Dataset Information

Interpretability of Input Representations for Gait Classification in Patients after Total Hip Arthroplasty.

ABSTRACT: Many machine learning models show black box characteristics and, therefore, a lack of transparency, interpretability, and trustworthiness. This strongly limits their practical application in clinical contexts. For overcoming these limitations, Explainable Artificial Intelligence (XAI) has shown promising results. The current study examined the influence of different input representations on a trained model's accuracy, interpretability, as well as clinical relevancy using XAI methods. The gait of 27 healthy subjects and 20 subjects after total hip arthroplasty (THA) was recorded with an inertial measurement unit (IMU)-based system. Three different input representations were used for classification. Local Interpretable Model-Agnostic Explanations (LIME) was used for model interpretation. The best accuracy was achieved with automatically extracted features (mean accuracy Macc = 100%), followed by features based on simple descriptive statistics (Macc = 97.38%) and waveform data (Macc = 95.88%). Globally seen, sagittal movement of the hip, knee, and pelvis as well as transversal movement of the ankle were especially important for this specific classification task. The current work shows that the type of input representation crucially determines interpretability as well as clinical relevance. A combined approach using different forms of representations seems advantageous. The results might assist physicians and therapists finding and addressing individual pathologic gait patterns.

SUBMITTER: Dindorf C

PROVIDER: S-EPMC7471970 | biostudies-literature | 2020 Aug

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Interpretability of Input Representations for Gait Classification in Patients after Total Hip Arthroplasty.

Dindorf Carlo C Teufl Wolfgang W Taetz Bertram B Bleser Gabriele G Fröhlich Michael M

Sensors (Basel, Switzerland) 20200806 16

Many machine learning models show black box characteristics and, therefore, a lack of transparency, interpretability, and trustworthiness. This strongly limits their practical application in clinical contexts. For overcoming these limitations, Explainable Artificial Intelligence (XAI) has shown promising results. The current study examined the influence of different input representations on a trained model's accuracy, interpretability, as well as clinical relevancy using XAI methods. The gait of ...[more]

PMID: 32781583

Similar Datasets

Project description:Background and purpose - Gait analysis performed under increased physical demand may detect differences in gait between total (THA) versus resurfacing hip arthroplasty (RHA), which are not measured at normal walking speed. We hypothesized that patients after RHA would reach higher walking speeds and inclines compared with THA. Additionally, an RHA would enable a more natural gait when comparing the operated with the healthy contralateral hip. Patients and methods - From a randomized controlled trial comparing THA with RHA with at least 5 years' follow-up patients with a UCLA score of more than 3 points (n = 34) were included for an instrumented treadmill gait analysis. 25 patients with a unilateral implant (primary analysis-16 THA versus 9 RHA) and 9 patients with a bilateral implant (sub-analysis-n = 5 RHA + THA; n = 4 THA + THA). Spatiotemporal parameters, ground reaction forces, and range of motion were recorded at increasing walking speeds and inclines. Functional outcome scores were obtained. Results - At a normal walking speed of 1.1 m/s and at increasing inclines no differences were recorded in gait between the 2 groups with a unilateral hip implant. With increasing walking speed the RHA group reached a higher top walking speed (TWS) (adjusted difference 0.07 m/s, 95% CI -0.11 to 0.25) compared with THA. Additionally, RHA patients tolerated more weight on the operated side at TWS (155 N, CI 49-261) and as such weight-bearing approached the unaffected contralateral side. For the RHA group a "between leg difference" of 8 N (CI 3-245) was measured versus -129 N (CI -138 to -29) for THA (adjusted difference 144 N, CI 20-261). Hip flexion of the operated side at TWS was higher after RHA compared with THA (adjusted difference 8°, CI 1.7-14). Interpretation - In this study RHA patients reached a higher walking speed, and preserved a more normal weight acceptance and a greater range of hip flexion against their contralateral healthy leg as compared with patients with a THA.

Project description:PurposeLiterature regarding total hip arthroplasty for pediatric hip diseases is scarce. This review aims to portray the various orthopedic conditions of childhood that can lead to significant impairment of the hip joint and, ultimately, to total hip arthroplasty in adolescence and adulthood.MethodsIn total, 61 out of 3666 articles were selected according to (1) the diagnosis of one of the 12 pediatric hip pathologies (Legg-Perthes-Calvé disease, developmental dysplasia of the hip, slipped capital femoral epiphysis, neuromuscular hip dysplasia, post-traumatic avascular necrosis of the proximal femur, juvenile rheumatoid arthritis, achondroplasia, spondyloepiphyseal dysplasia, mucopolysaccharidosis, mucolipidosis, hip infections, and tumors) that required total hip arthroplasty; (2) minimum follow-up of 16 months; (3) assessed outcome with a clinical or radiologic score; (4) Methodological Items for Non-Randomized Studies quality score of 9 or higher. The following information for each pathology was retrieved: mean age at total hip arthroplasty, reason for total hip arthroplasty, type of total hip arthroplasty, surgical technique, mean follow-up, and outcomes.ResultsOverall, the mean age at total hip arthroplasty for pediatric hip disease is in the sixth and seventh decade, except for tumors and skeletal dysplasias. The reason for performing total hip arthroplasty is often osteoarthrosis and abnormal anatomy. Prosthesis types change based on patient's conditions and technological advances; custom-made implants are used for tumors, juvenile rheumatoid arthritis, and skeletal dysplasias; for other diseases, the most frequent are modular cementless implants. Outcomes are generally good, and all studies portray functional and pain improvements.ConclusionTotal hip arthroplasty is performed more frequently than in the past in patients with pediatric hip pathologies; it enhances patients' quality of life by reducing pain and improving function. However, revision rate in these patients is not negligible.

Dataset Information

Interpretability of Input Representations for Gait Classification in Patients after Total Hip Arthroplasty.

Publications

Interpretability of Input Representations for Gait Classification in Patients after Total Hip Arthroplasty.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets