Zhijun Li1, Prasanna Karunanayaka1, Matthew Muller2, Christopher Sica1, Jian-Li Wang1, Lawrence Sinoway2, and Qing X. Yang1,3
1Center for NMR Research, Department of Radiology, College of Medicine, The Pennsylvania State University, Hershey, PA, United States, 2Heart and Vascular Institute, College of Medicine, The Pennsylvania State University, Hershey, PA, United States, 3Department of Neurosurgery, College of Medicine, The Pennsylvania State University, Hershey, PA, United States
Synopsis
Unlike in human brain
imaging, normalization to a common template during
exercising is a difficult proposition in muscle-imaging studies. Still, motion
artifact has been an issue for dynamic analysis of exercise paradigm. We used individual Independent Component
Analysis (ICA) to identify the “motion component” during exercise (rhythmic
plantar-flexion) and anatomical and temporal features of BOLD signal. We
simultaneously identified the lower leg muscle groups and their common
hemodynamic behaviors under a low-level exercise paradigm and
revealed an intriguing hemodynamic respond characteristic with a prominent
transient increase and followed by a negative BOLD signal sustained to the end
of exercise.Introduction
There
has been great interest in the application of the BOLD effect in studying the
physiology of skeletal muscle [1]. Compare to the BOLD signal in the brain,
however, the muscle hemodynamics during exercise is more complicated because it
is controlled by both local and systemic factors such as the sympathetic
nervous system (SNS) [2]. In order the separate the local vascular effect of
BOLD as observed in the brain from systemic cardiovascular effects during
exercise, this study is designed to elucidate BOLD signal characteristics
during low-level exercise with minimal involvement of the SNS. Motion artifact is one of the obstacles for
functional studies of skeletal muscle. We implemented individual Independent
Component Analysis (ICA), which allowed us to simultaneously identify the
muscle groups and their hemodynamic behaviors, while teasing out the “motion
component” during an exercise paradigm (single leg dynamic plantar-flexion). Our
results may help future studies.
Methods
The
subjects included 9 healthy
adults (mean age = 51 years, 1 female). Using a Siemens 3T scanner (Magnetom
Trio), gradient echo EPI of the calf muscles during exercise were acquired with
the following parameters: voxel size = 2.5 x 2.5 mm, number of slices = 10,
slice thickness = 5.0 mm,
FOV= 160 mm
2, flip angle = 70 deg, bandwidth = 2112 Hz/Px,
TE = 25 ms
and TR = 3 s.
The exercise paradigm consisted of a 1-min
baseline, a 14-min
dynamic plantar-flexion exercise and a 5-min
recovery. Using an MRI compatible, custom made exercise machine, all subjects
performed plantar-flexion with a 2-kg load at a pace of 1/3 Hz. To reduce the
motion artifact, the exercise was performed only during the 2211-ms interval
when no image acquisition was performed within each TR. MRI data were first
motion-corrected before preprocessed using SPM8 (The Welcome Trust Centre for
Neuroimaging, University College London, UK). The single subject ICA analysis
was performed according to the methods outlined elsewhere [3, 4].
Results
Fig. 1 shows representative “motion component” from two
subjects. For both subjects, it’s seen located on the surface of the leg with
its temporal pattern more correlated with the exercise paradigm and distributed
in low frequency region of the corresponding power spectrum. Accordingly, Fig. 2 shows the 3 muscle
groups identified by ICA from one subject and their corresponding time-course
of BOLD signal. We validated the selected IC with ROI analysis by manually segmenting
the three muscle groups and their BOLD signal time-courses are shown on the
right column.
Discussion
All three ICA components captured the muscle
groups through their hemodynamic characteristics during the exercise paradigm. ICA requires no prior physiology or anatomy knowledge,
and was able to differentiate and identify different muscles based on their
unique hemodynamic characteristics during exercise. This data driven method can
overcome the problem of subject dependent variations in utilization of muscle
groups, muscle strength and conditions. Our
results support the notion of an intimate causal connection between the
hemodynamic characteristic and muscle physiological conditions during exercise [5]. As such, the current
study demonstrates a novel and straight forward methodology to investigate
muscle exercise physiology and related diseases such as peripheral arterial disease (PAD).
As seen in Fig. 2, the BOLD signals in the three
muscle groups exhibited a similar character: a rapid initial rise (usually within
30 seconds since onset) and followed by a drop to below the baseline level
(negative BOLD) that usually sustained to the end of the exercise. This negative BOLD effect in the muscle is interesting
as it indicates that a higher oxygen extraction than at rest by the skeletal muscle during steady state exercise at a level that
systemic sympathetic activities are not necessary as indicated by the blood
pressure and heart rate. Concurrent studies performed on a separate day in
these same subjects showed that popliteal artery blood flow increases with this
exercise paradigm (~ 100% increase above baseline), consistent with previously
published studies [6]. In addition, a transient reactive hyperemia occurred
rapidly as indicated by the initial rise of the BOLD signal, which requires
further investigation.
Conclusion
Single subject
ICA allowed for automatically segmenting muscle groups with respect to the
matching hemodynamic characteristics while removing motion artifacts. Application
of the single-subject ICA, we revealed an intriguing hemodynamic respond
characteristic with a prominent initial transient increase and followed by a sustained
negative BOLD signal. With further investigations on assessing its feasibility
in differentiating BOLD behavior in healthy and diseased subjects, it can be a
valuable tool for clinical applications.
Acknowledgements
Funding has been partially provided by NIH P01
HL096570.References
[1] Jacobi et al. J Magn Reson Imaging. 2012;
35(6):1253-65, [2] Joyner et al. Physiol Rev. 2015; 95(2):549-601 [3] Calhoun, et al. Hum. Brain
Mapp. 14(2001) 140-51, [4] Karunanayaka et al. HBM (2013), [5] Wasserman et al. Am
J Physiol. 1994; 266:E519–E539, [6] Villar et al. Physiol Meas. 2013; 34(3):291-306