twente spine model - development, validation, and application of a complete and coherent musculoskeletal model of the human spine
Riza Bayoglu is a PhD student in the department of Biomechanical Engineering. His supervisors are prof.dr.ir. H.F.J.M. Koopman and prof.dr.ir. N.J.J. Verdonschot from the faculty of Engineering Technology (ET).
The human spine is a complex biologic structure composed of several articulating bones, intervertebral discs, and ligaments. The spine interacts with the shoulder- arm complex and lower extremity and carries important functions such as protecting the spinal cord, supporting the trunk, and providing motion. While performing these functions, the spine is exposed to considerable loads which might affect its performance and lead to musculoskeletal and neuromuscular complications. Understanding the nature of the loads acting on the spinal tissues can assist with diagnosing and treating spinal pathologies, thus assessment of the spinal loads has been the focus of much research. In the past, several studies measured in vivo spinal loads for most activities of daily life through intradiscal pressures or telemetrized implants. Unfortunately, these measurements were limited to some thoracic and lumbar levels due to difficulties primarily associated with ethical considerations. A comprehensive assessment of the spinal loads throughout the spine can advance our understanding of its functioning but is largely unavailable.
Musculoskeletal modeling offers a non-invasive means to estimate in vivo spinal loads and can thus provide clinical insights into the spine’s functioning. The primary objective of this thesis was to develop a validated, complete and coherent musculoskeletal model of the entire human spine for investigating the spinal loads. To achieve this goal, in Chapters 2 & 3, an anatomical dataset (the Twente Spine Dataset) including necessary musculoskeletal parameters for creating this model was measured. This dataset was obtained by medical imaging and detailed dissection of a male cadaver. Spinal and abdominal muscles located in the ribcage and cervical, thoracic, and lumbar regions were identified and divided into several muscle-tendon elements. For each element, locations of the attachment sites at the origin, insertion, and via points were digitized, and its morphology was measured. The morphological parameters consisted of the fiber length, tendon length, sarcomere length, optimal fiber length, pennation angle, mass, and physiological cross-sectional area. New
quantitative data were reported for several muscles such as rotatores, multifidus, serratus posterior inferior, levatores costarum, spinalis, semispinalis, subcostales, transversus thoracis, and intercostales. Muscle parameters reported in this dataset were within the range of data found in earlier studies. This dataset facilitates the development of a coherent musculoskeletal model for the entire human spine and prevents uncertainties intrinsic in combining musculoskeletal data from different studies. In essence, this dataset makes up the first complete and consistent atlas for modeling the human spine.
In Chapter 4, a complete and coherent musculoskeletal model of the entire human spine (the Twente Spine Model) was developed based on the previously acquired anatomical dataset. In this model, cervical, thoracic, and lumbar vertebrae, a flexible ribcage, and comprehensive muscular anatomy were incorporated. This spine model in principle represents a subject-specific musculoskeletal system and, unlike generic models, does not require combining anatomical data from several studies. An inverse dynamics based static optimization routine minimizing muscle fatigue was used for calculating muscle and joint forces during basic neck and trunk movements. For validation of the predicted internal loads, quasi-static trunk tasks as measured in previous in vivo studies were simulated, and calculated intradiscal pressures at thoracic and lumbar discs and normalized resultant loads were compared. In addition, disc compression forces were calculated during upright standing, flexion, extension, lateral bending, axial rotation and were reported as percentages of the total body weight. Disc compression forces, in general, increased caudally and with increasing trunk motion, yet decreased at the lumbar and lower thoracic levels in extension. The cervical compression forces were lower than those of the thoracolumbar discs during neck tasks. Although there were some discrepancies, the results, in general, agreed well with previously measured in vivo spinal loads. This indicated the first validation of the Twente Spine Model in predicting the internal loads at the intervertebral discs during basic activities of daily life.
In Chapter 5, the sensitivity of muscle and intervertebral disc force computations against potential errors in modeling muscle attachment sites (muscle origin, insertion, and via points) were investigated. For this, every muscle attachment location was perturbed in the Twente Spine Model during upright standing, flexion, lateral bending, and axial rotation of the trunk. The changes in the T6/T7, T12/L1, and L4/L5 disc forces were analyzed, and an overall sensitivity index value was calculated for every perturbed muscle. Most of the muscles that were studied resulted in minor changes (less than 5%) in disc forces. However, certain muscle groups had a large influence. For example, fluctuations in the quadratus lumborum caused changes in the shear forces as high as 353% and changes in the compressive forces as high as 17%. Musculoskeletal simulations identified certain muscles in the ribcage and lumbar spine as being more sensitive for calculating muscle and disc forces. The findings suggest that modeling muscle attachment sites simply based on anatomical illustrations might lead to an erroneous evaluation of the internal forces.
In Chapter 6, electromyographic activities and trunk movements during isometric and dynamic trunk activities were simultaneously measured. Although differences were observed between women and men and between body heights, consistent muscle activation trends were found in most static and dynamic exertions. Measured trunk tasks could be isometrically or dynamically simulated in inverse dynamics based spinal musculoskeletal models, and predicted and measured muscle activations could then be compared to verify musculoskeletal predictions.
Finally, in Chapter 7, musculoskeletal and patient-specific finite element models were used in combination to investigate if modeling more physiological load regimes (compared to simpler loading regimes) can significantly affect the vertebral fracture risk prediction. The Twente Spine Model was scaled to the body height and weight of a patient who had spinal metastasis and was used to extract the loading and boundary conditions at the L2/L3 joint during physiological trunk movements. These boundary conditions were implemented in a finite element model of the L3-L5 spinal unit for studying its mechanical behavior. Results indicated that only trunk flexion caused plasticity at the L4 vertebra and that the disc force during 45° flexed position was significantly lower than the axial force which resulted in the same plasticity level. This implied that the metastasized vertebra was likely to fail under flexion loads than under only a compression load.
