In this paper, we present a novel approach to early diagnosis, through a video processing-based approach, of the presence of neonatal seizures. In particular, image processing and gesture recognition techniques are first used to characterize typical gestures of neonatal seizures. More precisely, gesture trajectories are characterized by extracting some relevant features. In particular, selecting the point with the maximum amplitude of the optical flow vector of the video frame sequence, during a newborn movement, is selected and then tracked through an algorithm based on template matching and optical flow. The observed features are then clustered using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. The proposed approach allows to efficiently differentiate pathological repetitive movements (e.g., clonic and subtle seizures) from random ones.

Video processing-based detection of neonatal seizures by trajectory features clustering / G. Kouamou; F. Lofino; G. Ferrari; R. Raheli; F. Pisani. - (2012), pp. 3456-3460. ((Intervento presentato al convegno IEEE Int. Conf. on Communications (ICC 2012), Selected Areas in Communications Symposium ('ICC'12 SAC'), E-Health Track tenutosi a Ottawa, Canada nel giugno [10.1109/ICC.2012.6364396].

Video processing-based detection of neonatal seizures by trajectory features clustering

KOUAMOU NTONFO, Guy Mathurin;FERRARI, Gianluigi;RAHELI, Riccardo;PISANI, Francesco
2012

Abstract

In this paper, we present a novel approach to early diagnosis, through a video processing-based approach, of the presence of neonatal seizures. In particular, image processing and gesture recognition techniques are first used to characterize typical gestures of neonatal seizures. More precisely, gesture trajectories are characterized by extracting some relevant features. In particular, selecting the point with the maximum amplitude of the optical flow vector of the video frame sequence, during a newborn movement, is selected and then tracked through an algorithm based on template matching and optical flow. The observed features are then clustered using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. The proposed approach allows to efficiently differentiate pathological repetitive movements (e.g., clonic and subtle seizures) from random ones.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11381/2439415
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 5
  • ???jsp.display-item.citation.isi??? 0
social impact