The way a person moves is as unique as their fingerprints or iris pattern. Researchers have tracked muscle movements of study participants during exercise, gait and pedaling, and believe these differences may go deeper than individual movement styles. By identifying a person’s unique movement style could help identify future musculoskeletal problems, and assist in better tailoring treatments to the patient’s specific needs.
In this study, reported in the Oct. 2019 Journal of Applied Physiology, researchers analyzed movement patterns of 53 individuals, using surface electromyography (EMG) on their legs as they pedaled on a stationary bicycle and walked on a treadmill. By using a machine learning protocol, they were able to tracked activation patterns from 8 muscles of the right leg: the vastus lateralis (VL), rectus femoris (RF), vastus medialis (VM), gastrocnemius lateralis (GL), gastrocnemius medialis (GM), soleus (SOL), tibialis anterior (TA), and biceps femoris-long head (BF). They used the data gleaned from this study to establish unique muscle activation signatures recorded during an initial session. Participants then returned for a second round of the same physical activities between 1 and 41 days after the first (average, 13 days), allowing researchers to evaluate the similarities between activation patterns observed at each session.
According to the researchers, the participants were in good health, the majority male (77 percent) with an average age of 23.1 years and average BMI of 23.2 for males and 21 for females.
Taking a Look at the Study’s Findings
The study’s findings indicated that a “substantial” variability in activation patterns among individuals, especially in the RF, GL, BF, and SOL muscles, with the same types of variability recorded on both days of activity. The researchers noted that the machine learning system was able to identify individual muscle activation patterns during the first session with a high degree of accuracy, particularly when more of the tracked muscles were factored into the mix. The classification rate was just over 99 percent for pedaling and 98.86 percent for treadmill gait.
Recognition rates were nearly as accurate when focused on the second session, where accuracy was 89.80% for 7 muscles in pedaling, and 86.20 percent for 7 muscles during walking, say the researchers noting that the differences between the first and second sessions are due to variations in placement of the EMG sensors, but they believe that given the highly similar results, the differences in placement only strengthen their conclusions.
Finally, the study concluded that the RF, GM, GL and SOL muscles provided the best recognition data for pedaling, while the TA and BF muscles were tied strongly to better recognition data related to gait.
Although the researchers didn’t find a single explanation for why muscle activation patterns might be individualized, they note that both “optimal feedback control” and “good enough” theories of motor control could be at play in activation signatures.
Activation patterns may be consistent with the optimal feedback control theory in that “it is possible that each individual optimizes their movement with the muscle activation strategies that are best, given that individual’s mechanical and/or neural restraints,” the researchers say, adding that it is possible that the signatures develop according to the “good-enough” concept, “through motor exploration, experience, and training, leading to habitual rather than optimal strategies.
” It’s a debate that likely won’t be settled without “retrospective studies on large cohorts or longitudinal studies performed at different lifespans,” note the researchers.
As to study’s limitations, the research population was small, and homogenous. While the homogeneity was intentional to tease out the accuracy of the machine learning process, the approach limited researchers’ ability to identify potential motor control theories at play, and whether at least some of the activation strategies are innate.