In this study, we used the 2D ML video-based analysis system, at several habilitation clinics in Sweden, to explore its feasibility and possible usefulness in children with CP. We identified and quantitatively assessed several minor and major deviations in hip, knee, ankle and foot orientation kinematics and in gait speed and step length.
Kinematics, hip, knee, ankle and foot orientation and classification
The classification system of Rodda et al. is based on kinematics in the sagittal plane for children with bilateral and unilateral CP, using primarily the position of the foot, secondarily the kinematics of the ankle and knee, and lastly the kinematics of the hip and pelvis, to cover the full complexity of spastic gait disorders [3]. Rodda et al. however acknowledge the importance of coronal and transverse plane deviations in children with CP and a 3D GA is required for a full gait evaluation of all planes (Rodda 2001).
Based on ankle kinematics and foot orientation, we could identify two of the gait patterns described by Rodda et al and present in our cohort: true equinus (Fig. 3B) and apparent equinus (Figs. 2B and 4B).
In the true equinus gait pattern (Fig. 3B), toe-walking with ankle plantarflexion throughout stance is the dominant sign and hip and knee kinematics reveal full extension. This pattern is often caused by calf muscle spasticity and can be managed by spasticity reduction (Botox injections), hamstring lengthening surgery and/or ankle-foot orthosis. True equinus is common in the young child, but rarely remains throughout childhood [3, 4].
In the apparent equinus gait pattern (Figs. 2B and 4B), there is also toe-walking. However, the ankle has a normal range of motion and it is increased knee and hip flexion during the stance phase that causes toe-walking (apparent equinus). It is crucial to distinguish apparent equinus from true equinus since the former seldom requires surgical treatment with calf muscle lengthening, which is commonly performed in true equinus [3, 4]. In addition to the gait analysis, careful assessment is needed through physical examination to distinguish between short gastrocnemius muscle only and shortening of both gastrocnemius and soleus muscles, so as to determine the level of calf muscle lengthening required. Hence, it is important not to limit the evaluation of a child to one tool only, but rather to use several: in this case, gait analysis and physical examination.
In the jump gait pattern, not present in our study cohort, toe-walking is present, the ankle is in plantarflexion and the knee and hip reveal increased flexion in stance. Management can include orthosis, spasticity reduction and/or surgery, depending on age and severity [3, 4].
Lastly, in the crouch gait pattern, not present in our study cohort, there is excessive ankle dorsiflexion with increased knee and hip flexion. Crouch gait is most commonly caused by earlier overlengthening of the soleus muscle (see the discussion above regarding apparent equinus pattern), and is notoriously difficult to manage [3, 4].
With both the 3D GA and the 2D ML systems, there are several challenges to making exact measurements. The commonly developed foot deformities hind foot equinus with midfoot abduction and pes planovalgus with midfoot break are especially problematic and difficult to assess reliably. From this perspective, it can be difficult to identify if a child toe-walks or not and if the ankle is in plantarflexion or in dorsiflexion, since this can be obscured by the foot deformity. We found the foot orientation in the room helpful for this purpose (Figs. 4A and 4B).
Rodda et al. described a mild gait deviation pattern: sagittal kinematics within the 1 SD band of their laboratory normal range [9]. We also identified mild gait deviations in our cohort (Figs. 2A, 3A and 4A).
Although the hip kinematics with the 2D ML system used in this study was based on the thigh in relation to a horizontal line, not in relation to the pelvis, as in the 3D GA system, the gait patterns described by Rodda and Graham could be determined almost fully. Furthermore, even with the offset between the 2D ML and 3D GA systems, it was possible to determine and identify deviating patterns with the 2D ML method, making comparisons over time possible. Additionally, it has been pointed out that performing classification of gait patterns aids prognosis, enabling awareness of expected later deviations of gait pattern [3, 36].
Gait speed and step length
Both subjective perceptions of gait and the ability to control gait speed have been reported in children at various GMFCS levels [37, 38]. In our cohort, the mean gait speed was 0.69 m/s (range 0.12–0.91), far from the average gait speed in typically developed children, where a 7-year-old walks at an average of 1.14 m/s [39].
Adequate step length is considered a prerequisite for acceptable gait speed, with associated stability in stance [23, 40]}. Abel et al. reported shorter step length, 0.79 m (SD 0.19) in children with bilateral CP compared with typically developed children, 1.08 m (SD 0.14) [41]. Our cohort had even shorter step length, mean 0.37 m (SD 0.08, range 0.12–0.49). The 2D ML system provided reliable measurements of gait speed and step length.
Setup, acquisition and processing
The setup of camera, light sources, walkway and background was easy to perform. The equipment can be moved to any location as long as there is a room that measures at least 5 times 10 meters, which is not ideal for a 3D GA system.
Acquisition always depends on the child’s cooperation and walking ability, but the 2D ML system saves time compared with a 3D GA. The 2D ML also has the advantage of no markers being attached to the body, which can be uncomfortable for the child and are at risk of falling off during acquisition.
The processing is automated, requiring limited manual involvement, and thus saves time. An additional general advantage is that 2D ML does not require advanced training or knowledge. This, together with low cost of equipment and that the system is portable, makes it suitable for screening and follow-up purposes.
Follow-up programs and limited assessment of gait
In most follow-up programs for children with CP, extensive physical examinations are performed on a regular basis with a focus on preventing severe complications, such as hip dislocation, common at GMFCS levels IV and V. However, around 70% of children with CP are at GMFCS levels I, II or III and are ambulatory [12]. Nevertheless, no objective quantification of gait variables is currently included in the follow-up programs, to screen and follow the development of gait over time. Static physical examination alone has been shown to be insufficient to assess gait [42, 43].
Haumont et al. pointed out in a study from a pediatric specialty center that a treatment program with careful orthopedic follow-up based on medical history, physical examination and gait analysis leads to improved gait function throughout childhood [44].
Limitations
Limitations include the small number of participants in the study and that we did not collect data on typically developed children. In the 2D ML we did not evaluate the transverse or coronal plane, both of which is important in the evaluation of gait deviation in children with CP. A 3 D GA was not performed and thus a comparison with the classification by Rodda et al. (ref) between the two systems was not performed. Another limitation was that we did not follow our study cohort over time, although this study did establish 2D ML gait analysis as a possible dynamic complement in the evaluation of gait. Nevertheless, we think these children can serve as examples, illustrating various variables and deviations possible to obtain with the 2D ML technique and the practicalities surrounding the assessments and processing of data.