A hierarchical multimodal system for motion analysis in patients with epilepsy

Ahmedt-Aristizabal, David; Fookes, Clinton; Denman, Simon; Nguyen, Kien; Fernando, Tharindu; Sridharan, Sridha; Dionisio, Sasha

doi:10.1016/j.yebeh.2018.07.028

Cited by 29 publications

(27 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our study shows that the analysis of patients considering all body motions simultaneously is viable using the existing monitoring technology in the hospital. As it was confirmed in previous studies [5,6], the performance of LOSO-CV schemes is strictly related to semiological patterns contained in the dataset. This is evident from the difference in classification performance between patients based on the deep learning architecture.…”

Section: Discussionsupporting

confidence: 79%

“…A cross-validation evaluation is performed to verify the flexibility of the system to capture and quantify human motion behavior, and to demonstrate the sensitivity and specificity of each MoCap library to distinguish known behaviors. We attempt to discriminate between clinical presentations, which may help to address limitations of current automated approaches for semiology assessment [6] by incorporating the identification of unusual clinical manifestations within the system. Details of each phase are described in the following subsections.…”

Section: Methodsmentioning

confidence: 99%

“…Verification performance using the trained architecture is shown in the "Test Accuracy CNN-LSTM" column, which reached an average accuracy of 66.48% and 62.19% for patients with MTLE and ETLE, respectively. These results represent a large variability in the verification of each patient because of the high variation in the data when classifying the semiology of particular patients, which is one of the significant limitations of such strategies [6].…”

Section: Verification Of Mocap Libraries and Identification Of Aberramentioning

confidence: 99%

“…Given the importance of body motion patterns in the assessment of epilepsy, prior works have demonstrated that automated analysis of semiological patterns based on computer vision can support diagnosis by standard and objective assessment methods among evaluators [3,4]. Previously, we proposed a facial semiology analysis approach to identify patients with mesial temporal lobe epilepsy (MTLE) [5], and a hierarchical multi-modal system to quantify and classify patients with MTLE and extra-temporal lobe epilepsy (ETLE) based on face, body and hand motions [6].…”

Section: Introductionmentioning

confidence: 99%

“…data for a new patient) does not come from one of the training categories, the trained model is of no use. This is evident in low performance when classifying the semiology of unseen patients with potential aberrant or unusual seizures (leaveone-subject-out cross-validation scheme) [6]. In this scenario, the ictal patterns from the test patients with MTLE or ETLE are not strongly correlated to the semiological patterns of other patients previously observed and contained within the training and validation data.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Aberrant epileptic seizure identification: A computer vision perspective

Ahmedt-Aristizabal

Fookes

Denman

et al. 2019

Seizure

Self Cite

View full text Add to dashboard Cite

A B S T R A C TPurpose: The recent explosion of artificial intelligence techniques in video analytics has highlighted the clinical relevance in capturing and quantifying semiology during epileptic seizures; however, we lack an automated anomaly identification system for aberrant behaviors. In this paper, we describe a novel system that is trained with known clinical manifestations from patients with mesial temporal and extra-temporal lobe epilepsy and presents aberrant semiology to physicians. Methods: We propose a simple end-to-end-architecture based on convolutional and recurrent neural networks to extract spatiotemporal representations and to create motion capture libraries from 119 seizures of 28 patients. The cosine similarity distance between a test representation and the libraries from five aberrant seizures separate to the main dataset is subsequently used to identify test seizures with unusual patterns that do not conform to known behavior. Results: Cross-validation evaluations are performed to validate the quantification of motion features and to demonstrate the robustness of the motion capture libraries for identifying epilepsy types. The system to identify unusual epileptic seizures successfully detects out of the five seizures categorized as aberrant cases. Conclusions: The proposed approach is capable of modeling clinical manifestations of known behaviors in natural clinical settings, and effectively identify aberrant seizures using a simple strategy based on motion capture libraries of spatiotemporal representations and similarities between hidden states. Detecting anomalies is essential to alert clinicians to the occurrence of unusual events, and we show how this can be achieved using prelearned database of semiology stored in health records.

show abstract

Section: Discussionsupporting

confidence: 79%

Section: Methodsmentioning

confidence: 99%

Section: Verification Of Mocap Libraries and Identification Of Aberramentioning

confidence: 99%