Over 795,000 people in the United States experience a stroke each year and over half of stroke survivors over age 65 have reduced mobility . Regaining walking ability is an important goal of rehabilitation as walking speed is a critical predictor of long-term health  and stroke survivors who achieve limited or full community walking speeds report an overall higher quality of life than those who remain household ambulators . Successful walking requires the execution of the critical pre-swing biomechanical subtasks of body propulsion and leg swing initiation, which are often impaired post-stroke [4, 5].
Deficits in these late stance subtasks may also have important implications for achieving adequate knee flexion during the swing phase. For example, a lack of push-off acceleration rather than over-activity of the knee extensors may be the primary cause of stiff knee gait in some individuals post-stroke  and stimulating the plantarflexors in pre-swing increased peak knee flexion . In addition, late stance phase braking forces in stroke survivors predict less knee flexion during swing, and late braking and low propulsion have been shown to predict circumduction . However, interestingly the total propulsive force is not correlated to knee flexion  and it is unknown if the relationships between braking, propulsion and swing phase kinematics are causal or correlative.
Braking and propulsion deficits are common in individuals post-stroke . While healthy individuals produce braking only in the first half of stance and propulsion in the second half, stroke survivors often have prolonged braking and low propulsion output, which predicts slower walking speed [8, 9]. The plantarflexor muscles are primary contributors to propulsion [10, 11] and decreased plantarflexor contributions to propulsion have been observed in stroke survivors [12–14], with post-training increases in walking speed being predicted by greater plantarflexor contributions to propulsion . However, a lack of propulsion can occur not only because of low plantarflexor output, but propulsion can be offset by prolonged activity from the vasti, which are primary contributors to braking forces . On average, individuals with impaired plantarflexor coordination do not have lower propulsion, but rather greater braking than individuals with normal plantarflexor coordination , which is likely due to co-activation of the plantarflexors and vasti muscles. Due to the characteristically high variability between stroke survivors, there are a number of mechanisms that can cause the propulsion deficits.
Stroke survivors also experience deficits in leg movement throughout swing [18, 19], with modeling studies having identified knee flexion velocity at toe-off as the primary contributor to peak knee flexion during swing [20–22]. Knee velocity at toe-off may be diminished by late braking forces because muscles such as the vasti and rectus femoris that contribute to braking also contribute to knee extension and oppose leg swing initiation in late-stance . Deficits in leg swing initiation may also be caused by decreased gastrocnemius contributions to leg-swing initiation, leading to lower knee velocity at toe-off and consequently less knee flexion during swing. For example, medial gastrocnemius contribution to knee flexion acceleration increased after gait retraining . However, a representative stroke survivor with a limited community walking speed had similar contributions from the gastrocnemius and vasti compared to a healthy control walking at the same speed but less power delivered to the leg in pre-swing by the iliopsoas . Thus, the high degree of variability in stroke survivors makes these results difficult to generalize and it is unknown how these deficits in leg-swing initiation impact swing phase kinematics.
Impaired knee flexion is often attributed to rectus femoris activity [18, 23]. In a previous modeling study, eliminating rectus femoris activity in pre-swing was more effective than eliminating rectus femoris activity in early swing for improving knee flexion . The gluteus medius, vasti, and rectus femoris have the greatest potential to decrease knee flexion velocity in late stance, while the sartorius, gracilis, biceps femoris short head, gastrocnemius, iliopsoas and hamstrings have the greatest potential to increase knee flexion velocity in late stance , although it is unknown which muscles most affect pre-swing knee flexion velocity in stroke survivors.
Previous work has established the importance of pre-swing conditions to achieving adequate swing phase knee flexion. However, actual muscle contributions to propulsion, knee velocity and leg-swing initiation in stroke survivors and their relationship to swing-phase knee flexion has not been established. Moreover, most simulation studies are limited by a low number of participants, and therefore may not be generalizable to the overall stroke population. Thus, the objectives of this study were to: 1) analyze a large number of stroke survivors and determine the underlying causes of propulsion and braking deficits in late-stance, 2) identify primary muscle contributors to knee velocity and leg power in late stance, and 3) determine whether these muscle contributions to knee velocity and leg power predict knee flexion in swing. We hypothesized that 1) braking and propulsion asymmetries would be caused by both low plantarflexor contributions to propulsion and high vasti contributions to braking, 2) vasti and plantarflexor contributions to propulsion and braking in pre-swing would predict swing phase knee flexion, 3) the rectus femoris would be a major contributor to knee extension in pre-swing in individuals with stiff knee gait, and 4) greater knee flexion would be correlated with greater power delivered to the leg in pre-swing. The outcomes of this work will highlight specific deficits in propulsion and leg swing initiation post-stroke and their implications for swing phase knee flexion, which will provide a basis for developing targeted walking interventions.