Keywords
Apparent diffusion coefficient - Intravoxel incoherent motion (IVIM) - spleen - aging - normative value
Introduction
The liver and spleen are the two biggest solid organs in the upper abdomen. Diffusion weighted imaging (DWI) has been applied to these organs, particularly in the oncological setting. For quantification, ADC (apparent diffusion coefficient) is already in widespread use, while IVIM (intravoxel incoherent motion) imaging remains in a research setting. Common malignant lesions of the spleen include metastases, lymphoma and leukemia. Benign lesions of the spleen include cystic lesions, infective lesions, and inflammatory pseudotumor. Though much of the published research has been focused on the liver with the spleen receiving less attention, diffusion imaging of the spleen is still commonly conducted [1]
[2]
[3]. Diffusion measurement from a healthy spleen is also used to normalize the measures of other abdominal organs such as the liver and pancreas [4].
Upper abdominal organ IVIM quantification has been generally considered to suffer from measurement instability [5]. With the goal to improve the stability of liver IVIM parameters, a number of measures have been adopted. For data acquisition, a 16 b-value IVIM protocol has been used [5]. Before image processing, an image data quality assessment is conducted, with images with severe respiratory motion or artifacts discarded [6]. Then, the signal is measured on the liver with an ROI (region-of-interest) based approach. The ROI-based analysis offers better parameter estimation than pixelwise fitting when the signal-to-noise ratio is low [7]. If segmented fitting is applied, a threshold b-value of 60 s/mm2 is chosen [8]. In liver IVIM parameter calculation, b=0 image data is not included for bi-exponential curve fitting [5]
[9]
[10]. The relationship between liver DWI signal and b-value below a diffusion weighting of 1000 s/mm2 does not follow bi-exponential decay; instead, it can be better fitted by an addition of a very fast component with a tri-exponential decay model [11]. However, with image data acquired from routine clinical scanners, the fitting of a tri-exponential decay model can be quite unstable at individual study subject’s levels [11 ]. With fitting starting from a non-zero low b-value, the relationship between DWI signal and b-value better follows a bi-exponential decay pattern [5]
[10]
[12]. Datasets with unacceptable curve fitting are also excluded [6]. With these approaches, good scan-rescan stability for liver IVIM can be achieved. In two studies conducted on 1.5T and 3.0T scanners, scan-rescan repeatability intraclass correlation coefficients (ICC) for IVIM-PF (perfusion fraction) of 0.837 and 0.824, respectively, in the same scan session were achieved [10]
[13]. In one study conducted on a 3.0T scanner, a scan-rescan reproducibility ICC of 0.738 for PF in two different scan sessions was achieved [10]. Based on the approaches, in three clinical studies of medium sample size, a high detection rate for even early-stage liver fibrosis was achieved [5]
[12].
It has been noted that there are age and gender differences in normative values of liver diffusion parameters, with older subjects being associated with a lower ADC and IVIM-Dslow, and older females being associated with lower blood perfusion [8]
[14]
[15]
[16]. The liver and spleen are both large solid organs with rich blood supply, both are susceptible to respiratory motion, and both function as an iron storage organ. With careful imaging data processing as described above, we aim in this study to evaluate potential age and gender differences in normative values of spleen diffusion parameters. Reliable normative values are essential, particularly for the assessment of diffused spleen lesions and when using diffusion measurement from a healthy spleen to normalize measured values of other organs.
Materials and Methods
The MRI data acquisition from healthy volunteers was approved by the institutional ethical committee, and informed consent was obtained for all subjects ([Fig. 1]). All study participants were known to be healthy at the time of the MRI exam and at the 6-month follow-up after the exam, with no liver, spleen, or other abdominal organ disease history, and with no regular medication intake. Dataset 1 was acquired in the period from Apr 22, 2019 to Dec 11, 2019. Study subjects were scanned twice during the same session as long as the study subject was able to tolerate being in the magnet and was able to lie still, with the subjects’ position and selected scan planes remaining unchanged. Imaging used a 1.5T magnet ([Table 1]) The IVIM imaging b-value selection was based on the following considerations: 1) in most clinical scanners, the maximum b-value number allowed is 16; 2) images with b-value > 600 s/mm2 tend to be too noisy; 3) to densely sample images with b-value <10s/mm2 to model the initial fast signal decay, and 4) to relatively densely sample images with b-values around 60 s/mm2 which is the critical point to separate a fast component and a slow component in perfusion-rich organs such as the liver and spleen [8]
[12]. Dataset 2 was acquired during the period from Jan. 9, 2021 to Dec. 6, 2022, using the same scanner and same IVIM protocol, with the goal of sampling more older male subjects. Dataset 3 was acquired during the period from Dec. 1, 2019 to July 23, 2021, using a 3.0T scanner ([Table 1]). For all scans, participants were asked to fast for 6 hours before imaging.
Table 1 Diffusion parameters for datasets 1 and 2.
|
Datasets 1 and 2
|
Values#
|
Median
|
Slope
|
r
|
Pslope
|
Pgender
|
|
# value: mean ± standard deviation (range), 95% CI. slope: trend for age-related change. §: mean age (years). r: Pearson correlation r value for a diffusion parameter and age. Pslope
: P-value for the fitted slope (Pslope < 0.05 suggests the slope of increasing or decreasing along age-axis is significant). Pgende
r: P-value for comparison of results for males vs. results for females. **P < 0.05; *P < 0.1; ¶: unit in 10 –3 mm2/s
|
|
ADC
(2 b-values)
|
Males, 50.41y§
|
1.037 ± 0.194 (0.705–1.710), 95%CI: 0.990–1.083
|
1.011
|
0.0010
|
0.06
|
0.64
|
0.13
|
|
Females, 45.94y§
|
0.968 ± 0.115 (0.725–1.247), 95%CI: 0.928–1.008
|
0.987
|
–0.0032
|
–0.42
|
0.01**
|
|
|
All, 48.93y§
|
1.014 ± 0.175 (0.705–1.710), 95%CI: 0.980–1.048
|
0.995
|
–0.0005
|
–0.04
|
0.72
|
|
|
ADC
(3 b-values)
|
Males, 50.41y§
|
0.975 ± 0.181 (0.670–1.593), 95%CI: 0.932–1.019
|
0.932
|
0.0004
|
0.03
|
0.82
|
0.47
|
|
Females, 45.94y§
|
0.928 ± 0.114 (0.702–1.218), 95%CI: 0.889–0.968
|
0.949
|
–0.0034
|
–0.46
|
0.007**
|
|
|
All, 48.93y§
|
0.960 ± 0.163 (0.670–1.593), 95%CI: 0.928–0.992
|
0.945
|
–0.0010
|
–0.08
|
0.42
|
|
|
Dslow¶
|
Males, 49.48y§
|
0.880 ± 0.186 (0.600–1.468), 95%CI: 0.826–0.934
|
0.852
|
0.0017
|
0.10
|
0.49
|
0.66
|
|
Females, 44.85y§
|
0.865 ± 0.112 (0.653–1.107), 95%CI: 0.813–0.918
|
0.886
|
–0.0031
|
–0.45
|
0.046**
|
|
|
All, 48.12y§
|
0.875 ± 0.167 (0.600–1.468), 95%CI: 0.835–0.916
|
0.867
|
–0.0004
|
–0.03
|
0.79
|
|
|
PF
|
Males, 49.48y§
|
0.105 ± 0.035 (0.055–0.216), 95%CI: 0.095–0.115
|
0.100
|
0.0007
|
0.22
|
0.13
|
<0.0001*
|
|
Females, 44.85y§
|
0.069 ± 0.014 (0.042–0.093), 95%CI: 0.063–0.076
|
0.073
|
0.0004
|
0.49
|
0.03**
|
|
|
All, 48.12y§
|
0.095 ± 0.035 (0.042–0.216), 95%CI: 0.086–0.103
|
0.088
|
0.0008
|
0.28
|
0.02**
|
|
|
Dfast¶
|
Males, 49.48y§
|
121.4 ± 65.4 (7.6–246.2), 95%CI: 102.4–140.3
|
92.5
|
1.0070
|
0.17
|
0.24
|
0.41
|
|
Females, 44.85y§
|
110.9 ± 64.1 (31.2–244.0), 95%CI: 80.9–140.9
|
84.7
|
0.9657
|
0.25
|
0.29
|
|
|
All, 48.12y§
|
118.3 ± 64.7 (7.6–246.2), 95%CI: 102.6–133.9
|
91.1
|
1.022
|
0.21
|
0.09*
|
|
|
DDVD
|
Males, 50.88y§
|
11.93 ± 7.40 (0.56–44.68), 95%CI: 10.10–13.77
|
11.05
|
0.0849
|
0.09
|
0.48
|
0.50
|
|
Females, 45.72y§
|
11.46 ± 8.62 (0.60–39.47), 95%CI: 8.35–14.56
|
9.96
|
–0.1022
|
–0.18
|
0.31
|
|
|
All, 49.18y§
|
11.78 ± 7.78 (0.56–44.68), 95%CI: 10.21–13.34
|
10.55
|
0.0625
|
–0.03
|
0.81
|
|
Fig. 1 Study volunteer inclusion information. Data-1 and Data-2 were scanned at 1.5T and Data-3 were scanned at 3.0T. After excluding images with artifacts or data with insufficient fitting quality, the final analysis includes 69 males (44 had two scans) and 34 females (26 had two scans) for 1.5T ADC, 7 males (5 had two scans) and 15 females (13 had two scans) for 3.0T ADC, 48 males (33 had two scans) and 20 females (15 had two scans) for IVIM (only at 1.5T), and 65 males (37 had two scans) and 32 females (24 had two scans) for DDVD (only at 1.5T). DDVD < 0 was considered an MRI scanner scaling artifact, as tissues in b=0 s/mm2 image should always have a higher signal than in b=2 or 4 s/mm2 image. For the cases with two scans, the mean values of the two scans are presented as the final results.
Image segmentation was conducted with ITK-SNAP (http://www.itksnap.org) and data analysis was implemented in MATLAB (MathWorks, Natick, MA, USA). Blinded to the demographic information of the image data, ROI placement was conducted initially by a radiology trainee, then was checked by a senior radiologist until consensus was achieved. For ADC analysis, ROIs were placed on the b=0 s/mm2 image to cover a large portion of the spleen parenchyma while avoiding large vessels ([Fig. 2]A) and then copied to the images of other b-values (b=60 and/or b=600 s/mm2) of this slice.
Fig. 2 Spleen segmentation and IVIM curve-fitting. a1 b=0 s/mm2 diffusion weighted image. a2 Segmentation of spleen parenchyma. a3 Segmentation of spleen parenchyma. b example of accepted IVIM curve fitting. c example of accepted IVIM curve fitting. d example of accepted IVIM curve fitting. e example of accepted IVIM curve fitting. f example of excluded IVIM curve fitting. g example of excluded IVIM curve fitting.
ADC(b0b600) was calculated according to
where b2 and b1 refer to b=600 and b=0 s/mm2
, respectively, where S(b1) and S(b2) denote the image signal-intensity acquired at the b-factor values of b=0 and b=600 s/mm2, respectively.
ADC(b0b60b600) was calculated according to
Where bi
is the ith b value(unit: s/mm2), S(bi) is the signal intensity at bi
.
For IVIM analysis for Dslow (D), PF (perfusion fracture, f), and Dfast (perfusion related diffusion, D*), ROIs were placed on the b=0 image to cover a large portion of the spleen parenchyma while avoiding large vessels and then copied to the images of other b-values of this slice. IVIM parameters were calculated based on the mean signal intensity of the whole ROI. The signal value at each b-value was normalized by attributing a value of 100 at b=0 s/mm2 (S(b) norm=(S(b)/S0)×100, where S(b) and S(b)norm are the signal and normalized signal at a given b-value, respectively, S0 means the signal at b=0 s/mm2). Segmented fitting was conducted with the threshold b-value of 60 s/mm2. For the bi-compartmental model, the signal attenuation was modelled according to Eq 3:
The measurement of the spleen DDVD (diffusion-derived vessel density) followed recent reports [5]
[17]. The ROI for the spleen parenchyma was segmented on the b=0 s/mm2 image (resulting in ROI area of area0) and then copied onto the b=2 s/mm2 and b=4 s/mm2 images (resulting in ROI areas of area2 and area4, respectively). The DDVD was calculated according to Eq 4–6.
For all acquired MRI data, we performed a data quality assessment prior to diffusion parameter qualification ([Fig. 1]). Images with notable motion and artifacts were initially discarded. For IVIM analysis, the quality of curve fitting was also checked, with only data with good fitting being accepted ([Fig. 2]). For the participants scanned twice, when the two scans for a subject were of good image and fitting quality, the fittings taking the mean of each b-value’s image signal were used, and if one scan was of good quality and one scan was of unacceptable quality, the scan with unacceptable quality was discarded.
For all analyses, the mean of all included slice measurements was then regarded as the value of the examination, with the last step being weighted by the ROI area of each slice. For the total of 124 subjects scanned in datasets 1 and 2, 103 subjects (83.1%) and 97 subjects (78.2%) had image data suitable for ADC and DDVD analysis, respectively, while only 68 subjects (54.8%) had image data suitable for IVIM analysis. All of dataset 3 was suitable for ADC analysis ([Fig. 1]). The total ROI areas (sum of ROI areas of all included slices) were around 60 cm2 for a study subject (supplementary table 1). For statistical analysis, data were processed using GraphPad Prism (San Diego, CA, USA). Comparisons were performed using independent 2 sample t test or Mann-Whitney U test as appropriate, and the tests were all two-sided. The significance of diffusion measures and age was tested with Pearson correlation. A p-value of less than 0.05 was considered statistically significant, > 0.1 not significant, and between 0.05 and 0.1 with a trend of significance. The liver diffusion parameter analysis of dataset 1 has been published [8]. The spleen DDVD(b0b2) results for dataset 1 have been measured and published earlier [17].
Results
The age- and gender-related trends and differences are shown in [Table 1] and [Fig. 3], [Fig. 4], [Fig. 5], [Fig. 6].
Fig. 3 Relationships between age and ADC values, and comparisons between males and females, between 1.5T magnet (pink dots for females and blue dots for males) and 3.0T magnet (black dots), and between ADC(b0b600) and ADC(b0b60b600) values. a females results [ADC(b0b600)]. b males results [ADC(b0b600)]. c males (blue dots) and females (pink dots) 1.5T results together. [ADC(b0b600)]. d 3.0T results young males (blue dots, age < 31 years) and young females (pink dots). [ADC(b0b600)]. e females results. [ADC(b0b60b600) results.]. f males ADC(b0b60b600) results. g males (blue dots) and females (pink dots) 1.5T results together. [ADC(b0b60b600)]. h A comparison of ADC(b0b600) 1.5T results (2b ADC) and ADC(b0b60b600) 1.5T results (3b ADC).
Fig. 4 Relationships between age and Dslow values, and comparison between males and females. a results for females. b results for males. c results for males (blue dots) and females (pink dots) together. d males (blue dots) and females (pink dots) results together.
Fig. 5 Relationships between age and PF/Dfast values, and comparison between males and females. a PF results for females. b PF results for males. c PF results for males (blue dots) and females (pink dots) together. d PF results for males (blue dots) and females (pink dots) together. e Dfast results for females. f Dfast results for males. g Dfast results for males (blue dots) and females (pink dots) together. h Dfast results for males (blue dots) and females (pink dots) together.
Fig. 6 Relationships between age and DDVDmean values, and comparison between males and females. a results for females. b results for males. c results for males (blue dots) and female (pink dots) together. d results for males (blue dots) and female (pink dots) together.
An age-related decrease of ADC was noted for females ([Fig. 3]A, E), while for males such a trend was not noted ([Fig. 3]B, F). A very high level of heterogeneity was noted for the data for males, with the highest ADC(b0b600) value being 1.710 × 103mm2/s and the lowest value being 0.705 × 10–3mm2/s ([Fig. 3]B). A male-female data comparison did not show a statistically significant difference between the median values ([Table 1]). ADC values > 1.3 × 10–3mm2/s were only seen among males. There did not appear to be a notable difference between 1.5T ADC data and 3.0T ADC data ([Fig. 3]A, B, E, F). ADC(b0b60b600) values were systematically lower than ADC(b0b600) values (p < 0.0001 both for the data for males and the data for females, [Fig. 3]H).
An age-related decrease in Dslow was noted for females ([Fig. 4]A) but not for males ([Fig. 4]B). A very high level of heterogeneity was noted for Dslow for males, with the highest value being 1.468 × 10–3mm2/s and the lowest value being 0.600 × 10–3mm2/s. With available data, a male-female comparison did not show a statistically significant difference between the median values of Dslow ([Fig. 3]D). However, Dslow > 1.2 × 10–3mm2/s was only seen among males.
A trend of an age-related increase is noted for both PF and Dfast, and both for males and females ([Fig. 5]A, B, E, F). Statistical significance for such a trend was achieved for PF values for females (p=0.03) . A trend of marginal significance was achieved for Dfast data for males and females grouped together (p=0.09).
PF values were on average higher among males than females ([Fig. 5]C, D, P < 0.0001). However, no difference was noted for Dfast between males and females ([Fig. 5]G, H).
DDVDmean did not show an age-related trend both for females and males ([Fig. 6]A, B). No notable difference was noted in DDVDmean value between males and females ([Fig. 6]C, D).
Discussion
Iron level and the associated T2* value in the organs are well known to influence an organ’s diffusion measurements [8]
[15]
[16]
[18]. Both the liver and the pancreas show an age-related increase in iron deposition and a decrease in ADC [8]
[15]
[19]. An age-related increase in iron deposition in the spleen has also been well documented [16]
[20]. Due to women’s menstruation and pregnancy, iron deposition in the liver and spleen are lower in adult pre-menopausal women than in age-matched adult men. In women, iron deposition in the liver and spleen substantially increases after menopause [20]
[21]. However, a few earlier studies that attempted to investigate the potential age and gender-related changes for spleen diffusion MRI parameters reported conflicting results. Lavdas et al. [22] studied 51 healthy volunteers (mean age: 38 years; age range: 23–68 years) with a 1.5T scanner and a DWI protocol with b = 0, 150, 400, 750, and 1000 s/mm2. They noted that no significant differences for the spleen were found between males and females, and there were no significant correlations between the ADCs and age. Nazarlou and Abdolmohammadi [23] studied 69 patients (males=29, females =40, without notable spleen abnormalities) with a 1.5T scanner and a DWI protocol with b-values of 50, 400, and 800 s/mm2. The mean age was 52.5 years (age range: 11–84 years) for males and 46.8 years (age range 10–85 years) for females. They concluded that no differences were observed in ADC values of the spleen among the female and male participants or those from various ages. On the other hand, Li et al.
[24] studied 127 patients (age range: 10–79 years, mean age: 44.4 years, without notable spleen abnormalities) with a 1.5T scanner and a DWI protocol with b-values of 0, 800 s/mm2. They reported an age-related decrease in ADC values for older patients. However, Li et al. did not separately analyze male and female data. Chen et al.
[25] studied 1243 patients (age range: 18–91 years, mean age: 57 years, without notable spleen abnormalities) using 3.0T scanners and a DWI protocol with b-values of 50, 800 s/mm2. They reported that the spleen ADC values for the males were lower than those for the females, and the spleen ADC values increased with patient age.
With a relatively large sample size and meticulous care taken for data analysis, our study aims to clarify earlier conflicting reports. We hypothesized that for females there is an age-related decrease of ADC and Dslow for the spleen, which was confirmed in this study. However, despite the relatively large sample size, we did not see an apparent trend of age-related change for ADC and Dslow among males. On the contrary, very high ADC and Dslow values were seen in male subjects aged > 45 years. Despite the fact that the same criteria and the same care were taken to measure the data for females and males, the female results were relatively ‘clean’ and consistent with our expectations, and the results for the male spleen were very heterogeneous. We double-checked the images and the curve-fitting patterns. There was no reason for us to exclude the ‘extreme’ values in the results of males. Moreover, the two scans for each subject, when included for analysis, all showed the same pattern. Schwenzer et al.
[16] reported a negative and statistically significant correlation between age and spleen T2* in healthy females (r=–0.44, P < 0.0001), and a minimally negative and statistically non-significant correlation between age and spleen T2* in healthy males (r=–0.13, P > 0.05). There was a much greater T2* variation in the spleen than in the liver, with spleen T2* varying between 14.4–113.6 ms (mean 48.3 ms) for females and 15.8–69.0 (mean: 36.1) for males, liver T2* varying between 14.7–45.96 ms (mean 29.6 ms) for females and 13.6–43.1 (mean: 25.4) for males. Our results can help to explain the results of Lavdas et al. and those of Nazarlou and Abdolmohammadi. It is more difficult to acquire satisfactory diffusion data fitting in the spleen than in the liver. For the liver, we commonly can have satisfactory IVIM diffusion data fitting for 80–85% of the scanned cases [8]
[12]
[13], but for the spleen satisfactory IVIM diffusion data fitting was only achieved for 54.8% of the scanned cases in this study. Compared with the liver, the spleen is substantially smaller in volume, so that fewer slices and a smaller ROI were available for signal averaging. The data fitting for ADC can also be subject to instabilities [26]. Lavdas et al. observed age-related reduction of ADC for the liver, but not for the spleen. The quality of spleen diffusion measurements can be hampered both by the difficulties in data fitting and also by their relatively small sample size with 27 females most of whom were pre-menopausal or para-menopausal [22]. For ADC, we excluded 17% of the scanned cases which were considered to be of insufficient quality. Chen et al.
[25] reported spleen ADC values increased with the age of patients, which differs from the results of Li et al. and also differs from our female results. It was not described whether the studies of Lavdas et al. [22], Nazarlou and Abdolmohammadi [23], Li et al.
[24], and Chen et al. [25] excluded data of insufficient quality. Note that the studies of Nazarlou and Abdolmohammadi, Li et al. and Chen et al. were all on patients. It is unknown how and whether they excluded data regarding metabolic diseases. Patients with more subtle alterations of spleen diffusion and perfusion might not have been totally excluded, for example. The data of Nazarlou and Abdolmohammadi are also likely affected by extreme values [23].
With three b-values to fit the results, ADC(b0b60b600) can be potentially more stable (i.e., with higher scan-rescan reproducibility) than ADC(b0bb600). While ADC(b0b60b600) values were systematically lower than ADC(b0bb600) values, which can be explained by the fast diffusion (perfusion) effect as shown in Supplementary Fig. 1, ADC(b0b60b600) and ADC (b0bb600) show similar age- and gender-related trends. With limited data, this study did not show a notable difference between 3.0T and 1.5T for the spleen ADC value ([Fig. 3]).
Our study is the first to investigate age and gender differences of normative values of spleen IVIM parameters. Consistent with earlier results for the liver with standard IVIM modeling [4], there were trends of an age-related increase in the spleen of older subjects both for observed PF and observed Dfast. It was reported earlier that, with standard IVIM modeling, a decrease in liver T2* value leads to an increase in observed PF, and a decrease in observed Dslow [18]. On the other hand, an increase in liver T2 value will lead to a decrease in observed PF [27]. Thus, IVIM observed PF and Dfast are not interpreted as a true physiological values in liver perfusion. Instead, they should be considered as ‘composite’ biomarkers that are still clinically meaningful [5]
[12]
[28]
[29]. The same as the results for ADC and Dslow, a tendency is noted for the PF in the male spleen to be more heterogeneous than the PF in the female spleen. The spleen PFs in males were on average higher than the spleen PFs in females. This can be partially explained by higher iron content among males [16]
[18]. IVIM observed Dslow and PF are known to be negatively correlated [8]
[28]
[29]. Dfast values were heterogeneous for both males and females, which is a known feature of Dfast due to its instability in data fitting [5]
[11].
ADC and IVIM values have been tested for quantitative evaluation of spleen pathologies. For example, Jang et al.
[1] described that the lesion-to-parenchyma ADC ratios were significantly different between malignant lesions and benign lesions. Bian et al.
[2] studied spleen DWI for acute leukemia patients with splenomegaly, acute leukemia patients with normal spleen volume, and healthy controls. IVIM parameters were all significantly different among the three groups. Dslow was correlated with white blood cell counts, lactate dehydrogenase, and bone marrow blasts [2]. Klasen et al.
[3] reported that compared with controls, patients with cirrhosis and portal hypertension had significantly higher spleen ADCs. There was a statistically significant correlation between Child–Pugh grade and spleen ADC. After transjugular intrahepatic portosystemic shunt implantation, a reduction in spleen ADC values was noted [3]. Note that there have been many efforts to use spleen diffusion parameters as a reference measurement to normalize the data for other abdominal organs [4]. The results in this study suggest caution should be taken for such an approach, particularly for men.
On DWI, blood vessels (including micro-vessels) show a high signal when there is no diffusion gradient (b=0 s/mm2), while they show a low signal even when very low b-values (such as b=2 or 4 s/mm2) are applied [9]. DDVD can be interpreted as a physiological surrogate of the area of micro-vessels per unit tissue area, which can be conceptually converted to a surrogate of the volume of micro-vessels per tissue unit volume if multiple slices are integrated. Zheng et al.
[17] described a decreased spleen DDVD in viral hepatitis-b liver fibrosis patients. Earlier a relatively ‘clean’ decrease in liver DDVD value for older females was observed [8]. This is in agreement with the known physiological age-dependent reduction in liver blood flow. However, an earlier report did not see such a trend both for the female spleen DDVD (n=35) and the male spleen DDVD (n=32) [17]. In this study, we increased the sample size for males to n=65 and re-measured DDVDmean which is the average of DDVD(b0b2) and DDVD(b0b4) which in theory increases the stability of DDVD values. However, spleen DDVDmean still did not show an age-related change. In addition, DDVD values were more heterogeneous in the spleen than in the liver ([Fig. 2] in [17], [Fig. 3] in [8] and [Fig. 6] in this study). It is possible that the much wider variation in iron concentration in the spleen than in the liver [16] may affect the spleen DDVD value. Although physiologically, the per unit volume blood perfusion to the spleen is not less than that to the liver, IVIM measured PF is only half of that of the liver [30], and this is considered at least partially due to the higher spleen T2 relative to liver T2 [27]. Due to the much smaller PF quantified with standard IVIM [30], we included b=0 data for the IVIM curve fitting in this study.
There are many limitations to this study. We could not explore the biophysical mechanisms behind the observed trends or variation, but we think iron concentration and T2* may explain parts of the trends. Spleen iron concentration does not change a lot in male adults (or can increase slightly in older males), while spleen iron concentration increases substantially in females after menopause [16]
[21] and this may cause the decrease in ADC and Dslow observed in older females. It can be argued that the sample size used in this study was small for females. However, we expect that while more samples may lead to some of the trends becoming statistically more significant, it is unlikely that the trend directions will be altered. The data were mainly collected for 1.5T, with only limited ADC data for 3.0T. Another limitation is that we do not have data for pediatric and adolescent populations where large changes can be expected. This is a healthy subject study, and we did not provide any pathologic data under the same conditions. Though we did not provide intra- and inter-observer variability data, we do not expect this will be an issue as the ROI placement for the normal spleen is reasonably straightforward.
To summarize, this study shows an age-related decrease in spleen ADC and Dslow for older females. ADC and Dslow show high heterogeneity for males. The high heterogeneity for ADC and Dslow in males suggests that it is difficult to define what values are abnormally elevated. However, the lowest normal value of ADC or Dslow appeared to be consistent among males and females, i.e., higher than 0.6 × 10–3 mm/s in this study. This study shows the normative value of PF is higher among men than among women. Due to the high instability, Dfast is not commonly used for disease assessment, and this study supports this practice for the spleen.
Clinical Relevance
-
Interpretation of normal spleen ADC and IVIM parameters should take age and gender factors into consideration.
-
The high heterogeneity for ADC and IVIM-Dslow in males suggests that it is difficult to define what values are abnormally elevated.
-
IVIM modelled spleen perfusion fraction and Dfast demonstrate an artificial trend of an age-related increase for older subjects.
Data statement
Raw data can be obtained by external researchers for analysis by contacting the corresponding author of this article.
