in any other segments and the reproducibility of the right lobe was better than the left. It has been shown that cardiac motion is the reason causing liver ADC variation, but to what extend does cardiac motion affect the ADCs in different liver anatomical locations is still unknown. Therefore, it is necessary for us to elucidate the effect of anatomical location on ADC values and its reproducibility.Purposes1. To compare the reproducibility of normal liver ADC measurements by using different respiratory motion compensation techniques with MBH, FB, RT, and NT DW imaging2. To compare the reproducibility of the ADC values in different normal liver anatomic locations.Materials and Methods Study populationThis prospective study was approved by our institutional review board, and written informed consent was obtained from each participant. Forty-seven healthy volunteers were referred for MR imaging of the liver from August2012to November2012. The inclusion criteria for this study were as follows:(a) no history of drug, viral hepatitis, alcohol abuse and abdominal surgery, and no current medication during the study,(b) normal appearance of the liver on ultrasound study (no focal and diffuse liver disease, including mild steatosis), and (c) ability to breath hold for up to20seconds. The exclusion criteria included:(a) a history or findings related to liver disease and contraindications to MR imaging;(b) failing to hold their breath for up to20seconds during DWI acquisition;(c) failing to complete the DWI procedure for any reasons, and (d) poor imaging quality not quantify to imaging analysis. MR DW Imaging ProtocolAll volunteers were imaged on a1.5-T MR scanner(Magnetom Espree, Siemens Medical Solutions, Germany) with a dedicated six-channel body matrix coil and a twelve-channel spine coil. End-expiratory MBH(4breath-holds), FB, RT (using an air-filled pressure sensor capable of measuring respiratory-induced pressure changes fixed to the hypochondrial region by a respiration belt around the volunteer) and NT(utilizing a100-mm long pencil-beam excitation prepulse at the interface between liver and lung to detect the position of the right diaphragm) single-shot echo-planar DWI were performed in axial view with three b-values(0,100,500s/mm2). Fat suppression was achieved with Spectral Adiabatic Inversion Recovery(SPAIR). Each volunteer was removed from the imager for a rest of approximately15minutes and then placed inside the magnet again for a second DWI series. The same imaging parameters were used for the repeated DWI series. In total, each volunteer got8DWI scans.Image AnalysisMR images were transferred to a Workstation(Viewforum; iMAC; Cupertino, CA, USA) for post-processing. ADC maps were calculated by mono-exponentialy fitting the three b-value data using the following equation ln(S)=-b*ADC+ln(So), where S was the signal intensity and b-equals100and500, and So was the signal intensity at b=0. The curve fitting used the linear least square method to find the optimum ADC that minimizes the summed square of the residuals. The algorithm was developed in house using Matlab7.7(Mathworks, USA).ADCs were measured with ImageJ (National Institutes of Health, Bethesda, MD) by two radiologists independently(Reader1and Reader2with5and3years of clinical experience in liver MRI, respectively). Reader1measured ADCs twice in a1-week interval following the same procedure to assess the intra-observer reproducibility, and his first measurement was used to compare with Reader2’s measurement to assess the inter-observer agreement.The right and left liver lobe was divided using the middle hepatic vein as a reference, and ADCs were evaluated separately. Three representative slices were selected for each DWI dataset. For the right liver lobe, the central slice was through the level of right portal vein, and the superior and inferior slices were3-4slice levels above or below the central slice, depending on the slice thickness(3slices for MBH,4slices for FB, RT and NT). For the left lobe, the central3continuous slices were selected.Three equally spaced circular ROIs, each measuring80mm2, were placed in the left and right liver lobes respectively in all three selected slices. All18ROIs were placed in the homogeneous liver parenchyma (visible vascular and biliary structures were excluded) with a margin of at least5mm from liver border. For the repeated DWI series, ROIs were placed in areas as similar as possible, which were achieved visually using anatomical landmarks, such as the main portal vein and its right branches in liver, etc.Therefore, a total of144ADCs were collected for each volunteer(two repeated series, three slices per lobe, two lobes, three ROIs per lobe per slice, four techniques).Statistical Analysis Estimation of sample sizeThe sample size was determined by using power analysis according to our pretest values between two repeated scanning for each technique by presuming an expected difference of0, an allocation ratio of1, alpha value of0.05and power of0.9, yielding an expected sample size of36. We enrolled a total of46subjects assuming a failure rate of about25%for the possibility of unpredicable events such as poor image quality or incomplete DWI scans.The expression of ADC values and Normality and F testADCs were expressed as mean ADCs±standard deviation (SD), and first tested with the Kolmogorov-Smirnov test for normality, then tested with Levene’s test for Variance homogeneity.The intra-observer and inter-observer agreementThe intra-observer and inter-observer agreement of ADC measurements were evaluated by using intra-class and inter-class correlation coefficient (ICC). The average ADCs from all the9ROIs was used for ICC calculation. Intra-observer ICC was computed from Reader l’s two measurements. Inter-observer ICC was computed from Reader l’s first measurements and Reader2’s measurements. An ICC greater than0.75was considered good agreement.The reproducibility test of ADC measurementsThe reproducibility of ADC measurements was evaluated with Bland-Altman method. The mean absolute difference (bias) and95%confidence interval (CI) of the mean difference, i.e. limits of agreement (LOA) between the first and second DWI series were compared.The comparison of ADC valuesTo evaluate the systematic bias of repeated ADC measurements, the average ADC from the9ROIs of both right and left liver lobe were compared between the two repeated series using paired-sample t-test for each technique.Next, for each MBH, FB, RT and NT DWI series, the differences between left and right liver lobes were evaluated by comparing the average ADCs from the9ROIs on each lobe using paired-sample t-test. The difference among these4techniques was assessed by using two-way classification analysis of variance(ANOVA).Finally, the difference in ADCs among the nine individual ROIs on each liver lobe was assessed by using three-way classification ANOVA. Bonferroni method was used to adjust for multiple comparisons when necessary.Statistical softwareStatistical analyses were performed using Statistical Package for the Social Sciences v.19.0software (SPSS, Chicago, IL) and MedCalc Software (MedCalc, Mariakerke, Belgium). Differences were considered significant when P values were less than0.05.ResultPopulation demographicsForty-six volunteers were enrolled in this study. Seven of them didn’t complete the study because of poor image quality due to their frequent bulk movement during acquisition (n=4) or incomplete acquisition of all DWI sequences ascribed to long acquisition time (n=3). Thirty-nine volunteers successfully completed the scans (15 men and24women; mean age,23.97years±2.75[standard deviation], age range,20-32years; mean age of men,25.07years±2.37, age range22-29years; mean age of women23.29years±2.79, age range20-32years.Intra-observer and inter-observer agreement ofADCs measurementThe intra-observer ICC calculated based on Reader l’s two measurements ranged from0.853to0.982. The lowest ICC0.853was from MBH scan, and the highest ICC0.982was from FB. Inter-observer agreement between Reader l’s first measurements and Reader2’s measurements was good for the4techniques, with ICCs ranging from0.829to0.952.For both intra and inter-observer agreement test, FB, NT and RT offer consistently higher ICC than MBH.Reproducibility of ADC measurements of left and right liver lobe with4techniquesOverall, ADC reproducibility of the right liver lobe was better than that of the left lobe with any one of the four techniques, and FB, RT and NT were always superior to MBH. For instance, the LOA between the2repeated DWI in the superior right ROI of the right liver lobe was0.640－10-3mm2/s with MBH,0.395×10-3mm2/s with FB,0.315×10-3mm2/s with RT and0.375×10-3mm2/s with NT, while the LOA in the superior right ROI of the left liver lobe was much higher,1.090×10-3mm2/s with MBH,0.600×10-3/s with FB,0.515×10-3mm2/s with RT and0.590×10-3mm2/s with NT.In the right liver lobe, reproducibility of ADC measurements of the9ROIs in different anatomical locations varied for each technique. The LOA ranged from0.365×10-3mm2/s to0.640×10-3mm2/s with MBH,0.325×10-3mm2/s to0,450×10-3mm2/s with FB,0.285×10-3mm2/s to0.435×10-3mm2/s with RT, and0.325×10-3mm2/s to0.440×10-3mm2/s with NT. Reproducibility of ADC measurements in the central-middle location was superior to other anatomical locations for all the techniques, giving the mean absolute differences of ADCs±LOA of (-0.03±0.365)×10-3mm2/s with MBH,(0.02±0.325)×10-3mm2/s with FB,(-0.02±0.285)×10-3mm2/s with RT and (-0.04±0.325)×10"3mm2/s with NT.For the left liver, the LOA of the9ROIs were all larger than those of the right liver, in which the LOA ranged from0.590×10-3mm2/s to1.090×10-3mm2/s with MBH,0.425×10-3mm2/s to0.675×10-3mm2/s with FB,0.360×10-3mm2/s to0.810×10-3mm2/s with RT and0.395×10-3mm2/s to0.630×10-3mm2/s with NT. Taking both the mean absolute difference and LOA into consideration, ADC of the inferior-right ROI showed the best reproducibility, as (0.01±0.590)×103mm2/s with MBH,(0.01±0.365)×10-3mm2/s with FB,(0.02±0.360)×10-3mm2/s with RT, and(0.07±0.435) x10-3mm2/s with NT.Mean ADCs of left and right liver lobe with4techniquesFor both readers, the left liver lobe ADCs were significantly higher than the right with all4techniques (P<0.001).ADCs with MBH (Right[R]:(91.641-1.662)×10-3mm2/s; Left[L]:(2.034-2.054)×10-3mm2/s) were significantly higher than those with FB (R:(1.349-1.391)×10-3mm2/s; L:(1.630-1.700)×10-3mm2/s), RT (R:(1.439-1.455)×10-3mm2/s; L:(1.720-1.755)×10-3mm2/s) and NT (R:(1.387-1.400)×10-3mm2/s; L:(1.661-1.736)×10-3mm2/s) for both readers on both repeated scans and for both left and right liver lobes (P<0.001), but no statistical difference was noticed among FB, RT and NT techniques (P=0.130-1.000).ADCs of different anatomical liver locationsFor all4techniques, ADCs showed a decreasing trend in the left-right direction. The differences of ADCs in that direction for both left and right liver lobes were all significantly different (P<0.001for right lobe, and P=0.001-0.016for left lobe). On the other hand, in the superior-inferior direction, ADCs of the left lobe clearly decreased (P<0.001), but no significant difference was observed in the right lobe (P=0.144-0.450). Standard deviation (SD) of ADCs of the left liver lobe was in general much larger than that of the right liver lobe.ConclusionsIn conclusion, we recommend that ADC measurement of the liver parenchyma be performed with FB DW imaging in clinical practice and research because of its good reproducibility and shorter acquisition time compared with that of MBH, RT, and NT DW imaging. ADC of the central middle segment in the right lobe has the best reproducibility and should be used as the reference standard when liver ADC is used as a biomarker for clinical applications, such as monitoring treatment responses. |
PDF Full Text Request