An objective evaluation of a segmented foot model Nori Okita a,c, Steven A. Meyers a,c, John H. Challis a, Neil A. Sharkey a,b,* a Biomechanics Laboratory, Department of Kinesiology, The Pennsylvania State University, University Park, PA, USA b Department of Orthopaedics and Rehabilitation, Penn State College of Medicine, Hershey, PA, USA c Department of Mechanical and Nuclear Engineering, The Pennsylvania State University, University Park, PA, USA 1. Introduction Traditional gait analysis considers the foot as a single rigid body with no intrinsic motion. The quantitative data that this approach provides has proven useful in clinical studies and in research aimed at assessing the behavior of the hip, knee and ankle under various conditions [1���3]. Unlike other segments of the lower extremity (i.e., the pelvis, shank, and thigh), the foot is composed of multiple bones and joints with complex interactions. Furthermore, aberrant motion and flexible deformities of the foot are common clinical problems, particularly in neuromuscular conditions such as cerebral palsy and stroke. The traditional single-body foot model fails to provide clinically meaningful information regarding the kinematic behavior of the foot in health and disease [4], thus improved methods of examination are warranted. To rectify this situation, recent extensions of the non-invasive three-dimensional photogrammetry method divide the foot into multiple segments and treat each one as a separate rigid body [4��� 15]. To date little objective work has been undertaken to determine if these multi-segmented foot models are robust enough to confidently infer internal skeletal behavior from measurements of external marker motion. There are two potential sources of error associated with modeling the foot as a multi-segmented structure: (1) departure from the rigid body assumption [10,16,17,21], and (2) motion artifact due to skin and soft tissue motion in relation to the underlying bone or bones of interest [18���21]. The goal of this research was to objectively evaluate the fidelity of a generalized three-segment foot and ankle model [5,22���25]. An established apparatus that reproduces the kinematics and kinetics of gait in cadaver lower extremities [26���28] was used to independently examine the validity of the rigid body assumption and the magnitude of soft tissue marker artifact by comparing data derived from skin-mounted markers, such as would be used in clinical applications, with data derived from bone-mounted marker clusters. We examined two null-hypotheses: (1) clinical segments constructed from externally mounted skin markers behave as rigid bodies (2) there are no differences in the kinematic Gait & Posture 30 (2009) 27���34 A R T I C L E I N F O Article history: Received 19 August 2008 Received in revised form 3 February 2009 Accepted 16 February 2009 Keywords: Foot and ankle modeling Gait simulation Motion analysis Skin artifact A B S T R A C T Segmented foot and ankle models divide the foot into multiple segments in order to obtain more meaningful information about its functional behavior in health and disease. The goal of this research was to objectively evaluate the fidelity of a generalized three-segment foot and ankle model defined using externally mounted markers. An established apparatus that reproduces the kinematics and kinetics of gait in cadaver lower extremities was used to independently examine the validity of the rigid body assumption and the magnitude of soft tissue artifact induced by skin-mounted markers. Stance phase simulations were conducted on ten donated limbs while recording the three-dimensional kinematic trajectories of skin-mounted and then bone-mounted marker constructs. Segment kinematics were compared to underlying bone kinematics to examine the rigid body assumption. Virtual markers were calculated from the bone mounted marker set and then compared to the skin-mounted markers to examine soft tissue artifact. The shank and hindfoot segments behaved as rigid bodies. The forefoot segment violated the rigid body assumption, as evidenced by significant differences between motions of the first metatarsal and the forefoot segment, and relative motion between the first and fifth metatarsals. Motion vectors of the external skin markers relative to their virtual counterparts were no more than 3 mm in each direction, and 3���7 mm overall. Artifactual marker motion had mild affects on inter- segmental kinematics. Despite errors, the segmented model appeared to perform reasonably well overall. The data presented here enable more informed interpretations of clinical findings using the segmented model approach. �� 2009 Elsevier B.V. All rights reserved. * Corresponding author at: 201 Henderson Building, Pennsylvania State University, University Park, PA 16802, USA. Tel.: +1 814 863 2426 fax: +1 814 863 8698. E-mail address: nas9@psu.edu (N.A. Sharkey). Contents lists available at ScienceDirect Gait & Posture journal homepage: www.elsevier.com/locate/gaitpost 0966-6362/$ ��� see front matter �� 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.gaitpost.2009.02.010

data derived from skin-mounted and bone-mounted markers (absence of soft tissue artifact). 2. Methods 2.1. Experiment A total of 10 normal fresh frozen cadaver extremities (5M/5F, 42���84 years of age) were evaluated. Specimen preparation included removal of all soft tissues 5 cm superior to the malleoli, taking care to preserve the entire lengths of tendons from six muscle groups: the gastrocnemius and soleus complex (TS) tibialis anterior and extensor digitorum longus (TA) tibialis posterior (TP) flexor digitorum longus (FDL) flexor hallucis longus (FHL) peroneus longus and peroneus brevis muscles (PER). Skin and retinaculi about the ankle were preserved. The tibia and fibula were transected approximately 23 cm superior to the sole of the foot, and the proximal ends were cemented to a coupling device via an intramedullary tibial rod and polymethylmethacrylate. Simulations of the stance phase of gait were conducted at 1/20th of normal walking speed using the robotic dynamic activity simulator (RDAS) [26���28]. Our current simulations utilize an in-house library of shank kinematics and corresponding target ground reaction forces enabling each specimen to be size- matched with input data taken from similarly sized live subjects. Temporally based muscle force profiles are constructed from rectified EMG profiles [29] adjusted for force���length and force���velocity properties [30]. Peak contractile abilities are normalized according to estimated body weight of the donor. In the present experiment body weight ranged from 35 kg to 50 kg. Shank kinematics and the actions of the six muscle groups listed above were adjusted until target vertical ground reaction forces with peaks equal to 1.1 times estimated body weight were attained. Once established, simulation parameters were held constant for all trials in both the skin marker and bone marker configurations. Fig. 1. (A) The reference poses from the skin-mounted and bone-mounted marker trials were superimposed to define fixed local vectors of skin markers with respect to the corresponding bone-mounted clusters. Virtual markers were calculated from these vectors and coordinate transformation matrices of each bone. (B) Global and segmental coordinate system definition. The foot was oriented in the global coordinate system with the X axis oriented posterior ( ) to anterior (+), the Y axis inferior ( ) to superior (+), and the Z axis left ( ) to right (+). The mismatch between the segment coordinate frames based on skin-mounted and bone-mounted markers are illustrated. Unit (A, B): unit vector in the direction from A to B, mid(A, B): mid-point between A and B. N. Okita et al. / Gait & Posture 30 (2009) 27���34 28