Theses supervised by Prof. Dr. Can Ali Yücesoy
10 theses · Boğaziçi University
Development of a recurrent neural network based estimation algorithms for ankle power by using surface EMG
Estimation of ankle power can be used in identification of gait abnormalities and establishing timings of net power generation in powered prosthetic devices. Current inverse dynamics calculations of ankle power rely on gait analysis data collected in specialized, expensive laboratories, which limits its applicability and accessibility for prosthetic device users. The aim of this study is to develop a Recurrent Neural Network system to estimate ankle power during level walking by using only surface electromyography (sEMG) as algorithm inputs. For this purpose, an open access data set which includes 50 participants with 25 males and 25 females aged between 6 to 72. In the dataset there are sEMG data from upper leg muscles: Biceps Femoris (BF), Gastrocnemius Medialis (GM), Gluteus Maximum (GMax), Rectus Femoris (RF), Vastus Medialis (VM) and lower body muscles; Peroneus Longus (PL), Soleus (SO), Tibialis Anterior (TA). Algorithms for combinations of all these muscles have been developed. A correlation coefficient of 0.90 between the actual (result of gait analysis) and predicted ankle power is considered to perform successfully. 25 muscle combinations yielded successful correlations with 1 set of 1 muscle, 3 sets of 2 muscles, 5 sets of 3 muscles, 9 sets of 4 muscles, 5 sets of 5 muscles, 1 set of 6 muscles, 1 set of 7 muscles. Note that, all successful muscle groups include either PL or GM muscle. Our findings suggest that our system can be used in powered prosthetics control and detection of gait abnormalities.
Assessment of active state titin's effects on muscle mechanics using finite element modeling
Calcium dependent mechanical behaviors characterize titin's contribution to force production in three-myofilament paradigm: (1) Stiffening of PEVK (Proline, Glutamate, Valine, Lysine) segment, and (2) reduction of free-spring length via N2A-titin binding. This thesis is focused on the introduction of an alternative perspective to the analysis of titin with incorporating epimuscular myofascial loads. Isolated and integrated rat muscle finite element model variations were used with three titin models: passive state titin, active state titin-I and active state titin-II. Results of isolated model showed that active state titin-I and II limits sarcomere shortening (lm = 32.7mm: up to 10% and 20%, respectively). Such shorter sarcomere effect characterizes active state titin's mechanism of effects. Integrated models showed that the shorter sarcomere effect becomes an inconsistent and variable mechanism: Shorter sarcomere effect is further enhanced for proximal fascicle interfaces (by 30.2% and 31.0%, respectively) whereas it is also diminished for remaining fascicles (by 10.3% and 14.0%, respectively), but even a longer sarcomere effect is shown. Overall, titin's mechanism of effect and functionality are manipulated by epimuscular myofascial force transmission. This implies a new approach for the 3-miyofilament model: For the analysis of the components of the contractile machinery, contribution to force production and contribution to muscle mechanics should be assessed with alternative perspectives. Titin's calcium dependent mechanical behaviors belong to former as these increases its stiffness, whereas shorter sarcomere effect belongs to latter as this mechanism further translates its effect to other components as well as to length-force characteristics. These together comprehensively define titin's contribution as a third myofilament.
Effects of hamstring lengthening surgery on muscle–tendon velocities of patients with cerebral palsy
Cerebral Palsy (CP) is a permanent movement disorder seen in early childhood, consisting of muscle spasticity and/or contracture, and difficulty in walking due to poor selective control. In CP patients, crouch gait with excessive knee flexion is usually corrected by hamstring lengthening surgery, which is believed to improve gait, by increasing the length or velocity of the spastic muscle. However, hamstring muscles that may not be not short and slow before the surgery can also be operated on. The thesis aims to assess whether the gait of CP patients improved after the surgery by testing the following hypotheses: (i) knee joint movement does not improve postsurgery, (ii) the hip joint movement was impaired pre-surgery, (iii) the gait deviation index (GDI) increases post-surgery, (iv) the muscle lenghtening velocity remains unchanged in postsurgery, and (v) the pre-surgery psoas lenghtening velocity is slower than post-surgery. 8 limbs of 4 CP patients who had undergone hamstring lengthening surgery were included in the study. Pre-and post-surgery muscle lengthening velocity changes of patients were compared with reference to age-matched TD children (14 limbs of 7 participants) based on gait analysis data and using musculoskeletal modeling (OpenSim). Our results showed that post-surgery, mean knee angular velocity did not change significantly. No significant effect of surgery was shown in hip angular velocity or GDI. Moreover, no significant changes were shown in hamstring muscle lenghtening velocities. Only two of the preoperative patients had slow psoas muscle lenghtening velocity. As a result, post-surgery improvement in knee movement was achieved without a significant change in hamstring muscle lenghtening velocity. Keywords: Muscle Shortness, Cerebral Palsy, Contracture, OpenSim, Hamstring Lengthening Surgery, Spasticity, Psoas, Lengthening Velocity
Skeletal muscle mechanics and spasticity management: Human and animal experiments
Being the most common motor disability in childhood, cerebral palsy (CP) describes a movement disorder for which the exact underlying mechanism is unclear, and no cure is available. Yet, local injection of botulinum toxin type-A (BTX-A) is used for spasticity management. In this thesis, the relationship between the mechanics of spastic muscles and the impaired joint motion was investigated in patients, and the long-term effects of BTX-A on muscular mechanics were assessed in animals. Experiments on spastic knee flexors showed that passive muscle forces are much less than active forces (e.g., 26%), and epimuscular myofascial force transmission (EMFT) arising from intermuscular mechanical interactions significantly increases active forces (up to 132%). Combined with musculoskeletal models developed based on gait analysis data, EMFT effects were shown to be compatible with metrics characterizing patients' pathological gait, indicating that intermuscular mechanical interactions may be a source of high flexor forces in flexed joint positions. Experiments in the rat anterior crural compartment showed that long-term after injection, BTX-A yields in addition to decreased active forces, both unintended (a narrower range of force exertion by 23% and increased passive forces by 12%, for the injected muscle) and uncontrolled effects (similar effects on compartmental muscles due to the spread of the toxin). BTX-A also leads to collagen content increase (by several folds) for muscles exposed, which explains elevated passive forces and impacts also active forces. These effects are of high potential clinical importance as they conflict with therapeutic goals. Particularly, controlling the effects of BTX-A on connective tissue adaptation is critical for better spasticity management.
Human muscle structure-function relation in-vivo using magnetic resonance imaging modalities
Non-uniform muscle deformation has become a frequent finding in biomechanics research, using imaging modalities operating at different resolution levels from sarcomeres to fascicles. Mainly due to technical limitations, interpretations of these findings are detached from a theoretical foundation that considers the muscle with mechanical links to its surrounding. To enable this vital consideration, this thesis aims at developing and testing the validity of a multimodal MRI method that bridges the understanding between non-uniform mechanical deformations and their myofascial origins, in-vivo. 1) Supplemented with DTI tractography, registration-based fiber direction deformations and principal strains on NVTs characterized the myofascial loads in relation to the strain heterogeneity pattern in active muscle (proximally shortened (up to 22%), distally lengthened (up to 108%) fascicles). Inter-subject deviations from the general pattern were in agreement with subject specific anatomy. 2) A multiverse analysis was performed on the tuning parameters of the demons registration algorithm to assess the validity of strain distribution pattern against algorithmic choices. Results showed that the overall deformation pattern was immune to such perturbations, yet the strains amplitudes underwent significant changes. 3) To add orthogonal information to the myofascial origin assessment and validation of strain distributions, quantitative and velocimetry MRI were used. T1 mapping showed promising results in associating microstructural content with the strain distribution pattern. SR patterns from 2D VEPC showed weak similarities with registration-based principal strains, whereas those from compressed sensing 4D-PC showed much better agreement. Collectively, these studies show a way forward for the understanding of in-vivo muscle structure function relationship with implications for muscle physiology in health and disease.
Investigating skeletal muscle adaptations due to btx-a injections using agent-based modelling
Local application of Botulinum Toxin Type-A (BTX-A) has been the gold standard for spasticity management in children with Cerebral Palsy. The treatment aims to reduce the passive resistance of force at the joint and increase the joint range of motion. However, recent studies have reported results contradictory to treatment aims including increased passive force, increased muscle stiffness and decreased length range of force exertion which have been attributed to the increased collagen content in the muscle Extracellular Matrix (ECM) confirmed by histological findings. Moreover, a recent finite element analysis study has reported that BTX-A injections were found to increase the injected muscle fibers' strain. Hence, to understand these muscle adaptations and the effects of strain on muscle structure and function, the Agent-Based Modelling (ABM) method was used. The advantages of this method over other methods is that it allows studying muscle adaptations at the cellular level with the complex interactions modelled. This study was modelled in three cases comparing the BTX-A model, BTX-A-Free model and a middle half paralyzed (MHP) model similar to that used in the recent finite element model. Two of the cases compared the BTX-A-Free fascicle against the BTX-A and the MHP fascicles and the third case compared the BTX-A-Free case to a modified MHP case which included the reported BTX-A induced atrophy and the decaying effects of BTX-A 7-10 days after injection. The collagen increase in the cases was approximately 33%, 33%, and 20.3% respectively. This study revealed that although the BTX-A induced strain increase was found to increase collagen content of the ECM, other important factors such as the BTX-A induced atrophy may also play a significant role in collagen increase in the muscle ECM.
sEMG-based ankle position and moment prediction in silico: Neural network approach and muscle selection
Lower limb amputation is the partial or complete removal of a limb, and powered prostheses are the best solution for restoring amputees' locomotion abilities. Although recent advancements have enhanced their hardware, autonomous adaptation is required to achieve natural ambulation. The utilization of surface electromyogram (sEMG) holds promise, whereas real-time analysis is challenging. Also, a systematic analysis should be conducted for muscle selection to ensure compatibility with different levels of amputations. Therefore, the feature extraction was implemented for non-normalized sEMG amplitudes, and an economic algorithm minimizing sEMG input was sought. For the sake of different amputation level compatibility, a practical algorithm was aimed to limit the use of lower leg muscles. In this context, neural network-based algorithms with timing-based approaches utilizing sEMG amplitudes as inputs have been developed to (1) predict sagittal ankle position and moment during ground-level walking and (2) rank all muscle combinations based on success. Eight leg muscles were studied: tibialis anterior (TA), soleus (SO), medial gastrocnemius (MG), peroneus longus (PL), rectus femoris (RF), vastus medialis (VM), biceps femoris (BF) and gluteus maximus (GMax). The results showed the best-performing muscle variation was MG+RF+VM whereas, PL and GMax+VM were distinguished as the economic and practical variations (rposition>0.90, rmoment>0.97), respectively. The analysis regarding the effect of window size selection for feature extraction on prediction accuracy revealed that a window size of 150 ms demonstrated the best performance for the proposed neural network architecture. The cross-validation results supported the repeatability of the structure and methodology.
Patient specific musculoskeletal modeling to relate intra operative muscle force data to gait in cerebral palsy
Increased pathological resistance against knee extension is a characteristic pathological condition in cerebral palsy (CP). Active and passive forces of spastic knee flexor muscles have been associated with the high forces which constrain the knee movement in flexed positions. However, studies quantifying these forces directly are rare. The aim of this study was to determine (1) if the range for spastic muscle-tendon complex length is comparable to that of healthy muscle and (2) how spastic muscle force changes as a function of muscle-tendon complex length. Musculoskeletal patient specific models were developed using OpenSim software by using the gait analysis data of CP patients. The results suggest that the muscle-tendon complex length trends of the patients are similar to those of healthy individuals. The difference between the minimum and maximum MTL is 13% and 8% for healthy ST-GRA muscles respectively, while this difference oscillates between 7% and 14% for spastic ST muscles and 4%-9% for spastic GRA muscles of the patients. In addition, at longer muscle-tendon complex lengths higher passive and active forces of spastic muscles were measured and the dominant component of the total force is the active muscle force. For the GRA, peak active muscle force is more than 5 times that of passive force. For the ST, peak active muscle force is approximately 4 times that of passive force. The outcome of this assessment suggests that mechanical characteristics of spastic muscles may not necessarily differ from those of healthy muscles. This implies that, an explanation for the restricted joint movement of CP patients via the consideration of spastic muscles as shortened and mechanically abnormal muscles is a simplistic one.
Assessment of effectiveness of muscle lengthening surgery in cerebral palsy using musculoskeletal modeling
Cerebral Palsy is a permanent movement disorder that manifests itself at early childhood, as poor coordination, and gait difficulty due to muscle spasticity and/or contracture. Excessive knee flexion during gait e.g., crouch gait is a common impairment and often corrected by hamstring lengthening surgery. Such crouch gait is presumed to originate from shortness (i.e., contracture) and/or slowness in lengthening (i.e., spasticity) of affected muscles and remedial surgery is considered to improve gait by increasing muscle length or its velocity. However, a third group of patients who neither has short nor slow hamstrings pre-operatively can still undergo surgery. The aim of the thesis is to investigate whether the gait of those patients improved after the surgery by testing the hypotheses: post-operatively, (i) the knee joint movement is improved (ii), the hip joint movement is deteriorated (iii), the gait deviation index (GDI) is increased (iv), muscle unit length do not change, and (v) pre-operative psoas muscle lengths were shorter. Findings showed that mean knee angle decreased significantly at the initial contact (0-3%), terminal stance (28-43%), and the terminal swing phases (95-100%) (p<0.05). No significant effects were detected in the hip angle or the GDI. Additionally, no significant changes in the hamstring muscle lengths were found. Only half of the patients had shorter psoas muscle lengths pre-operatively. In conclusion, the excessive knee flexion of the patients was decreased without negatively affecting hip movement or gait overall. The improvement in the knee angle was achieved without any change in the muscle length of the hamstrings, suggesting that an isometric position shift of the target muscle occurs, which may be ascribed to post-surgical alterations in the epimuscular connections.
Epimuscular myofascial force transmission: insights on BTX-A intervention and muscle spindle sensory feedback
Mechanical equilibrium between skeletal muscles is maintained via connective tissues that surround the musculoskeletal system as a continuum. Muscle contrac- tion generates forces that spread within the muscle compartment through connective tissues, facilitating epimuscular myofascial force transmission (EMFT). This thesis ex- plored mechanical and sensory implications of EMFT using in vivo animal experiments. First, the long-term effects of botulinum toxin type-A (BTX-A), a common spasticity management drug, on individual and intermuscular muscle mechanics was studied in rats. Although BTX-A-induced paralysis was sustained, findings contradicted clinical aims at the individual muscle level (e.g., 12% increase in passive forces in the injected muscle and narrowed range of force exertion in the non-targeted bi-articular muscle by 15% distally and 32% proximally). Importantly, unlike in acute observations, dimin- ished EMFT was not observed. Given its clinical relevance in spastic muscle mechanics, BTX-A's efficacy seems tied to its impact on EMFT. These studies highlighted complex and uncontrolled BTX-A effects on intermuscular mechanics, suggesting limitations in long-term outcomes for managing spasticity. Then, the influence of EMFT on muscle spindle (MS) sensory feedback was tested in mice. Synergistic muscle stimulation re- sulted in significant increase in MS afferent firing (53% on average) due to intermuscular mechanical interactions. Variability in MS responses indicated that MSs sense local length changes across the muscle that arise from muscle fiber-extracellular matrix in- teractions. Findings provided insights for optimizing spasticity management strategies and EMFT's role in sensorimotor integration.