Dataset Information

Does expert knowledge improve automatic probabilistic classification of gait joint motion patterns in children with cerebral palsy?

ABSTRACT:

Background

This study aimed to improve the automatic probabilistic classification of joint motion gait patterns in children with cerebral palsy by using the expert knowledge available via a recently developed Delphi-consensus study. To this end, this study applied both Naïve Bayes and Logistic Regression classification with varying degrees of usage of the expert knowledge (expert-defined and discretized features). A database of 356 patients and 1719 gait trials was used to validate the classification performance of eleven joint motions.

Hypotheses

Two main hypotheses stated that: (1) Joint motion patterns in children with CP, obtained through a Delphi-consensus study, can be automatically classified following a probabilistic approach, with an accuracy similar to clinical expert classification, and (2) The inclusion of clinical expert knowledge in the selection of relevant gait features and the discretization of continuous features increases the performance of automatic probabilistic joint motion classification.

Findings

This study provided objective evidence supporting the first hypothesis. Automatic probabilistic gait classification using the expert knowledge available from the Delphi-consensus study resulted in accuracy (91%) similar to that obtained with two expert raters (90%), and higher accuracy than that obtained with non-expert raters (78%). Regarding the second hypothesis, this study demonstrated that the use of more advanced machine learning techniques such as automatic feature selection and discretization instead of expert-defined and discretized features can result in slightly higher joint motion classification performance. However, the increase in performance is limited and does not outweigh the additional computational cost and the higher risk of loss of clinical interpretability, which threatens the clinical acceptance and applicability.

SUBMITTER: De Laet T

PROVIDER: S-EPMC5453476 | biostudies-literature | 2017

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Does expert knowledge improve automatic probabilistic classification of gait joint motion patterns in children with cerebral palsy?

De Laet Tinne T Papageorgiou Eirini E Nieuwenhuys Angela A Desloovere Kaat K

PloS one 20170601 6

<h4>Background</h4>This study aimed to improve the automatic probabilistic classification of joint motion gait patterns in children with cerebral palsy by using the expert knowledge available via a recently developed Delphi-consensus study. To this end, this study applied both Naïve Bayes and Logistic Regression classification with varying degrees of usage of the expert knowledge (expert-defined and discretized features). A database of 356 patients and 1719 gait trials was used to validate the c ...[more]

PMID: 28570616

Similar Datasets

Project description:BackgroundSimultaneous motion of the knee and ankle joints is required for many activities including gait. We aimed to evaluate the influence of surgery involving tendons around the knee on ankle motion during gait in the sagittal plane in cerebral palsy patients.MethodsWe included data from 55 limbs in 34 patients with spastic cerebral palsy. Patients were followed up after undergoing only distal hamstring lengthening with or without additional rectus femoris transfer. The patients' mean age at the time of knee surgery was 11.2 ± 4.7 years, and the mean follow-up duration was 2.2 ± 1.5 years (range, 0.9-6.0 years). Pre- and postoperative kinematic variables that were extracted from three-dimensional gait analyses were then compared to assess changes in ankle motion after knee surgery. Outcome measures included ankle dorsiflexion at initial contact, peak ankle dorsiflexion during stance, peak ankle dorsiflexion during swing, and dynamic range of motion of the ankle. Various sagittal plane knee kinematics were also measured and used to predict ankle kinematics. A linear mixed model was constructed to estimate changes in ankle motion after adjusting for multiple factors.ResultsImprovement in total range of motion of the knee resulted in improved motion of the ankle joint. We estimated that after knee surgery, ankle dorsiflexion at initial contact, peak ankle dorsiflexion during stance, peak ankle dorsiflexion during swing, and dynamic range of motion of the ankle decreased, respectively, by 0.4° (p = 0.016), 0.6° (p < 0.001), 0.2° (p = 0.038), and 0.5° (p = 0.006) per degree increase in total range of motion of the knee after either knee surgery. Furthermore, dynamic range of motion of the ankle increased by 0.4° per degree increase in postoperative peak knee flexion during swing.ConclusionsImprovement in total knee range of motion was found to be correlated with improvement in ankle kinematics after surgery involving tendons around the knee. As motion of the knee and ankle joints is cross-linked, surgeons should be aware of potential changes in the ankle joint after knee surgery.

Project description:Experts recently identified 49 joint motion patterns in children with cerebral palsy during a Delphi consensus study. Pattern definitions were therefore the result of subjective expert opinion. The present study aims to provide objective, quantitative data supporting the identification of these consensus-based patterns. To do so, statistical parametric mapping was used to compare the mean kinematic waveforms of 154 trials of typically developing children (n = 56) to the mean kinematic waveforms of 1719 trials of children with cerebral palsy (n = 356), which were classified following the classification rules of the Delphi study. Three hypotheses stated that: (a) joint motion patterns with 'no or minor gait deviations' (n = 11 patterns) do not differ significantly from the gait pattern of typically developing children; (b) all other pathological joint motion patterns (n = 38 patterns) differ from typically developing gait and the locations of difference within the gait cycle, highlighted by statistical parametric mapping, concur with the consensus-based classification rules. (c) all joint motion patterns at the level of each joint (n = 49 patterns) differ from each other during at least one phase of the gait cycle. Results showed that: (a) ten patterns with 'no or minor gait deviations' differed somewhat unexpectedly from typically developing gait, but these differences were generally small (?3°); (b) all other joint motion patterns (n = 38) differed from typically developing gait and the significant locations within the gait cycle that were indicated by the statistical analyses, coincided well with the classification rules; (c) joint motion patterns at the level of each joint significantly differed from each other, apart from two sagittal plane pelvic patterns. In addition to these results, for several joints, statistical analyses indicated other significant areas during the gait cycle that were not included in the pattern definitions of the consensus study. Based on these findings, suggestions to improve pattern definitions were made.

Project description:During a Delphi consensus study, a new joint gait classification system was developed for children with cerebral palsy (CP). This system, whose reliability and content validity have previously been established, identified 49 distinct joint patterns. The present study aims to provide a first insight toward the construct validity and clinical relevance of this classification system. The retrospective sample of convenience consisted of 286 patients with spastic CP (3-18 years old, GMFCS levels I-III, 166 with bilateral CP). Kinematic and kinetic trials from three-dimensional gait analysis were classified according to the definitions of the Delphi study, and one classified trial was randomly selected for each included limb (n = 446). Muscle weakness and spasticity were assessed for different muscle groups acting around the hip, knee, and ankle. Subsequently, Pearson Chi square tests, Cramer's V, and adjusted standardized residuals were calculated to explore the strength and direction of the associations between the joint patterns, and the different patient-specific characteristics (i.e., age, GMFCS level, and topographical classification) or clinical symptoms (muscle weakness and spasticity). Patient-specific characteristics showed several significant associations with the patterns of different joints, but the strength of most identified associations was weak. Apart from the knee during stance phase and the pelvis in the sagittal plane, the results systematically showed that the patterns with "minor gait deviations" were the most frequently observed. These minor deviations were found significantly more often in limbs with a lower level of spasticity and good muscle strength. Several other pathological joint patterns were moderately associated with weakness or spasticity, including but not limited to "outtoeing" for weakness and "intoeing" for spasticity. For the joints in the sagittal plane, significantly stronger associations were found with muscle weakness and spasticity, possibly because most of the evaluated muscles in this study mainly perform sagittal plane motions. Remarkably, the hip patterns in the coronal plane did not associate significantly with any of the investigated variables. Although further validation is warranted, this study contributes to the construct validity of the joint patterns of the Delphi consensus study, by demonstrating their ability to distinguish between clinically relevant subgroups in CP.

Project description:The interaction between brain damage and motor function is not yet fully understood in children with spastic cerebral palsy (CP). Therefore, a semi-quantitative MRI (sqMRI) scale was used to explore whether identified brain lesions related to functional abilities and gait pathology in this population. A retrospective cohort of ambulatory children with spastic CP was selected [N = 104; 52 bilateral (bCP) and 52 unilateral (uCP)]. Extent and location-specific scores were defined according to the sqMRI scale guidelines. The gross motor function classification system (GMFCS), the gait profile score (GPS), GPSs per motion plane, gait variable scores (GVS) and multiple-joint (MJ) gait patterns were related to brain lesion scores. In all groups, the global total brain scores correlated to the GPS (total: r s = 0.404, p ≤ 0.001; bCP: r s = 0.335, p ≤ 0.05; uCP: r s = 0.493, p ≤ 0.001). The global total hemispheric scores correlated to the GMFCS (total: r s = 0.392, p ≤ 0.001; bCP: r s = 0.316, p ≤ 0.05; uCP: r s = 0.331, p ≤ 0.05). The laterality scores of the hemispheres in the total group correlated negatively to the GMFCS level (r s = -0.523, p ≤ 0.001) and the GVS-knee sagittal (r s = -0.311, p ≤ 0.01). Lesion location, for the total group demonstrated positive correlations between parietal lobe involvement and the GPS (r s = 0.321, p ≤ 0.001) and between periventricular layer damage and the GMFCS (r s = 0.348, p ≤ 0.001). Involvement of the anterior part of the corpus callosum (CC) was associated with the GVS-hip sagittal in all groups (total: r pb = 0.495, p ≤ 0.001; bCP: r pb = 0.357, p ≤ 0.05; uCP: r pb = 0.641, p ≤ 0.001). The global total hemispheric and laterality of the hemispheres scores differentiated between the minor and both the extension (p ≤ 0.001 and p ≤ 0.001) and flexion (p = 0.016 and p = 0.013, respectively) MJ patterns in the total group. Maximal periventricular involvement and CC intactness were associated with extension patterns (p ≤ 0.05 and p ≤ 0.001, respectively). Current findings demonstrated relationships between brain structure and motor function as well as pathological gait, in this cohort of children with CP. These results might facilitate the timely identification of gait pathology and, ultimately, guide individualized treatment planning of gait impairments in children with CP.

Dataset Information

Does expert knowledge improve automatic probabilistic classification of gait joint motion patterns in children with cerebral palsy?

Background

Hypotheses

Findings

Publications

Does expert knowledge improve automatic probabilistic classification of gait joint motion patterns in children with cerebral palsy?

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets