Abstract
Background
The magnetic resonance imaging (MRI)-based proton density fat fraction (PDFF) has become popular for quantifying liver fat content. However, the variability of the region-of-interest (ROI) sampling strategy may result in a lack of standardisation of this technology. In an effort to establish an accurate and effective PDFF measurement scheme, this study assessed the pathological correlation, the reader agreement, and time-burden of different sampling strategies with variable ROI size, location, and number.
Methods
Six-echo spoiled gradient-recalled-echo magnitude-based fat quantification was performed for 50 patients with obesity, using a 3.0-T MRI scanner. Two readers used different ROI sampling strategies to measure liver PDFF, three times. Intra-reader and inter-reader agreement was evaluated using intra-class correlation coefficients and Bland‒Altman analysis. Pearson correlations were used to assess the correlation between PDFFs and liver biopsy. Time-burden was recorded.
Results
For pathological correlations, the correlations for the strategy of using three large ROIs in Couinaud segment 3 (S3 3L-ROI) were significantly greater than those for all sampling strategies at the whole-liver level (P < 0.05). For inter-reader agreement, the sampling strategies at the segmental level for S3 3L-ROI and using three large ROIs in Couinaud segment 6 (S6 3L-ROI) and the sampling strategies at the whole-liver level for three small ROIs per Couinaud segment (27S-ROI), one large ROI per Couinaud segment (9L-ROI), and three large ROIs per Couinaud segment (27S-ROI) had limits of agreement (LOA) < 1.5%. For intra-reader agreement, the sampling strategies at the whole-liver level for 27S-ROI, 9L-ROI, and 27L-ROI had both intraclass coefficients > 0.995 and LOAs < 1.5%. The change in the time-burden was the largest (100.80 s) when 9L-ROI was changed to 27L-ROI.
Conclusions
For hepatic PDFF measurement without liver puncture biopsy as the gold standard, and for general hepatic PDFF assessment, 9L-ROI sampling strategy at the whole-liver level should be used preferentially. For hepatic PDFF with liver puncture biopsy as the gold standard, 3L-ROI sampling strategy at the puncture site segment is recommended.
Similar content being viewed by others
Explore related subjects
Find the latest articles, discoveries, and news in related topics.Background
Nonalcoholic fatty liver disease (NAFLD) involves the excessive deposition of triglycerides in the liver and has become a major public health concern globally [1]. Obesity is an important risk factor for NAFLD and nonalcoholic steatohepatitis (NASH) [2, 3]. Without intervention, fatty liver may develop into steatohepatitis, liver fibrosis, liver cirrhosis, and hepatocellular carcinoma [1, 4, 5]. Recently, a consensus of international experts has proposed a change in the disease name from NAFLD to metabolic-associated fatty liver disease (MAFLD), which is a positive diagnosis that reflects the close relationship between fatty liver and overnutrition, sedentary lifestyle, and metabolic conditions [6]. The criteria are based on evidence of hepatic steatosis, plus any of the following three conditions: Overweight/obesity, presence of type 2 diabetes mellitus, or presence of metabolic disorder [7].
Currently, the gold standard for the diagnosis and grading of hepatic steatosis is liver biopsy. However, this procedure can occasionally lead to complications, such as bleeding, liver hematoma, and puncture of the gallbladder or adjacent organs [8,9,10,11,12]. For these reasons, noninvasive detection might be the preferred method for detecting MAFLD/NAFLD.
Over the past 20 years, there has been a shift from qualitative to quantitative assessment of liver steatosis by imaging, with current research investigating the use of ultrasound, computed tomography, and magnetic resonance imaging (MRI) for improving the accuracy of fat quantification [13,14,15]. MRI-based proton density fat fraction (MRI-PDFF) with a multi-echo chemical-shift-encoded sequence is now routinely used at different centres to quantify hepatic steatosis, with excellent diagnostic performance [13,14,15,16,17,18]. This assessment utilises multi-echo water‒lipid separation technology to eliminate the interference of T1 and T2* effects, as well as main magnetic field (B0) inhomogeneity in the fat quantification process, which makes accurate quantification of visceral fat possible [19].
The optimal sampling strategy based on the regions-of-interest (ROIs) has not yet been established. The PDFF of the liver is usually evaluated by radiologists by manually placing different numbers of ROIs at different locations of the liver and averaging these values [16,17,18,19,20,21,22]. Furthermore, the spatial distribution of liver fat is not uniform, which results in different PDFFs at different locations [22,23,24]. This lack of standardisation leads to further PDFF variability, which affects the widespread use of this technology as a quantitative imaging biomarker. Currently, the common approach in clinical trials is to place 1–3 ROIs from each Couinaud liver segment [22, 23, 25, 26]. Several reports have indicated that large ROIs with a 4-ROI (anterior, posterior, medial, and lateral), balanced 4-ROI (2 ROIs per lobe), 9-ROI (1 ROI per Couinaud segment), or 27-ROI strategy (3 ROIs per Couinaud segment) are preferred for good intra- and inter-reader agreements [20, 21, 24, 26,27,28,29]. The common sampling method measures PDFF at the whole-liver level, which requires a longer time [28, 29]. Some studies have suggested the use of a balanced 4-ROI (2 ROIs per lobe) approach, which provides high repeatability and low error, in order to reduce time consumption [20, 21]. Although the above multi-ROI method provide good reader agreements, which can avoid the variability of sampling, the gold standard for detecting "steatosis" is liver biopsy, which can accurately reflect the histological situation of the liver at the puncture site. Changing of ROI sampling strategies may vary in performance in terms of pathological correlation between liver biopsy and PDFF.
Therefore, in an effort to establish an accurate and effective PDFF measurement scheme for clinicians and radiologists, this study assessed the pathological correlation, the intra- and inter-reader agreement, and time-burden of different sampling strategies with variable ROI size, location, and number.
Methods
Study design and patients
We performed a secondary analysis of cross-sectional data from a prospective, self-controlled observational study conducted by Li et al. [26] at the Beijing Friendship Hospital, Beijing, China, from December 2017 to November 2018. The study used MRI-PDFF to determine the changes in hepatic PDFF characteristics after bariatric surgery, to determine the correlation with the clinical NASH score [30], and to determine the predictors of postoperative NASH score changes. The inclusion criteria were an age between 18 and 65 years, diagnosis of obesity (BMI ≥ 28 kg/m2) [31, 32], and willingness to accept bariatric surgery intervention. The exclusion criteria were a history of obesity surgery interventions or general contraindications for MRI.
Sixty-nine patients were enrolled and underwent abdominal MRI at baseline and again at 3 months after surgery. Fifty patients underwent liver biopsy at baseline. We used the pathological biopsy as the pathological reference standard of these 50 patients and measured their liver PDFFs at baseline. Eleven of 50 patients had a history of type 2 diabetes for at least half a year and regularly used insulin, metformin, or other hypoglycaemic drugs. Standard clinical, anthropometric, and biochemical measurements of these 50 patients at baseline were also obtained. The parameters included sex, age, body weight, body mass index (BMI), and serum levels of alanine aminotransferase, alanine aminotransferase, triglycerides, and glycated haemoglobin.
The study was approved by the Ethics Committees of the Beijing Friendship Hospital where the study was conducted (NO. 2017-P2-131–02). Informed consent was obtained from the patients for the use and publication of their data.
Liver biopsy and histopathological analysis
According to the international surgical standards, all patients underwent laparoscopic Roux-en-Y gastric shunt or sleeve gastrectomy bariatric surgery. Wedge liver biopsy was performed in the left liver lobe (area of Segment 3) at the same time as the surgery, by a single operator.
Liver biopsy material were fixed in 10% buffered formalin, and was sent to the Biobank of Beijing Friendship Hospital for unified pathological sectioning and staining (haematoxylin‒eosin, reticular fibre, Masson, copper, and iron staining). Haematoxylin‒eosin staining was used to evaluate hepatic steatosis, as the pathological reference standard. A single liver pathologist with 12 years of experience, who was blinded to the clinical and radiological data, used the near continuous scale (0%, 5%, 10%, 20%, 30%, …, 100%) to score hepatic steatosis. The mean value of the fatty degeneration score was recorded and converted to the following four-point ordinal score, as defined by the NASH Clinical Research Network scoring system [33]: 0 (< 5% hepatocytes), 1 (5–33% hepatocytes), 2 (33–66% hepatocytes), and 3 (> 66% hepatocytes).
MRI quantification technique
MRI was performed within 1‒2 days before surgery. All patients underwent non-contrast scans using a 3.0-Tesla MRI scanner (MR750, GE Healthcare, Waukesha, WI, USA) with an eight-channel phased-array body coil centred over the liver. The patients were instructed to fast for 6 h before undergoing scanning with feet first in the supine position. A six-echo spoiled gradient-recalled-echo magnitude-based fat quantification technique was used. The parameters were as follows: repetition time, 7.3 ms; echo time, six different echoes ranging from 1.0 to 5.0 ms to permit correction for R2* signal decay and chemical-shift-based separation of fat and water signals; matrix, 160 × 160; field-of-vision, 40 cm; slice-section thickness, 8 mm; number of excitations, 0.5; flip angle, low (4°) to minimise T1 bias [34]; bandwidth, 111.11; and acquisition time, 21 s. Respiratory bellows were applied to monitor breathing, and the patients were instructed to hold their breath during image acquisition. The sequences were planned such that the whole liver was covered.
PDFF measurements
A radiologist with 8 years of experience in abdominal imaging diagnosis (Reader 1) and another with 3 years of experience in abdominal imaging diagnosis (Reader 2) used the Picture Archiving and Communication System (PACS) to view the images.
To verify the impact of different ROI sampling strategies (different location, number, and size) on intra- and inter-reader agreement, pathological correlation, and time-burden, we used ROIs at different locations, different numbers, and different sizes for analysis according to the sampling strategies reported in previous studies [20, 21, 24, 26,27,28,29]. Locations included each of the Couinaud segments of the liver at the segmental level (including segment 1 [S1], segment 2 [S2], segment 3 [S3], segment 4a [S4a], segment 4b [S4b], segment 5 [S5], segment 6 [S6], segment 7 [S7] and segment 8 [S8]), as well as two points in the left and right lobes and all nine Couinaud segments at the whole liver level. Readers freely selected the slices. For quantity, first, each reader manually placed one and three ROIs in each of the nine Couinaud segments, and then two ROIs in the left and right lobes. Therefore, the number of ROIs included: one ROI (one ROI in a single liver segment), three ROIs (three ROIs in a single liver segment), nine ROIs (one ROI in each Couinaud segment), 27 ROIs (three ROIs in each Couinaud segment), and four ROIs (two ROIs in the left lobe, two ROIs in the right lobe). In terms of size, we used large (> 4 cm2; if the S1 volume was small, we used the largest area that fit inside each placement to make ROIs > 3 cm2) and small (< 4 cm2; if the S1 volume was small, we used an ROI < 3 cm2) ROIs to measure PDFF. Large vessels, hepatic margin, and obvious image artifacts were avoided.
Through the above methods, we achieved different sampling strategies as follows: one small ROI in a single Couinaud segment (1S-ROI strategy) at the segmental level, three small ROIs in a single Couinaud segment (3S-ROI strategy) at the segmental level, one large ROI in a single Couinaud segment (1L-ROI strategy) at the segmental level, three large ROIs in a single Couinaud segment (3L-ROI strategy) at the segmental level, one small ROI per Couinaud segment (9S-ROI strategy) at the whole-liver level, three small ROIs per Couinaud segment (27S-ROI strategy) at the whole-liver level, one large ROI per Couinaud segment (9L-ROI strategy) at the whole-liver level, three large ROIs per Couinaud segment (27S-ROI strategy) at the whole-liver level, two small ROIs in the left and right lobes (4S-ROI strategy), and two large ROIs in the left and right lobes (4L-ROI strategy) at the whole-liver level (Figs. 1, 2, 3).
Visual representation of sampling strategy of the location and number of regions-of-interest (ROIs) placed by the readers on proton density fat fraction (PDFF) maps. Yellow, red, and green circles represent the first, second, and third ROIs placed in each liver segment, respectively. The PDFF measured by yellow ROIs on each single liver segment represents 1S-ROI. The mean PDFF measured by yellow, red, and green ROIs on each single liver segment represents 3S-ROI. The mean PDFF measured by all yellow ROIs on all liver segments represents 9S-ROI. The mean PDFF measured by all yellow, red, and green ROIs on all liver segments represents 27S-ROI
Visual representation of sampling strategy of the location and number of large regions-of-interest (ROIs) placed by the readers on proton density fat fraction (PDFF) maps; yellow, red, and green circles represent the first, second, and third ROIs placed in each liver segment, respectively. The PDFF measured by yellow ROIs on each single liver segment represents 1L-ROI. The mean PDFF measured by yellow, red, and green ROIs on each single liver segment represents 3L-ROI. The mean PDFF measured by all yellow ROIs on all liver segments represents 9L-ROI. The mean PDFF measured by all yellow, red, and green ROIs on all liver segments represents 27L-ROI
Three months later, Reader 1 re-measured the PDFFs in the original image to assess the intra-reader agreement, using the same measurement method as that used the first time. Consequently, we obtained three measurement results for two readers.
The MRI-PDFF images were assessed by readers in a double-blind method at the image analysis. However, Reader 1 knew that the image of the second measurement was the same as that used in the first round. The two readers were also blinded to the clinical data and the histological data.
Time-burden
We recorded the time-burden for one person (Reader 1, first time) to measure the time for each sampling strategy. The recorded time encompassed opening the PDFF diagram in the PACS, selecting the slice, drawing the ROI on the PDFF diagram, and exporting the ROI measurement data to Excel (version 2016, Microsoft Corp, Redmond, WA, USA). The time-burden at the segmental level was recorded as the average of the total time spent in each Couinaud segment using one sampling strategy. The time-burden at the whole-liver level was recorded as the time spent using one sampling strategy.
Statistical analysis
Pearson correlations were used to assess the correlation between the histological degree of steatosis and PDFF. To determine whether the pathological correlations had significant differences, we compared the pathological correlations of each sampling strategy at the segmental and whole-liver levels, and pairwise comparison was made between the pathological correlations of two sampling strategies at the whole-liver level (Tables 1, 2, 3). Because the procedure involved three measurements for time, the above correlations were compared and analysed for each time measurement dataset.
Intra-reader and inter-reader agreement were evaluated using intra-class correlation coefficients (ICCs) and Bland‒Altman analyses. The consistency between the first and second PDFF measurements of Reader 1 represents the intra-reader agreement. The consistency between the first PDFF measurement of Reader 1 and PDFF measurements of Reader 2 represented the inter-reader agreement. We used two thresholds to define close agreement: ICCs > 0.995 and absolute limits of agreement (LOA) < 1.5%. These thresholds were selected a priori to represent clinically meaningful differences [20].
The pathological correlation differences were compared relied on tests implemented in the cocor R package [35] by calculating Fisher’s z. Other statistical analyses were performed using MedCalc v. 18.2.1 (MedCalc Software, Ostend, Belgium) and SPSS v25.0 (IBM Inc., Armonk, NY, USA). The significance threshold was set at p = 0.05.
Results
Patient characteristics
The study sample comprised 40 women (mean age: 32.85 ± 8.88 years; range, 17–54 years) and 10 men (mean age: 37.20 ± 8.19 years; range, 22–50 years). The overall average age of 50 patients was 33.72 ± 8.84 years (range, 17–54 years). The summaries of body weight, BMI, liver biochemical parameters, and blood lipid and HbA1c levels, and histological degree of steatosis in 50 patients are presented in Table 4.
PDFF measurement
Liver PDFFs measured by different strategies by Readers 1 and 2 are summarised in Table 5.
Correlations and correlation comparison
The pathological correlations (PDFF and biopsy) of different sampling strategies and correlation comparisons are listed in Tables 1, 2, 3. The results showed that the sampling strategy of S8 1S-ROI and S8 1L-ROI by Reader 2 had correlation coefficients < 0.8 (r = 0.796 and 0.788, respectively), while other sampling strategies had correlation coefficients > 0.8 (r = 0.811‒0.904). Only the correlations of S3 3L-ROI were greater than those of all sampling strategies at the whole-liver level (4S-ROI, 4L-ROI, 9S-ROI, 9L-ROI, 27S-ROI, and 27L-ROI), and the differences were significant (P < 0.05). All pairwise comparisons between the pathological correlations of the two sampling strategies at the whole-liver level had no significant differences (P > 0.05).
ICC and bland–altman analysis
ICC and Bland–Altman analysis of intra- and inter-reader agreements using different ROI sampling strategies are summarised in Tables 6 and 7. The results showed a general trend of increasing ICC and decreasing LOA bounds (closer intra- and inter-reader agreements) as the number and size of ROIs increased at both segmental and whole-liver levels (Figs. 4, 5, 6, 7).
Intra-class correlation coefficient (ICC) of sampling strategies at the segmental and whole-liver levels for inter-reader agreement increased with the number and size of regions-of-interest. a The ICC of sampling strategies at the segmental level. b The ICC of sampling strategies at the whole-liver level. Each dot or bar represents a particular sampling strategy. The dotted line indicates an ICC of 0.995
Intra-class correlation coefficient (ICC) of sampling strategies at the segmental and whole-liver levels for intra-reader agreement increased with the number and size of regions-of-interest. a The ICC of sampling strategies at the segmental level. b The ICC of sampling strategies at the whole-liver level. Each dot or bar represents a particular sampling strategy. The dotted line indicates an ICC of 0.995
The limits of agreement (LOA) of sampling strategies at the segmental and whole-liver levels for inter-reader agreement decreased with the number and size of regions-of-interest. a The LOA of sampling strategies at the segmental level. b The LOA of sampling strategies at the whole-liver level. Each dot or bar represents a particular sampling strategy. The dotted line indicates an LOA of 1.5%
The limits of agreement (LOA) of sampling strategies at the segmental and whole-liver levels for intra-reader agreement decreased with the number and size of regions-of-interest. a The LOA of sampling strategies at the segmental level. b The LOA of sampling strategies at the whole-liver level. Each dot or bar represents a particular sampling strategy. The dotted line indicates an LOA of 1.5%
For inter-reader agreement, the sampling strategies at the segmental level for S1 3L-ROI, S3 3L-ROI, S5 3L-ROI, S6 3L-ROI, S7 3L-ROI, and S8 3L-ROI had ICCs > 0.995. The sampling strategies at the segmental level for S3 3L-ROI and S6 3L-ROI had LOAs < 1.5%. The sampling strategies at the whole-liver level for 4L-ROI, 9S-ROI, 27S-ROI, 9L-ROI, and 27L-ROI had ICCs > 0.995, while the sampling strategies at the whole-liver level for 27S-ROI, 9L-ROI, and 27L-ROI had LOAs < 1.5%. The sampling strategy for 27L-ROI at the whole-liver level had the largest ICC and narrowest LOA.
For intra-reader agreement, the sampling strategies at the segmental level for S1 3L-ROI, S6 3L-ROI, and S7 3L-ROI had ICCs > 0.995. None of the sampling strategies at the segmental level had LOAs < 1.5%. The sampling strategies at the whole-liver level for 4L-ROI, 27S-ROI, 9L-ROI, and 27L-ROI had ICCs > 0.995, while the sampling strategies at the whole-liver level for 27S-ROI, 9L-ROI, and 27L-ROI had LOAs < 1.5%. The sampling strategy for 27L-ROI had the largest ICC and narrowest LOA.
Time-burden
The time measurement data are summarised in Table 8. Overall, the time-burden increased with the increase in the size and number of ROIs.
The size of ROI had little impact on time-burden. The change from small ROIs to large ROIs increased the measurement time by 6% (0.33 s) for the 1-ROI sampling strategy at the segmental level, by 4% (0.71 s) for the 3-ROI sampling strategy at the segmental level, by 13% (3.00 s) for the 4-ROI sampling strategy at the whole-liver level, by 6% (2.98 s) for the 9-ROI sampling strategy at the whole-liver level, and by 4% (6.39 s) for the 27-ROI sampling strategy at the whole-liver level.
The number of ROIs had a marked impact on time-burden. For the small ROI sampling strategy at the segmental level, the measurement time increased by 69% (10.82 s) when 1-ROI was changed to 3-ROI. For the large ROI sampling strategy at the segmental level, the measurement time increased by 68% (11.20 s) when 1-ROI changed to 3-ROI. Using the 4S-ROI sampling strategy at the whole-liver level increased the measurement time by 21% (4.06 s) as compared with the 3S-ROI sampling strategy at the segmental level. Using the 4L-ROI sampling strategy at the whole-liver level increased the measurement time by 28% (6.35 s) compared with the 3L-ROI sampling strategy at the segmental level. For the small ROI sampling strategy at the whole-liver level, the measurement time increased by 55% (24.28 s) when 4-ROI changed to 9-ROI and by 69% (97.39 s) when 9-ROI changed to 27-ROI. For the large ROI sampling strategy at the whole-liver level, the measurement time increased by 52% (24.26 s) when 4-ROI changed to 9-ROI and by 68% (100.80 s) when 9-ROI changed to 27-ROI.
Discussion
The ROI sampling strategy in MRI-PDFF may cause marked variability. To improve accuracy and effective PDFF measurement, we evaluated the pathological correlation, intra- and inter-reader agreement, and time-burden of different sampling strategies for MRI-PDFF, with variation in ROI size, location, and number. We found that whole-liver level sampling strategies of 27S-ROI, 9L-ROI, and 27L-ROI provided the best intra-reader agreement and inter-reader agreement. Correlations when using S3 3L-ROI on the puncture site segment were significantly greater than those for all sampling strategies at the whole-liver level. Time-burden increased the most (by 100.80 s) when 9L-ROI strategy was changed to 27L-ROI strategy.
Fat distribution in the liver may not be uniformly distributed [22,23,24, 36, 37], for numerous reasons. It is hypothesised that adiposis may be related to insulin, because insulin can stimulate the conversion of glucose to fatty acids. Due to the different concentrations of insulin in the "third liver inflow", which are some aberrant small veins (e.g., the gastric and cystic veins) that directly enter the liver, the fat content in specific areas of the liver can be increased or decreased [36, 37]. In addition, the mesenteric vein contains more nutrients, and the blood from the mesenteric vein is preferentially shunted to the right lobe through the right portal vein, while the blood of the splenic vein, with less nutrients, is preferentially shunted to the left lobe through the left portal vein, thereby resulting in more triglyceride deposition in the right lobe [22, 23]. For these reasons, an abnormal distribution may be missed by liver biopsy. Therefore, noninvasive imaging methods, such as MRI-PDFF, which can cover the whole liver and which have excellent diagnostic value, could be preferable for evaluating the fat distribution and longitudinal changes therein. Although MRI-PDFF, as a non-invasive quantitative imaging biomarker, has been proven to have excellent repeatability and reproducibility across different field strengths, manufacturers, and reconstruction approaches [38, 39], it is inevitable that the use of different sampling strategies by radiologists would result in varying PDFFs. Owing to this lack of standardisation in the sampling strategy used, further variability is introduced, which affects the widespread use of MRI-PDFF as a reproducible quantitative imaging biomarker.
Campo et al. [28] measured two ROIs (left and right liver lobes), four ROIs (anterior, posterior, medial, and lateral liver segments), and nine ROIs (nine Couinaud segments) with sizes of 1 cm2, 4 cm2, and large fit, respectively. They suggested that multiple large ROIs should be used to sample liver PDFF. Hooker et al. [29] performed a secondary cross-sectional and longitudinal analysis of MRI-PDFF data. PDFFs were measured by placing one primary and two additional ROIs in each segment. They found that the 27-ROI and 9-ROI PDFF measurements had ICCs of 0.998 and 0.997, respectively, and that the ICC for primary-ROI, single-segment PDFF measurements ranged from 0.957 to 0.990, depending on the segment. Our results support the above research. In our study, two readers placed different numbers of ROIs of varying sizes in each Couinaud segment to obtain PDFF, using different measurement methods, after which the intra-reader and inter-reader agreements were analysed. For a sampling strategy at the segmental level, the 1S-ROI strategy had the smallest mean ICC and the widest mean LOA, while the 3L-ROI strategy had the largest mean ICC and the narrowest mean LOA. At the segmental level, the 4S-ROI strategy had the smallest ICC and the widest LOA, while the 27L-ROI strategy had the largest ICC and the narrowest LOA. Regardless of intra-reader or inter-reader agreement, 27S-ROI, 9L-ROI, and 27L-ROI strategies at the whole-liver level had ICCs > 0.995 (0.998–0.999) and LOAs < 1.5% (0.56%–1.39%). These results also demonstrated that multiple larger ROIs can increase agreement at both the segmental and whole-liver levels. Hong et al. [20, 21] suggested using a balanced 4-ROI strategy (radius: 1 cm) to measure PDFF. In one previous study [20], all balanced 4-ROI strategies yielded an LOA < 1.5%. In another study [21], all balanced 4-ROI strategies achieved inter- and intra-examination ICCs > 0.998, 92% achieved an intra-examination repeatability coefficient (RC) < 1%, 83% achieved an inter-examination RC < 1%, and all achieved a mean absolute error < 1%. The methods of 4-ROI strategies in our study are similar to those of the balanced 4-ROI strategy. We evaluated the impact of the ROI size of this balanced 4-ROI strategy on reader agreements. We found that the 4S-ROI strategy and the 4L-ROI strategy had LOAs > 1.5%, which do not reach the threshold of LOA < 1.5%. However, the 4L-ROI strategy exceeded the threshold at ICC > 0.995. This suggests that, if PDFF is measured with a balanced strategy, more large ROIs may be required (e.g., 3 ROIs per lobe). The result may have been affected by the small sample size of this study.
In this study, we not only evaluated the reader agreements for some commonly used different sampling strategies at the whole-liver level, but also studied the influence of the number and size of ROIs at the segmental level on the reader agreements, to provide a reference for the determination of a clinical liver PDFF measurement scheme. Idilman et al. [17] previously placed a 4-cm2 ROI in segments V and VI and a 2-cm2 ROI in other segments to obtain the average PDFF of the liver. They found that the PDFF was closely related to pathological hepatic steatosis (r = 0.788–0.904).
Although MRI-PDFF has been proven to be a good indicator of the degree of hepatic steatosis, there is inhomogeneity of fat distribution, resulting in sampling error in liver biopsy [40]. We suspect that placement of ROIs in different regions of the liver may affect the correlation between PDFF and pathology. Liver biopsy is the gold standard for the diagnosis of MAFLD/NAFLD and evaluation of longitudinal liver histological changes [40,41,42,43,44,45]. For obese and bariatric surgery patients, questions about appropriate biopsy techniques and locations often arise because of the high prevalence of MAFLD/NAFLD and opportunity to perform liver biopsy. Since the criteria for the best selection of MAFLD/NAFLD biopsy technology and puncture location still need to be established [46], it may be necessary to investigate whether the ROI location should be consistent with the location of the liver puncture site.
According to our knowledge, no previous study has reported whether there are differences in the correlation of liver PDFF measured by different sampling strategies with biopsy as the gold standard. An important and unique advantage of the present study is that we used liver biopsy as the standard to compare the pathological correlations (PDFF and biopsy). Although all sampling strategies at the whole-liver level had high correlation coefficients (r = 0.864–0.875), the puncture site segment (S3) 3L-ROI sampling strategy yielded the highest correlation coefficient (r = 0.895–0.904) in all three measurements. In addition, the correlation of S3 3L-ROI was higher than the correlations of all sampling strategies at the whole-liver level in all three measurements, with statistically significant differences. The reason for this may be that multiple large ROIs placed on S3 may be more suitable for obtaining the PDFF of the liver tissue located at or around the puncture site, offsetting the influence of the sampling error.
Bonekamp et al. [22] used MRI to evaluate the spatial distribution and variability of liver fat in patients without NAFLD. In that study, PDFF in the left lobe was lower than that in the right lobe. Chaim et al. [47] found that the changes in fat content in different liver areas after weight loss surgery in obese patients were different, with right lobe segments having higher PDFF at baseline and showing more rapid reduction in liver fat content. PDFF measured at the whole-liver level may not accurately reflect the histological changes of liver biopsy at the puncture site. For such cases, the sampling strategy of using at least three large ROIs in the puncture site segment may be a reasonable choice, as PDFF measured by this sampling strategy can accurately reflect liver biopsy.
Our study recorded the measurement time-burden in an attempt to provide a compromise between measurement performance of different sampling strategies and the time-burden of liver PDFF. Our results were similar to those of Campo et al. [20], who found that an increase in the number and size of ROIs can increase the time-burden for the reader to obtain the ROI-based liver PDFF artificial measurements. Notably, the increase in the number of ROIs had more effect on time-burden than did the increase in ROI size. The reasons for this result are obvious: A larger ROI makes the reader spend more time observing other structures, such as large blood vessels and bile ducts, to avoid the influence of these anatomical structures, and an increase in the number of ROIs requires the reader to spend relatively more time looking for the corresponding level and location to draw and input data. Therefore, although the sampling strategy of 27L-ROI demonstrates the best consistency, it is the most time-consuming sampling strategy.
Based on the comprehensive consideration of reader agreement, correlation, and trade-off of time-burden, we suggest the 9L-ROI sampling strategy, with an analysis time of approximately 45 s, at the whole-liver level, for hepatic PDFF measurement without liver puncture biopsy as the gold standard and for hepatic PDFF in day-to-day analysis. For hepatic PDFF with liver puncture biopsy as the gold standard, 3L-ROI sampling strategy, with an analysis time of approximately 15 s, at the puncture point segment is suggested.
This study had some limitations. First, this was a single-centre study of adult obese patients in northern China, which may limit the universality of our results. Second, liver biopsy was only obtained at S3 in this study, and an uneven fat distribution in the liver of patients may have affected the results. In future, the corresponding liver biopsy may be used to evaluate and analyse PDFF in different liver lobes or even liver segments, to verify the results of this study. Previous studies have demonstrated the feasibility of whole-liver ROIs and automated MRI liver segmentation to measure PDFF [48,49,50]. The above methods were not evaluated in this study because such methods cannot avoid anatomical structures, such as large blood vessels, in the liver, which may affect the MRI-PDFF. In addition, no fixed slices were used for ROI measurements in this study, which may have led to variation of PDFF and time-burden.
Conclusions
Our study showed that PDFF measured by the sampling strategy of 3L-ROI on the puncture site segment can most accurately reflect the liver biopsy. Intra-reader and inter-reader agreements are improved when multiple large ROIs are used to measure PDFF. In addition, the number of ROI has a greater impact on the time-burden than does the size of ROIs. For hepatic PDFF measurement without liver puncture biopsy as the gold standard and for hepatic PDFF day-to-day analysis, 9L-ROI sampling strategy, with an analysis time of approximately 45 s, at the whole-liver level is suggested. For hepatic PDFF with liver puncture biopsy as the gold standard, 3L-ROI sampling strategy, with an analysis time of approximately 15 s, at the puncture point segment is suggested.
Availability of data and materials
The data that support the findings of this study are available from the corresponding author, but restrictions apply to the availability of these data, which were used under license for the current study, and thus are not publicly available. Data are, however, available from the authors upon reasonable request and with permission of the corresponding author.
Abbreviations
- BMI:
-
Body mass index
- ICC:
-
Intra-class correlation coefficients
- MRI:
-
Magnetic resonance imaging
- MAFLD:
-
Metabolic-associated fatty liver disease
- NAFLD:
-
Nonalcoholic fatty liver disease
- NASH:
-
Nonalcoholic steatohepatitis
- PACS:
-
Picture archiving and communication system
- PDFF:
-
Proton density fat fraction
- ROI:
-
Region-of-interest
References
Loomis AK, Kabadi S, Preiss D, Hyde C, Bonato V, St Louis M, et al. Body mass index and risk of nonalcoholic fatty liver disease: two electronic health record prospective studies. J Clin Endocrinol Metab. 2016;101:945–52.
Vernon G, Baranova A, Younossi ZM. Systematic review: The epidemiology and natural history of non-alcoholic fatty liver disease and non-alcoholic steatohepatitis in adults. Aliment Pharmacol Ther. 2011;34:274–85.
Leoni S, Tovoli F, Napoli L, Serio I, Ferri S, Bolondi L. Current guidelines for the management of non-alcoholic fatty liver disease: A systematic review with comparative analysis. World J Gastroenterol. 2018;24:3361–73.
Abd El-Kader SM, El-Den Ashmawy EM. Non-alcoholic fatty liver disease: The diagnosis and management. World J Hepatol. 2015;7:846–58.
Schwimmer JB. Definitive diagnosis and assessment of risk for nonalcoholic fatty liver disease in children and adolescents. Semin Liver Dis. 2007;27:312–8.
Wong VW, Wong GL, Woo J, Abrigo JM, Chan CK, Shu SS, et al. Impact of the new definition of metabolic associated fatty liver disease on the epidemiology of the disease. Clin Gastroenterol Hepatol. 2021;19:2161–71.
Eslam M, Sanyal AJ, George J. MAFLD: A consensus-driven proposed nomenclature for metabolic associated fatty liver disease. Gastroenterology. 2020;158:1999–2014.
Seeff LB, Everson GT, Morgan TR, Curto TM, Lee WM, Ghany MG, et al. Complication rate of percutaneous liver biopsies among persons with advanced chronic liver disease in the HALT-C trial. Clin Gastroenterol Hepatol. 2010;8:877–83.
Barbois S, Arvieux C, Leroy V, Reche F, Stürm N, Borel AL. Benefit-risk of intraoperative liver biopsy during bariatric surgery: review and perspectives. Surg Obes Relat Dis. 2017;13:1780–6.
Potretzke TA, Saling LJ, Middleton WD, Robinson KA. Bleeding complications after percutaneous liver biopsy: do subcapsular lesions pose a higher risk? AJR Am J Roentgenol. 2018;211:204–10.
Midia M, Odedra D, Shuster A, Midia R, Muir J. Predictors of bleeding complications following percutaneous image-guided liver biopsy: a scoping review. Diagn Interv Radiol. 2019;25:71–80.
Vernuccio F, Rosenberg MD, Meyer M, Choudhury KR, Nelson RC, Marin D. Negative biopsy of focal hepatic lesions: decision tree model for patient management. AJR Am J Roentgenol. 2019;212:677–85.
Kinner S, Reeder SB, Yokoo T. Quantitative imaging biomarkers of NAFLD. Dig Dis Sci. 2016;61:1337–47.
Zhang YN, Fowler KJ, Hamilton G, Cui JY, Sy EZ, Balanay M, et al. Liver fat imaging-a clinical overview of ultrasound, CT, and MR imaging. Br J Radiol. 2018;91:20170959.
Vernuccio F, Cannella R, Bartolotta TV, Galia M, Tang A, Brancatelli G. Advances in liver US, CT, and MRI: Moving toward the future. Eur Radiol Exp. 2021;5:52.
Reeder SB. Emerging quantitative magnetic resonance imaging biomarkers of hepatic steatosis. Hepatology. 2013;58:1877–80.
Idilman IS, Aniktar H, Idilman R, Kabacam G, Savas B, Elhan A, et al. Hepatic steatosis: quantification by proton density fat fraction with MR imaging versus liver biopsy. Radiology. 2013;267:767–75.
Gu J, Liu S, Du S, Zhang Q, Xiao J, Dong Q, et al. Diagnostic value of MRI-PDFF for hepatic steatosis in patients with non-alcoholic fatty liver disease: a meta-analysis. Eur Radiol. 2019;29:3564–73.
Serai SD, Dillman JR, Trout AT. Proton density fat fraction measurements at 1.5- and 3-T hepatic MR imaging: same-day agreement among readers and across two imager manufacturers. Radiology. 2017;284(1):244–54.
Hong CW, Wolfson T, Sy EZ, Schlein AN, Hooker JC, Fazeli Dehkordy S, et al. Optimization of region-of-interest sampling strategies for hepatic MRI proton density fat fraction quantification. J Magn Reson Imaging. 2018;47:988–94.
Hong CW, Cui JY, Batakis D, Xu Y, Wolfson T, Gamst AC, et al. Repeatability and accuracy of various region-of-interest sampling strategies for hepatic MRI proton density fat fraction quantification. Abdom Radiol (NY). 2021;46:3105–16.
Bonekamp S, Tang A, Mashhood A, Wolfson T, Changchien C, Middleton MS, et al. Spatial distribution of MRI-determined hepatic proton density fat fraction in adults with nonalcoholic fatty liver disease. J Magn Reson Imaging. 2014;39:1525–32.
Vu KN, Gilbert G, Chalut M, Chagnon M, Chartrand G, Tang A. MRI-determined liver proton density fat fraction, with MRS validation: comparison of ROIs sampling methods in patients with type 2 diabetes. J Magn Reson Imaging. 2016;43:1090–9.
Sofue K, Mileto A, Dale BM, Zhong X, Bashir MR. Interexamination repeatability and spatial heterogeneity of liver iron and fat quantification using MRI-based multistep adaptive fitting algorithm. J Magn Reson Imaging. 2015;42:1281–90.
Achmad E, Yokoo T, Hamilton G, Heba ER, Hooker JC, Changchien C, et al. Feasibility of and agreement between MR imaging and spectroscopic estimation of hepatic proton density fat fraction in children with known or suspected nonalcoholic fatty liver disease. Abdom Imaging. 2015;40:3084–90.
Li M, Cao D, Liu Y, Jin L, Zeng N, Wang L, et al. Alterations in the liver fat fraction features examined by magnetic resonance imaging following bariatric surgery: a self-controlled observational study. Obes Surg. 2020;30:1917–28.
Hines CDG, Frydrychowicz A, Hamilton G, Tudorascu DL, Vigen KK, Yu H, et al. T(1) independent, T(2) (*) corrected chemical shift based fat-water separation with multi-peak fat spectral modeling is an accurate and precise measure of hepatic steatosis. J Magn Reson Imaging. 2011;33:873–81.
Campo CA, Hernando D, Schubert T, Bookwalter CA, Pay AJV, Reeder SB. Standardized approach for ROI-based measurements of proton density fat fraction and R2* in the liver. AJR Am J Roentgenol. 2017;209:592–603.
Hooker JC, Hamilton G, Park CC, Liao S, Wolfson T, Dehkordy SF, et al. Inter-reader agreement of magnetic resonance imaging proton density fat fraction and its longitudinal change in a clinical trial of adults with nonalcoholic steatohepatitis. Abdom Radiol (NY). 2019;44:482–92.
Tai CM, Yu ML, Tu HP, Huang CK, Hwang JC, Chuang WL. Derivation and validation of a scoring system for predicting nonalcoholic steatohepatitis in Taiwanese patients with severe obesity. Surg Obes Relat Dis. 2017;13:686–92.
Zhou BF. Cooperative meta-analysis group of the working group on obesity in china. predictive values of body mass index and waist circumference for risk factors of certain related diseases in chinese adults--study on optimal cut-off points of body mass index and waist circumference in chinese adults. Biomed Environ Sci. 2002;15:83–96.
WHO Expert Consultation. Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies. Lancet. 2004;363:157–63.
Bedossa P. Diagnosis of non-alcoholic fatty liver disease/non-alcoholic steatohepatitis: why liver biopsy is essential. Liver Int. 2018;38(Suppl 1):64–6.
Johnson BL, Schroeder ME, Wolfson T, Gamst AC, Hamilton G, Shiehmorteza M, et al. Effect of flip angle on the accuracy and repeatability of hepatic proton density fat fraction estimation by complex data-based, T1-independent, T2*-corrected, spectrum-modeled MRI. J Magn Reson Imaging. 2014;39:440–7.
Diedenhofen B, Musch J. cocor: A comprehensive solution for the statistical comparison of correlations. PLoS ONE. 2015;10: e0121945.
Vilgrain V, Ronot M, Abdel-Rehim M, Zappa M, d’Assignies G, Bruno O, et al. Hepatic steatosis: a major trap in liver imaging. Diagn Interv Imaging. 2013;94:713–27.
Dioguardi Burgio M, Bruno O, Agnello F, Torrisi C, Vernuccio F, Cabibbo G, et al. The cheating liver: imaging of focal steatosis and fatty sparing. Expert Rev Gastroenterol Hepatol. 2016;10:671–8.
Yokoo T, Serai SD, Pirasteh A, Bashir MR, Hamilton G, Hernando D, et al. Linearity, bias, and precision of hepatic proton density fat fraction measurements by using mr imaging: a meta-analysis. Radiology. 2018;286:486–98.
Bachtiar V, Kelly MD, Wilman HR, Jacobs J, Newbould R, Kelly CJ, Gyngell ML, et al. Repeatability and reproducibility of multiparametric magnetic resonance imaging of the liver. PLoS ONE. 2019;14: e0214921.
Merat S, Sotoudehmanesh R, Nouraie M, Peikan-Heirati M, Sepanlou SG, Malekzadeh R, et al. Sampling error in histopathology findings of nonalcoholic fatty liver disease: a post mortem liver histology study. Arch Iran Med. 2012;15:418–21.
Adams LA, Sanderson S, Lindor KD, Angulo P. The histological course of nonalcoholic fatty liver disease: a longitudinal study of 103 patients with sequential liver biopsies. J Hepatol. 2005;42:132–8.
Jayakumar S, Middleton MS, Lawitz EJ, et al. Longitudinal correlations Between MRE, MRI-PDFF and liver histology in patients with nonalcoholic steatohepatitis: analysis of data from a phase 2 trial of Selonsertib. J Hepatol. 2018;70(1):133.
Tanaka N, Kimura T, Fujimori N, Nagaya T, Komatsu M, Tanaka E. Current status, problems, and perspectives of non-alcoholic fatty liver disease research. World J Gastroenterol. 2019;25:163–77.
Ampuero J, Aller R, Gallego-Durán R, Crespo J, Abad J, González-Rodríguez Á, et al. Definite and indeterminate nonalcoholic steatohepatitis share similar clinical features and prognosis: A longitudinal study of 1893 biopsy-proven nonalcoholic fatty liver disease subjects. Liver Int. 2021;41:2076–86.
Xian YX, Weng JP. Xu F MAFLD vs NAFLD: shared features and potential changes in epidemiology pathophysiology diagnosis and pharmacotherapy. Chin Med J (Engl). 2020;134:8–19.
Ooi GJ, Clouston A, Johari Y, Kemp WW, Roberts SK, Brown WA, et al. Evaluation of the histological variability of core and wedge biopsies in nonalcoholic fatty liver disease in bariatric surgical patients. Surg Endosc. 2021;35:1210–8. https://doi.org/10.1007/s00464-020-07490-y.
Chaim FDM, Pascoal LB, Chaim FHM, Palma BB, Damázio TA, da Costa LBE, et al. Histological grading evaluation of non-alcoholic fatty liver disease after bariatric surgery: a retrospective and longitudinal observational cohort study. Sci Rep. 2020. https://doi.org/10.1038/s41598-020-65556-2.
Procter AJ, Sun JY, Malcolm PN, Toms AP. Measuring liver fat fraction with complex-based chemical shift MRI: the effect of simplified sampling protocols on accuracy. BMC Med Imaging. 2019;19:14.
Wang K, Mamidipalli A, Retson T, Bahrami N, Hasenstab K, Blansit K, et al. Automated CT and MRI liver segmentation and biometry using a generalized convolutional neural network. Radiol Artif Intell. 2019;1(2):180022.
Martí-Aguado D, Jiménez-Pastor A, Alberich-Bayarri Á, Rodríguez-Ortega A, Alfaro-Cervello C, Mestre-Alagarda C, et al. Automated whole-liver MRI segmentation to assess steatosis and iron quantification in chronic liver disease. Radiology. 2022;302:345–54.
Acknowledgements
Not applicable
Funding
This study was supported by the National Natural Science Foundation of China (No. 61871276, 82071876); the National Key R&D Program of China (No. 2015BAI13B09, 2017YFC0110904); Beijing Natural Science Foundation (No. 7184199); Beijing Municipal Health Commission, Special Program of Scientific Research on health development in Beijing (No. 2018-2-2023); Beijing Hospitals Authority Clinical Medicine Development of Special Funding Support (N0. ZYLX202101); and Capital’s Funds for Health Improvement and Research (N0. 2020-1-2021).
Author information
Authors and Affiliations
Contributions
Guarantor of integrity of the entire study: DC; study concepts and design: DC, ML, YL, DY, HX, HL, JL, ZY; literature research: DC, ML; clinical studies: DC, ML, YL, PZ, ZZ; experimental studies/data analysis: DC, ML, YL, HJ; statistical analysis: DC, ML; manuscript preparation: all authors; manuscript editing: DC. All authors read and approved the final manuscript.
Corresponding authors
Ethics declarations
Ethical approval and consent to participate
The study was approved by the Ethics Committees of the Beijing Friendship Hospital where the study was conducted (NO. 2017-P2-131–02). All methods were carried out in accordance with relevant guidelines and regulations, and written informed consent was obtained from all participants. This study adhered to the tenets of the Declaration of Helsinki.
Consent for publication
Not applicable.
Competing interests
The authors declare that they have no competing interests.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
About this article
Cite this article
Cao, D., Li, M., Liu, Y. et al. Comparison of reader agreement, correlation with liver biopsy, and time-burden sampling strategies for liver proton density fat fraction measured using magnetic resonance imaging in patients with obesity: a secondary cross-sectional study. BMC Med Imaging 22, 92 (2022). https://doi.org/10.1186/s12880-022-00821-6
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s12880-022-00821-6