Evaluation of skeletal muscle microvascular perfusion of lower extremities by cardiovascular magnetic resonance arterial spin labeling, blood oxygenation level-dependent, and intravoxel incoherent motion techniques

Between February 2016 and October 2017, three groups of subjects were enrolled: 1) healthy young subjects (n = 15); 2) PAD patients (n = 14); and 3) age-matched healthy old subjects (n = 10). None of the healthy subjects showed clinical evidence of PAD, cardiac insufficiency, hypoxic pulmonary diseases, or lower extremity venous disorders. Each of these subjects had a normal peripheral pulse status and an ABI > 0.90, and they were all non-smokers. The healthy old group was age-matched to the PAD group to eliminate the confounding effect of age on CMR perfusion parameter measurements. Patients with PAD were recruited from the department of vascular surgery with symptoms of intermittent claudication, rest pain, or critical limb ischemia and an ABI ≤ 0.90. The PAD group was then stratified into two disease severity subgroups based on ABI: 1) the mild-to-moderate group corresponded to an ABI of 0.51 to 0.90 (n = 7); and 2) the severe group corresponded to an ABI ≤ 0.50 (n = 7). Details for each group are listed in Table 1.

In preparation for the scans, all subjects were asked to refrain from alcohol, caffeine, and vigorous exercise 12 h before imaging. The healthy young and healthy old subjects were subjected to a cuff-induced arterial occlusion experiment during the CMR imaging scans to test the sensitivities of ASL, BOLD, and IVIM to pressure variations. Ischemia via arterial occlusion was induced in the lower extremity by a sphygmomanometer cuff tied around the middle of one thigh. The contralateral lower extremity without intervention was imaged simultaneously as the control side. ASL, BOLD, and IVIM were conducted four times in the following order: baseline (Pre), cuff compression with a pressure of 20 mmHg above systolic pressure (Cuff-20), cuff compression with a pressure of 40 mmHg above systolic pressure (Cuff-40), and recovery period (Post). Two different cuff pressures were used to verify whether different air pressures could modulate the perfusion signal intensity and thus provoke different degrees of ischemia. The pressure of 40 mmHg above systolic pressure was chosen in accordance with the previously reported range of 30–50 mmHg above systolic pressure, which is recommended for provoking complete and reproducible ischemia [15]. During the Cuff-20 and Cuff-40 sessions, the cuff was kept inflated until all scans were completed (6 min 20 s). The last three sessions were performed at 30-min intervals to avoid the interference from the preceding session (Fig. 1). To test the interscan reproducibility, 5 healthy subjects were subjected to a second scan within 1 week. Patients with PAD received baseline examinations only at rest. All subjects were asked to lie still in the supine position for approximately 15 min before the onset of CMR imaging to ensure that their legs were at heart height [16], and remain still during the entire examination.

All the CMR measurements were carried out on a 3-T CMR system (HDxt, General Electric Healthcare, Waukesha, Wisconsin, USA) with an eight-channel cardiac coil. The subjects were investigated in the supine position. Prior to acquisition of the CMR perfusion images, axial three-dimensional (3D) spoiled gradient-recalled echo T1-weighted images (repetition time ms/echo time ms 4.1/1.5; flip angle = 12°; matrix, 320 × 320; FOV, 32 × 32 mm²; slice thickness = 5 mm; number of slices = 12) were acquired for use as anatomical landmarks. A FOV of 32 × 32 mm² could cover both legs without leaving anything outside, thus prohibiting wrapping artifacts [17]. Pseudocontinuous ASL was performed using an interleaved 3D stack of spiral fast spin-echo sequence with background suppression. Each spiral arm included 512 sampling points in k-space and a total of 6 spiral arms were acquired. Background suppression was achieved via 5 inversion pulses placed 0, 1465, 2100, 2600, and 2880 ms after the labeling start point to suppress a broad range of T1 values [18]. Other ASL parameters were as follows: repetition time ms/echo time ms = 4316/9.4; bandwidth = 62.5 kHz; FOV = 32 × 32 mm²; slice thickness = 5 mm; number of slices = 12; number of averages = 2; post-labeling delay time = 1525 ms. A post-labeling delay time of 1525 ms before imaging was used to allow the blood to perfuse all muscle groups, which was similar to that used in a previous study [19]. BOLD was performed using a multi-echo gradient-recalled echo sequence implementing the following parameters: repetition time ms/echo times ms 875/(2.5, 6.7, 10.9, 15.1, 19.3, 23.5, 27.7, 31.9, 36.1, 40.3, 44.5, 48.7, 52.9, 57.1, 61.3, and 65.5); matrix = 256 × 128; FOV = 32 × 32 mm²; slice thickness/gap = 5/0 mm; number of slices = 12. IVIM imaging was performed using a single shot spin-echo echo-planar imaging sequence at 9 b-values (0, 20, 50, 100, 150, 200, 300, 500, and 800 s/mm²) in three orthogonal gradient directions, with the following parameters: repetition time ms/echo time ms = 2800/70; matrix = 192 × 192; FOV = 32 × 32 mm²; slice thickness/gap = 5/0 mm; number of slices = 12; number of averages = 3; parallel imaging factor = two. A standard monopolar Stejskal-Tanner diffusion encoding scheme was applied with diffusion gradient pulse duration of 16 ms. The acquisition times for ASL, BOLD, and IVIM were 2 min 18 s, 1 min 59 s, and 2 min 3 s, respectively. All imaging was performed in the axial plane at the level of the middle calf.

ASL, BOLD, and IVIM data were post-processed on a pixel-by-pixel basis on a workstation (Advantage Workstation 4.5; General Electric Healthcare) to obtain corresponding parametric maps of blood flow, T2* relaxation time, perfusion fraction f, diffusion coefficient D, and pseudodiffusion coefficient D*.

Pseudocontinuous ASL perfusion was calculated using a one-compartment model for blood after subtracting the tagged images from the nontagged control images. Blood flow was quantified using the following equation:

where PLD is the post-labeling delay (1525 ms), λ is the tissue partition coefficient (0.9 ml/g) [19], T_{1, blood} is the longitudinal relaxation time of blood (1600 ms) [19], α is the labeling efficiency (0.80 × 0.75) (label pseudocontinuous ASL × background suppression) [18], τ is the labeling duration (1450 ms), SI _PD is the proton density reference without labeling or background saturation, and SI _control and SI _label are the control and tagged signals, respectively. The arterial transit time was ignored in the quantification.

T2* relaxation time was calculated from the multi-echo T2* gradient-recalled echo data using the least-square fit of monoexponential decay [20], according to the following equation: S(TE) = S_TE0·exp.(−TE/T2*), where TE is the gradient echo time and S(TE) and S_TE0 are the measured signal intensities for TE ≠ 0 and TE = 0, respectively.

The multi-b-value diffusion-weighted images were analyzed using the IVIM model according to the following equation: S(b) = S_b0·[f·exp.(−bD*) + (1-f) ·exp.(−bD)], where S(b) is the measured signal intensity obtained with a nonzero b-value and S_b0 is the measured signal intensity for b = 0. With this equation, perfusion fraction f together with diffusion coefficient D and pseudodiffusion coefficient D* were calculated using a nonlinear biexponential fit based on the Levenberg-Marquardt technique [21].

For BOLD and IVIM analysis, goodness of fit was assessed using R² = 1 − ESS/TSS, where ESS is the sum of squared errors between the data points and the fitting curve and TSS is the sum of squared differences between the data points and the mean value of all data points. Additional computations were performed to assess the signal-to-noise ratios (SNRs) of the images obtained with TE of 65.5 ms for BOLD and b-value of 800 s/mm² for IVIM. Noisy images were excluded from curve fitting.

Regions of interest (ROIs) were manually drawn on T1-weighted images around the 4 muscle groups (anterior, lateral, soleus, and gastrocnemius) at the largest cross-sectional area of the calf on both the experimental and control sides (Fig. 1). Attention was given to exclude areas influenced by bones and large vessels, and the inter-osseous muscle was not investigated because it contains a relatively large number of vessels [22]. ROIs were then copied and pasted into the corresponding functional perfusion maps. Average values within ROIs were recorded. Independent analysis of perfusion maps (from 7 randomly selected healthy subjects) by 2 radiologists blinded to the clinical outcomes was conducted to evaluate the interreader reproducibility. In addition, the repeat scans of 5 healthy subjects were analyzed by the same blinded radiologist to test the interscan reproducibility. Normalized values were obtained by dividing each imaging parameter value obtained under the experimental statuses (Cuff-20/Cuff-40/Post) by the baseline measurement (Pre).

TcPO2 measurements were acquired in all healthy young subjects the day after the CMR examinations using a TcPO2 monitoring system (Periflux System 5000; Perimed, Jarfalla, Sweden) in an air-conditioned room maintained at 22 °C. The cuff-induced ischemia paradigm followed the same process as that described for the CMR experiments. One author (LZ, 12 years of experience in TcPO2 measurement) placed the electrode at the same spot at which the CMR measurements were taken (i.e., the medial upper third of the calf adjacent to the gastrocnemius muscle at the maximal calf diameter) [7].

All data were expressed as the median (range) owing to non-normal data distributions. To assess interreader and interscan reproducibility, the intraclass correlation coefficient (ICC) was calculated. ICC values less than 0.40 indicated poor reproducibility, those ranging from 0.40 to 0.75 indicated fair to good reproducibility, and those higher than 0.75 indicated excellent reproducibility.

To determine whether differences existed in each imaging parameter value under different statuses (Pre/Cuff-20/Cuff-40/Post) in the experimental and control lower extremities, the Friedman test was used. In cases of statistical significance, further pairwise comparisons with the Dunn test were performed. The Wilcoxon signed rank test was used to compare imaging parameter measurements between left and right lower extremities in all healthy subjects under different statuses. The non-parametric Spearman rank correlation test was performed to assess correlations between imaging parameters derived from different methods, as well as between imaging parameters in gastrocnemius and TcPO2 measurements. Strength of correlation based on the Spearman rank correlation coefficient (ρ) was interpreted as follows: 0.00 to 0.20, very weak to negligible correlation; 0.21 to 0.40, weak correlation; 0.41 to 0.70, moderate correlation; 0.71 to 0.90, strong correlation; and 0.91 to 1.00, very strong correlation [23]. The effects of age on the baseline imaging parameter measurements were investigated by comparing the values between the healthy young and healthy old groups with the Mann-Whitney U test. The Wilcoxon signed rank test was used for comparisons of imaging parameter measurements between the left and right lower extremities in PAD patients. Then, imaging parameters were compared between PAD patients and healthy old subjects, as well as between mild-to-moderate and severe PAD patients using the Mann-Whitney U test. Finally, correlations between imaging parameters and ABI in PAD patients were assessed using the Spearman rank correlation test. Bonferroni correction for multiple comparisons for the number of muscle groups was applied when necessary.

The sample size of patients included in this study was estimated using one-sided calculations with α of 0.05 and a power of 80% to detect an absolute T2* decrease of 4 ms (compared with normal) with a standard deviation of 2 ms based on the results of previous studies [24, 25]. Assuming a 20% dropout rate, it was determined that 10 participants were required.

Changes in quantitative imaging parameters under the cuff compression paradigm in the healthy young and healthy old groups are illustrated in Figs. 2 and 3 for the experimental lower extremity and in Additional file 2: Figure S2 and Additional file 3: Figure S3 for the control side.

In the healthy young group on the experimental side, only blood flow and T2* values showed significant differences among the 4 statuses in all muscle groups (all P < 0.01 for blood flow, and all P < 0.001 for T2*). f and D values showed significant differences in the lateral, soleus, and gastrocnemius muscle groups, and D* values were significantly different in the soleus and gastrocnemius muscle groups (all P < 0.05). Under the Cuff-20/Cuff-40 compression statuses, T2* values were significantly lower than the baseline measurements in all muscle groups (all P < 0.05), and blood flow, f, D, and D* values showed marked differences in only some muscle groups. No significant differences for any parameter in either muscle group were observed between the Post and Pre statuses (all P > 0.05) (Fig. 2).

In the healthy old group on the experimental side, only T2*, f, and D values showed significant differences among the 4 statuses in all muscle groups (all P < 0.001 for T2*, and all P < 0.01 for f and D). D* values showed significant differences in the soleus and gastrocnemius muscle groups (both P < 0.01). Blood flow values did not significantly differ in any muscle group (all P > 0.05). Under the Cuff-20/Cuff-40 compression statuses, significant differences were found for T2* values compared with those at baseline in all muscle groups except for the anterior muscle group under the 20-mmHg status (all P < 0.05). Furthermore, f, D, and D* values were significantly different in several muscle groups under these statuses (all P < 0.05). Additionally, no significant differences between Pre and Post measurements were observed for any of the parameters (all P > 0.05) (Fig. 3).

In both the healthy young and healthy old groups on the experimental side, in the cases of significant differences, blood flow, T2*, and D values gradually decreased under the Cuff-20 and Cuff-40 compression statuses, while f values showed a tendency to increase. During the recovery period, all parameters nearly returned to normal (Figs. 2 and 3).

In both the healthy young and healthy old groups on the control side, among all parameters, only T2* values in some muscle groups showed significant differences among the 4 statuses (P < 0.05) (Additional file 2: Figure S2 and Additional file 3: Figure S3).

Results of Wilcoxon signed rank test comparing functional imaging parameters between the experimental and control sides under different statuses are illustrated in Table 2. No significant differences in any parameter were observed between the left and right lower extremities in healthy subjects at rest (all P > 0.0125). Most muscle groups exhibited significant differences in perfusion-related parameter values especially when the cuff compression pressure was increased to 40 mmHg above systolic pressure.

No significant correlations between functional imaging parameters derived from the different methods were observed (all P > 0.05) (Additional file 4: Table S1).

The normalized CMR imaging parameters blood flow, T2*, and f were all correlated with normalized TcPO2 measurements (Fig. 4). Blood flow and T2* showed significant moderate correlations with TcPO2 measurements (ρ = 0.465 (P = 0.001) and ρ = 0.522 (P < 0.001), respectively). A significant negative correlation was observed between f and TcPO2 measurements (P = 0.018), although the correlation was weak (ρ = − 0.351). No significant correlation was found for D (P = 0.054) or D* values (P = 0.340).

Within the muscle groups studied, blood flow for all muscle groups, f for the soleus and gastrocnemius groups, and D and D* for the gastrocnemius group demonstrated significant differences between the healthy young and healthy old groups at rest (all P < 0.0125). Only T2* was found to be independent of age in every muscle group (all P > 0.0125) (Table 3).

In PAD patients, T2* was markedly reduced on the affected side compared with that on the contralateral side; significance was reached in the anterior muscle group (P = 0.005), and no significance was reached in the lateral, soleus, or gastrocnemius muscle groups (all P > 0.0125) (Additional file 5: Table S2).

Results of Mann-Whitney U tests comparing functional perfusion parameters between PAD patients and age-matched healthy subjects and between mild-to-moderate and severe PAD patients at rest are illustrated in Table 4. Among all parameters, T2* showed the best performance for the discrimination, with significantly reduced values observed in PAD patients compared with age-matched healthy old subjects and severe PAD patients compared with mild-to-moderate patients in all muscle groups (all P < 0.0125). No significant differences in other parameters in any muscle group were observed except D* values between PAD patients and healthy old subjects in the lateral muscle group.

Spearman rank correlation analysis showed that in PAD patients, T2* was significantly correlated with ABI in the anterior (ρ = 0.837; P < 0.001), lateral (ρ = 0.820; P < 0.001), soleus (ρ = 0.785; P = 0.001), and gastrocnemius (ρ = 0.644; P = 0.012) muscle groups (Fig. 5), whereas no significant correlation was observed for the other parameters (all P > 0.05).