Freezing of gait assessment with inertial measurement units and deep learning: effect of tasks, medication states, and stops.

BACKGROUND: Freezing of gait (FOG) is an episodic and highly disabling symptom of Parkinson's Disease (PD). Traditionally, FOG assessment relies on time-consuming visual inspection of camera footage. Therefore, previous studies have proposed portable and automated solutions to annotate FOG. However, automated FOG assessment is challenging due to gait variability caused by medication effects and varying FOG-provoking tasks. Moreover, whether automated approaches can differentiate FOG from typical everyday movements, such as volitional stops, remains to be determined. To address these questions, we evaluated an automated FOG assessment model with deep learning (DL) based on inertial measurement units (IMUs). We assessed its performance trained on all standardized FOG-provoking tasks and medication states, as well as on specific tasks and medication states. Furthermore, we examined the effect of adding stopping periods on FOG detection performance. METHODS: Twelve PD patients with self-reported FOG (mean age 69.33 ± 6.02 years) completed a FOG-provoking protocol, including timed-up-and-go and 360-degree turning-in-place tasks in On/Off dopaminergic medication states with/without volitional stopping. IMUs were attached to the pelvis and both sides of the tibia and talus. A temporal convolutional network (TCN) was used to detect FOG episodes. FOG severity was quantified by the percentage of time frozen (%TF) and the number of freezing episodes (#FOG). The agreement between the model-generated outcomes and the gold standard experts' video annotation was assessed by the intra-class correlation coefficient (ICC). RESULTS: For FOG assessment in trials without stopping, the agreement of our model was strong (ICC (%TF) = 0.92 [0.68, 0.98]; ICC(#FOG) = 0.95 [0.72, 0.99]). Models trained on a specific FOG-provoking task could not generalize to unseen tasks, while models trained on a specific medication state could generalize to unseen states. For assessment in trials with stopping, the agreement of our model was moderately strong (ICC (%TF) = 0.95 [0.73, 0.99]; ICC (#FOG) = 0.79 [0.46, 0.94]), but only when stopping was included in the training data. CONCLUSION: A TCN trained on IMU signals allows valid FOG assessment in trials with/without stops containing different medication states and FOG-provoking tasks. These results are encouraging and enable future work investigating automated FOG assessment during everyday life.

Patterns and assessment of spastic hemiplegic gait.

Hemiparetic gait disorders are common in stroke survivors. A circumductory gait is often considered the typical hemiparetic gait. In clinical practice, a wide spectrum of abnormal gait patterns is observed, depending on the severity of weakness and spasticity, and the anatomical distribution of spasticity. Muscle strength is the key determinant of gait disorders in hemiparetic stroke survivors. Spasticity and its associated involuntary activation of synergistic spastic muscles often alter posture of involved joint(s) and subsequently the alignment of hip, knee, and ankle joints, resulting in abnormal gait patterns. Due to combinations of various levels of muscle weakness and spasticity and their interactions with ground reaction force, presentations of gait disorders are variable. From a neuromechanical perspective, a stepwise visual gait analysis approach is proposed to identify primary underlying causes. In this approach, the pelvic and hip joint movement is examined first. The pelvic girdle constitutes three kinematic determinants. Its abnormality determines the body vector and compensatory kinetic chain reactions in the knee and ankle joints. The second step is to assess the ankle and foot complex abnormality. The last step is to examine abnormality of the knee joint. Assessment of muscle strength and spasticity of hip, knee, and ankle/foot joints needs to be performed before these steps. Lidocaine nerve blocks can be a useful diagnostic tool. Recognizing different patterns and identifying the primary causes are critical to developing clinical interventions to improve gait functions.

Update on the reliability of gait analysis interpretation in cerebral palsy: Inter-institution agreement.

BACKGROUND: Studies have shown good reliability for gait analysis interpretation among surgeons from the same institution. However, reliability among surgeons from different institutions remains to be determined. RESEARCH QUESTION: Is gait analysis interpretation by surgeons from different institutions as reliable as it is for surgeons from the same institution? METHODS: Gait analysis data for 67 patients with cerebral palsy (CP) were reviewed prospectively by two orthopedic surgeons from different institutions in the same state, each with > 10 years' experience interpreting gait analysis data. The surgeons identified gait problems and made treatment recommendations for each patient using a rating form. Percent agreement between raters was calculated for each problem and treatment, and compared to expected agreement based on chance using Cohen's kappa. RESULTS: For problem identification, the greatest agreement was seen for equinus (85% agreement), calcaneus (88%), in-toeing (89%), and out-toeing (90%). Agreement for the remaining problems ranged between 66-78%. Percent agreement was significantly higher than expected due to chance for all issues (p ≤ 0.01) with modest kappa values ranging from 0.12 to 0.51. Agreement between surgeons for treatment recommendations was highest for triceps surae lengthening (89% agreement), tibial derotation osteotomy (90%), and foot osteotomy (87%). Agreement for the remaining treatments ranged between 72-78%. Percent agreement for all treatments was significantly higher than the expected values (p ≤ 0.002) with modest kappa values ranging from 0.22 to 0.52. SIGNIFICANCE: Previous research established that computerized gait analysis data interpretation is reliable for surgeons within a single institution. The current study demonstrates that gait analysis interpretation can also be reliable among surgeons from different institutions. Future research should examine reliability among physicians from more institutions to confirm these results.

Force Control Issues in Upper and Lower Limbs in Parkinson's Disease and Freezing of Gait.

Parkinson's Disease (PD) has been found to cause force control deficits in upper and lower limbs. About 50% of patients with advanced PD develop a debilitating symptom called freezing of gait (FOG), which has been linked to force control problems in the lower limbs, and some may only have a limited response to the gold standard pharmaceutical therapy, levodopa, resulting in partially levodopa-responsive FOG (PLR-FOG). There has been limited research on investigating upper-limb force control in people with PD with PLR-FOG, and without FOG. In this pilot study, force control was explored using an upper-and-lower-limb haptics-enabled robot in a reaching task while people with PD with and without PLR-FOG were on their levodopa medication. A healthy control group was used for reference, and each cohort completed the task at three different levels of assistance provided by the robot. Similar significant proportional force control deficits were found in the upper and lower limbs in patients with PLR-FOG versus those without FOG. Some aspects of force control were found to be retained, including an ability to increase or decrease force in response to changes in resistance while completing a reaching task. Overall, these results suggest there are force control deficits in both the upper and lower limbs in people with PLR-FOG.

Detection and prediction of freezing of gait with wearable sensors in Parkinson's disease.

Freezing of gait (FoG) is one of the most distressing symptoms of Parkinson's Disease (PD), commonly occurring in patients at middle and late stages of the disease. Automatic and accurate FoG detection and prediction have emerged as a promising tool for long-term monitoring of PD and implementation of gait assistance systems. This paper reviews the recent development of FoG detection and prediction using wearable sensors, with attention on identifying knowledge gaps that need to be filled in future research. This review searched the PubMed and Web of Science databases to collect studies that detect or predict FoG with wearable sensors. After screening, 89 of 270 articles were included. The data description, extracted features, detection/prediction methods, and classification performance were extracted from the articles. As the number of papers of this area is increasing, the performance has been steadily improved. However, small datasets and inconsistent evaluation processes still hinder the application of FoG detection and prediction with wearable sensors in clinical practice.

Utility of serum neurofilament light chain and glial fibrillary acidic protein as diagnostic biomarkers of freezing of gait in Parkinson's disease.

Freezing of gait (FOG) is one of the most distressing features of Parkinson's disease (PD), increasing the risks of fractures and seriously affecting patients' quality of life. We aimed to examine the potential diagnostic roles of serum neurofilament light chain (NFL) and glial fibrillary acidic protein (GFAP) in PD patients with FOG (PD-FOG). We included 99 patients, comprising 54 PD patients without FOG (PD-NoFOG), 45 PD-FOG and 37 healthy controls (HCs). Our results indicated serum markers were significantly higher in PD-FOG and postural instability and gait difficulty (PIGD) motor subtype patients than in PD-NoFOG and non-PIGD subtype patients (P < 0.05), respectively. Patients with high concentrations of the markers NFL and GFAP had higher PIGD scores and greater FOG severity than those with low concentrations. Moreover, serum levels of both NFL and GFAP were significantly positively associated with age, FOG severity, PD-FOG status, and negatively associated with Mini-Mental State Examination (MMSE) scores. Logistic regression analysis identified serum levels of NFL and GFAP as independent risk factors for PD-FOG. Mediation analysis revealed that MMSE scores fully mediated the relationship between serum GFAP levels and FOG-Q scores, accounting for 33.33% of the total effects (indirect effect = 0.01, 95% CI 0.01-0.02). NFL levels differentiated PD-FOG from PD-NoFOG with reliable diagnostic accuracy (AUC 0.75, 95% CI 0.66-0.84), and the combination of NFL, GFAP, duration and MMSE scores demonstrated high accuracy (AUC 0.84, 95% CI 0.76-0.91). Our findings support the notion that NFL and GFAP may be potential biomarkers for the diagnosis of PD-FOG.

Neurological gait assessment.

Gait disorders are a common feature of neurological disease. The gait examination is an essential part of the neurological clinical assessment, providing valuable clues to a myriad of causes. Understanding how to examine gait is not only essential for neurological diagnosis but also for treatment and prognosis. Here, we review aspects of the clinical history and examination of neurological gait to help guide gait disorder assessment. We focus particularly on how to differentiate between common gait abnormalities and highlight the characteristic features of the more prevalent neurological gait patterns such as ataxia, waddling, steppage, spastic gait, Parkinson's disease and functional gait disorders. We also offer diagnostic clues for some unusual gait presentations, such as dystonic, stiff-person and choreiform gait, along with red flags that help differentiate atypical parkinsonism from Parkinson's disease.

A Virtual Reality (VR) based Comprehensive Freezing of Gait (FOG) Neuro-electrophysiologic Evaluation System for People with Parkinson's Disease (PD).

Although Freezing of gait (FOG) is one of the most frustrating phenomena for people with Parkinson's Disease (PD), especially in their advanced stage, it is one of the least explained syndromes. The current studies only showed beta oscillations existed in frontal cortex-basal ganglia networks. Further studies need to be carried out. However, simultaneously recording neuro-electrophysiologic signals during walking is always a challenge, especially for Electroencephalogram (EEG) and Local Field Potential (LFP). This paper demonstrated a Virtual Reality (VR) based system which can trigger FOG and record biological signals at the same time. Moreover, the utilisation of VR will significantly decrease space requirements. It will provide a safer and more convenient evaluation environment for future participants. One participant with PD helped to validate the feasibility of the system. The result showed that both EEG and LFP could be recorded at the same time with trigger markers. This system design can be used to trigger freezing episodes in the controlled environment, differentiate subtypes of gait difficulties, and identify neural signatures associated with freezing episodes.Clinical relevance - This paper proposed a VR-based comprehensive FOG neuro-electrophysiologic evaluation system for people with PD. It had the advantages of minimum space requirement and wireless LFP data collection without externalised leads. This paper was to indicate a larger study which would formally recruit larger populations with PD and FOG. Future studies would explore FOG-related brain network coherence.

Factors influencing the level of stigma in Parkinson's disease in western Turkey.

Background and purpose:

Stigma is a widespread phenomenon in Parkinson’s disease (PD) and has been shown to affect the quality of life of individuals. This study aims to assess the level of stigma and identify the factors contributing to stigma in patients with PD in Turkey.

A total of 142 patients diagno­sed with PD between June 2022 and March 2023 were included in the study. Sociodemographic data including age, gender, marital status, education level, and duration of PD were collected using a sociodemographic information form. Motor symptom severity was assessed using the Unified Parkinson’s Disease Rating Scale (UPDRS part III). The disease stage was determined using the Hoehn and Yahr scale. Participants were classified as PIGD (postural instability/gait difficulty) or TD (tremor dominant) based on the UPDRS score. Patients with a UPDRS ratio greater than or equal to 1.5 were classified as TD, while subjects with a ratio less than or equal to 1.0 were classified as PIGD. Ratios between 1.0 and 1.5 were classified as mixed type. Depression was assessed using the Hamilton Depression Rating Scale (HAM-D), while stigma was measured using the Chronic Illness Anticipated Stigma Scale (CIASS) and the stigma sub-scale of the 39-item Parkinson’s Disease Questionnaire (PDQ-39 stigma sub-scale).

The mean score on the stigma sub-scale of the PDQ-39 was 7.60±4.39, while the mean total stigma score on the CIASS was 1.37±0.39. Our results indicated that stigma was more prevalent among patients with PD with the TD motor subtype, younger age, shorter disease duration, higher level of disability, and presence of depression symptoms.

Our study highlights the association between stigma and disease progression, duration, and depressive symptoms in patients with PD in western Turkey.

Classifying Tremor Dominant and Postural Instability and Gait Difficulty Subtypes of Parkinson's Disease from Full-Body Kinematics.

Characterizing motor subtypes of Parkinson's disease (PD) is an important aspect of clinical care that is useful for prognosis and medical management. Although all PD cases involve the loss of dopaminergic neurons in the brain, individual cases may present with different combinations of motor signs, which may indicate differences in underlying pathology and potential response to treatment. However, the conventional method for distinguishing PD motor subtypes involves resource-intensive physical examination by a movement disorders specialist. Moreover, the standardized rating scales for PD rely on subjective observation, which requires specialized training and unavoidable inter-rater variability. In this work, we propose a system that uses machine learning models to automatically and objectively identify some PD motor subtypes, specifically Tremor-Dominant (TD) and Postural Instability and Gait Difficulty (PIGD), from 3D kinematic data recorded during walking tasks for patients with PD (MDS-UPDRS-III Score, 34.7 ± 10.5, average disease duration 7.5 ± 4.5 years). This study demonstrates a machine learning model utilizing kinematic data that identifies PD motor subtypes with a 79.6% F1 score (N = 55 patients with parkinsonism). This significantly outperformed a comparison model using classification based on gait features (19.8% F1 score). Variants of our model trained to individual patients achieved a 95.4% F1 score. This analysis revealed that both temporal, spectral, and statistical features from lower body movements are helpful in distinguishing motor subtypes. Automatically assessing PD motor subtypes simply from walking may reduce the time and resources required from specialists, thereby improving patient care for PD treatments. Furthermore, this system can provide objective assessments to track the changes in PD motor subtypes over time to implement and modify appropriate treatment plans for individual patients as needed.

Freezing of gait in Parkinson's disease: Classification using computational intelligence.

Parkinson's disease (PD) is a neurodegenerative disease represented by the progressive loss of dopamine producing neurons, with motor and non-motor symptoms that may be hard to distinguish from other disorders. Affecting millions of people across the world, its symptoms include bradykinesia, tremors, depression, rigidity, postural instability, cognitive decline, and falls. Furthermore, changes in gait can be used as a primary diagnosis factor. A dataset is described that records data on healthy individuals and on PD patients, including those who experience freezing of gait, in both the ON and OFF-medication states. The dataset is comprised of data for four separate tasks: voluntary stop, timed up and go, simple motor task, and dual motor and cognitive task. Seven different classifiers are applied to two problems relating to this data. The first problem is to distinguish PD patients from healthy individuals, both overall and per task. The second problem is to determine the effectiveness of medication. A thorough analysis on the classifiers and their results is performed. Overall, multilayer perceptron and decision tree provide the most consistent results.

Normative dataset selection affects gait profile scores of children with cerebral palsy.

INTRODUCTION: The Gait Profile Score (GPS) requires a comparative dataset, to identify altered mechanics in persons with a gait abnormality. This gait index has been shown to be useful for identifying gait pathology prior to the assessment of treatment outcomes. Though studies have shown differences in kinematic normative datasets between different testing sites, there is limited information available on the changes in GPS score based on normative dataset selection. The aim of this study was to quantify the influence of normative reference data from two institutions, on the GPS and Gait Variable Scores (GVS), calculated on the same group of patients with Cerebral Palsy. METHODS: Seventy patients (Avg. age: 12.1 ± 2.9) diagnosed with CP underwent gait analysis during walking at a self-selected speed at Scottish Rite for Children (SRC). GPS and GVS scores were determined using normative kinematic data at a self-selected speed from, 83 typically developing children ages 4-17 from Gillette, and the same age range of children from SRC's normative dataset. Average normalized speed was compared between institutions. Signed rank tests were performed on the GPS and GVS scores using each institution's dataset. Spearman's correlations between scores using SRC and Gillette were determined within GMFCS level. RESULTS: Normalized speed was comparable between each institution's datasets. Within each GMFCS level, significant differences when using SRC vs. Gillette were found in most scores (p < 0.05). Scores were moderately to strongly correlated within each GMFCS level (range ρ = 0.448-0.998). CONCLUSIONS: Significant statistical differences were found in GPS and GVS scores but were within the range of previously reported variation across multiple sites. Caution and consideration may need to be taken when reporting GPS and GVS scores that are calculated utilizing different normative datasets as these scores may not be equivalent.

Gait Declines Differentially in, and Improves Prediction of, People with Parkinson's Disease Converting to a Freezing of Gait Phenotype.

BACKGROUND: Freezing of gait (FOG) is a debilitating, variably expressed motor symptom in people with Parkinson's disease (PwPD) with limited treatments. OBJECTIVE: To determine if the rate of progression in spatiotemporal gait parameters in people converting from a noFOG to a FOG phenotype (FOGConv) was faster than non-convertors, and determine if gait parameters can help predict this conversion. METHODS: PwPD were objectively monitored longitudinally, approximately every 6 months. Non-motor assessments were performed at the initial visit. Steady-state gait in the levodopa ON-state was collected using a gait mat (Protokinetics) at each visit. The rate of progression in 8 spatiotemporal gait parameters was calculated. FOG convertors (FOGConv) were classified if they did not have FOG at initial visit and developed FOG at a subsequent visit. RESULTS: Thirty freezers (FOG) and 30 non-freezers were monitored an average of 3.5 years, with 10 non-freezers developing FOG (FOGConv). FOGConv and FOG had faster decline in mean stride-length, swing-phase-percent, and increase in mean total-double-support percent, coefficient of variability (CV) foot-strike-length and CV swing-phase-percent than the remaining non-freezers (noFOG). On univariate modeling, progression rates of mean stride-length, stride-velocity, swing-phase-percent, total-double-support-percent and of CV swing-phase-percent had high discriminative power (AUC > 0.83) for classification of the FOGConv and noFOG groups. CONCLUSION: FOGConv had a faster temporal decline in objectively quantified gait than noFOG, and progression rates of spatiotemporal gait parameters were more predictive of FOG phenotype conversion than initial (static) parameters Objectively monitoring gait in disease prediction models may help define FOG prone groups for testing putative treatments.

Multi-muscle synergies in preparation for gait initiation in Parkinson's disease.

OBJECTIVE: We investigated changes in indices of muscle synergies prior to gait initiation and the effects of gaze shift in patients with Parkinson's disease (PD). A long-term objective of the study is to develop a method for quantitative assessment of gait-initiation problems in PD. METHODS: PD patients without clinical signs of postural instability and two control groups (age-matched and young) performed a gait initiation task in a self-paced manner, with and without a quick prior gaze shift produced by turning the head. Muscle groups with parallel scaling of activation levels (muscle modes) were identified as factors in the muscle activation space. Synergy index stabilizing center of pressure trajectory in the anterior-posterior and medio-lateral directions (indices of stability) was quantified in the muscle mode space. A drop in the synergy index in preparation to gait initiation (anticipatory synergy adjustment, ASA) was quantified. RESULTS: Compared to the control groups, PD patients showed significantly smaller synergy indices and ASA for both directions of the center of pressure shift. Both PD and age-matched controls, but not younger controls, showed detrimental effects of the prior gaze shift on the ASA indices. CONCLUSIONS: PD patients without clinically significant posture or gait disorders show impaired stability of the center of pressure and its diminished adjustment during gait initiation. SIGNIFICANCE: The indices of stability and ASA may be useful to monitor pre-clinical gait disorders, and lower ASA may be relevant to emergence of freezing of gait in PD.

Three-dimensional gait analysis of the effect of directional steering on gait in patients with Parkinson's disease.

INTRODUCTION: Deep Brain Stimulation (DBS) is an option to treat advanced Parkinson's Disease (PD), but can cause gait disturbance due to stimulation side efffects. This study aims to evaluate the objective effect of directional current steering by DBS on gait performance in PD, utilizing a three-dimensional gait analysis system. METHODS: Eleven patients diagnosed with PD and were implanted with directional lead were recruited. The direction of the pyramidal tract (identified by the directional mode screening) was set as 0°. Patients performed the six-meter-walk test and the time up-and-go (TUG) test while an analysis system recorded gait parameters utilizing a three-dimensional motion capture camera. The gait parameters were measured for the baseline, the directional steering at eight angles (0°, 45°, 90°, 135°, 180°, 225°, 270°, and 315°), and the conventional ring mode with 1, 2, and 3 mA. Pulse width and frequency were fixed. Placebo stimulation (0 mA) was used for a control. RESULTS: Eleven patients completed the study. No significant difference were observed between gait parameters during the directional, baseline, placebo, or ring modes during the six-meter-walk test (p > 0.05). During the TUG test, stride length was significantly different between 0° and other directions (p < 0.001), but no significant differences were observed for the other gait parameters. Stride width was non-significantly narrower in the direction of 0°. CONCLUSION: Controlling stimulation using directional steering may improve gait in patients with PD, while avoiding pyramidal side effects.

A Single-Sensor Approach to Quantify Gait in Patients with Hereditary Spastic Paraplegia.

Hereditary spastic paraplegia (HSP) is characterised by progressive lower-limb spasticity and weakness resulting in ambulation difficulties. During clinical practice, walking is observed and/or assessed by timed 10-metre walk tests; time, feasibility, and methodological reliability are barriers to detailed characterisation of patients' walking abilities when instrumenting this test. Wearable sensors have the potential to overcome such drawbacks once a validated approach is available for patients with HSP. Therefore, while limiting patients' and assessors' burdens, this study aims to validate the adoption of a single lower-back wearable inertial sensor approach for step detection in HSP patients; this is the first essential algorithmic step in quantifying most gait temporal metrics. After filtering the 3D acceleration signal based on its smoothness and enhancing the step-related peaks, initial contacts (ICs) were identified as positive zero-crossings of the processed signal. The proposed approach was validated on thirteen individuals with HSP while they performed three 10-metre tests and wore pressure insoles used as a gold standard. Overall, the single-sensor approach detected 794 ICs (87% correctly identified) with high accuracy (median absolute errors (mae): 0.05 s) and excellent reliability (ICC = 1.00). Although about 12% of the ICs were missed and the use of walking aids introduced extra ICs, a minor impact was observed on the step time quantifications (mae 0.03 s (5.1%), ICC = 0.89); the use of walking aids caused no significant differences in the average step time quantifications. Therefore, the proposed single-sensor approach provides a reliable methodology for step identification in HSP, augmenting the gait information that can be accurately and objectively extracted from patients with HSP during their clinical assessment.

Combination of Clinical and Gait Measures to Classify Fallers and Non-Fallers in Parkinson's Disease.

Although the multifactorial nature of falls in Parkinson's disease (PD) is well described, optimal assessment for the identification of fallers remains unclear. Thus, we aimed to identify clinical and objective gait measures that best discriminate fallers from non-fallers in PD, with suggestions of optimal cutoff scores. METHODS: Individuals with mild-to-moderate PD were classified as fallers (n = 31) or non-fallers (n = 96) based on the previous 12 months' falls. Clinical measures (demographic, motor, cognitive and patient-reported outcomes) were assessed with standard scales/tests, and gait parameters were derived from wearable inertial sensors (Mobility Lab v2); participants walked overground, at a self-selected speed, for 2 min under single and dual-task walking conditions (maximum forward digit span). Receiver operating characteristic curve analysis identified measures (separately and in combination) that best discriminate fallers from non-fallers; we calculated the area under the curve (AUC) and identified optimal cutoff scores (i.e., point closest-to-(0,1) corner). RESULTS: Single gait and clinical measures that best classified fallers were foot strike angle (AUC = 0.728; cutoff = 14.07°) and the Falls Efficacy Scale International (FES-I; AUC = 0.716, cutoff = 25.5), respectively. Combinations of clinical + gait measures had higher AUCs than combinations of clinical-only or gait-only measures. The best performing combination included the FES-I score, New Freezing of Gait Questionnaire score, foot strike angle and trunk transverse range of motion (AUC = 0.85). CONCLUSION: Multiple clinical and gait aspects must be considered for the classification of fallers and non-fallers in PD.

Video-Based Quantification of Gait Impairments in Parkinson's Disease Using Skeleton-Silhouette Fusion Convolution Network.

Gait impairments are among the most common hallmarks of Parkinson's disease (PD), usually appearing in the early stage and becoming a major cause of disability with disease progression. Accurate assessment of gait features is critical to personalized rehabilitation for patients with PD, yet difficult to be routinely carried out as clinical diagnosis using rating scales relies heavily on clinical experience. Moreover, the popular rating scales cannot ensure fine quantification of gait impairments for patients with mild symptoms. Developing quantitative assessment methods that can be used in natural and home-based environments is highly demanded. In this study, we address the challenges by developing an automated video-based Parkinsonian gait assessment method using a novel skeleton-silhouette fusion convolution network. In addition, seven network-derived supplementary features, including critical aspects of gait impairment (gait velocity, arm swing, etc.), are extracted to provide continuous measures enhancing low-resolution clinical rating scales. Evaluation experiments were conducted on a dataset collected with 54 patients with early PD and 26 healthy controls. The results show that the proposed method accurately predicted the patients' unified Parkinson's disease rating scale (UPDRS) gait scores (71.25% match on clinical assessment) and discriminated between PD patients and healthy subjects with a sensitivity of 92.6%. Moreover, three proposed supplementary features (i.e., arm swing amplitude, gait velocity, and neck forward bending angle) turned out to be effective gait dysfunction indicators with Spearman correlation coefficients of 0.78, 0.73, and 0.43 matching the rating scores, respectively. Since the proposed system requires only two smartphones, it holds a significant benefit for home-based quantitative assessment of PD, especially for detecting early-stage PD. Furthermore, the proposed supplementary features can enable high-resolution assessments of PD for providing subject-specific accurate treatments.