Support Vector Machines based multidimensional signals classification for fetal activity characterization - Université Toulouse 1 Capitole Accéder directement au contenu
Communication Dans Un Congrès Année : 2011

Support Vector Machines based multidimensional signals classification for fetal activity characterization

Résumé

Electronic fetal monitoring may be required during the whole pregnancy to closely monitor specific fetal and maternal disorders. Currently used methods suffer from many limitations and are not sufficient to evaluate fetal asphyxia. Fetal activity parameters such as movements, heart rate and associated parameters are essential indicators of the fetus well being, and no current device gives a simultaneous and sufficient estimation of all these parameters to evaluate the fetus well-being. We built for this purpose, a multi-transducer-multi-gate Doppler system and developed dedicated signal processing techniques for fetal activity parameter extraction in order to investigate fetus's asphyxia or well-being through fetal activity parameters. To reach this goal, this paper shows preliminary feasibility of separating normal and compromised fetuses using our system. To do so, data set consisting of two groups of fetal signals (normal and compromised) has been established and provided by physicians. From estimated parameters an instantaneous Manning-like score, referred to as ultrasonic score was introduced and was used together with movements, heart rate and associated parameters in a classification process using Support Vector Machines (SVM) method. The influence of the fetal activity parameters and the performance of the SVM were evaluated using the computation of sensibility, specificity, percentage of support vectors and total classification accuracy. We showed our ability to separate the data into two sets : normal fetuses and compromised fetuses and obtained an excellent matching with the clinical classification performed by physician.
Fichier non déposé

Dates et versions

hal-03146727 , version 1 (19-02-2021)

Identifiants

Citer

Sophie Ribes, Voicu Iulian, Jean Marc Girault, Morgane Fournier-Massignan, Franck Perrotin, et al.. Support Vector Machines based multidimensional signals classification for fetal activity characterization. SPIE Medical Imaging 2011, Society of Photo-Optical Instrumentation Engineers (SPIE), Feb 2011, Orlando, FL, United States. pp.(electronic medium), ⟨10.1117/12.877680⟩. ⟨hal-03146727⟩
18 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More