The central nervous system controls human force production by successive recruitment of motor units in the skeletal muscles and changing their neural firing rate. The motor unit comprises a motoneuron, its innervated muscle fibers, and its axons. The motor units’ function provides the basis for diagnosing neuromuscular diseases, analysis in exercise physiology and sports science, and prosthetic control. Electromyography is the gold standard to measure and analyze motor units under voluntary contractions, but the technique is limited in its field of view. Recent studies have demonstrated the possibility of using imaging techniques to study motor units providing a large field of view. However, these studies are based on electrical stimulation of the muscle, and therefore only provide partial information on the motor unit’s function in contrast to voluntary contractions.
The overall purpose of this thesis was to develop methods to identify and analyze motor units in ultrafast ultrasound image sequences of voluntary skeletal muscle contractions for neuromuscular diagnostics and muscle contraction characterization. The thesis is based on four studies.
In the first study, a methodological pipeline was developed to identify motor units by decomposing image sequences into spatiotemporal components. The firing pattern and territory of the components were evaluated using an in-house developed simulation model. It showed that this pipeline identified 75-95% of the simulated motor units at low force levels. The territory estimation had a 50-80% sensitivity and 100% specificity, and the firing pattern estimation had a 90% agreement with the true firing pattern. In general, the method’s performance decreased for more than 20 active motor units. Experimental isometric contractions from healthy subjects were recorded for feasibility assessment. The results showed that the number of components increased with force level, where the number of components at 1%, 2.5%, and 5% maximal voluntary contraction averaged 7, 9, and 12, respectively. The territory diameter (5-6 mm), contraction duration (40-50 ms), and firing rate (11-12 Hz) were similar for all force levels. Thus, the results were similar to motor units’ known characteristics, suggesting that these components could be motor units.
In the second study, the proposed pipeline was validated using ultrafast ultrasound and state-of-the-art needle electromyography simultaneously. The results showed that the method could identify 31% of the motor units in low force voluntary isometric contractions, and possible explanations for the unidentified 69% were discussed. The conclusion was that the proposed pipeline can identify motor units.
The third study focused on evaluating the influence of different decomposition algorithms on performance of identifying single motor units in the data from study 2. The results showed that a decomposition algorithm is required for motor unit identification. The algorithms performed similarly in estimating firing patterns and they did not influence the motor unit twitch waveform. It was also shown that the algorithms identify different motor units, where some identified completely different units. These results suggest that the precise choice of decomposition algorithm is not critical, and there may be an improvement potential to detect more motor units.
In the fourth study, data from the second study was used to estimate single motor units’ contractile parameters based on a subset of the data (14 motor unit contractions). Multiple single motor unit’s contraction parameters were estimated using two models. Both models’ contractile parameters were consistent and agreed with previous literature. The former and more detailed model had a better experimental fit, whereas the latter model captured the “average behavior” with fewer parameters. It was found that the single twitch waveforms within a motor unit change shape during a voluntary isometric contraction at a low force level. These results suggest that the motor unit’s contractile parameters can be estimated using ultrafast ultrasound image sequences in voluntary isometric contractions.
In summary, a methodological pipeline to identify motor units was developed, evaluated, and validated. The key module in the pipeline, i.e., decomposition algorithm, was evaluated by comparing different algorithms’ influence on identifying single motor units. Finally, the pipeline output can be used for estimating motor units’ contractile parameters. This pipeline may contribute to neuromuscular diagnostics and muscle contraction characterization. In general, it may allow the study of various motor unit-related questions that previously were difficult or not possible to address.