Dynamical 2D-DFA for movement analysis in obstetrics

While perineal tears continues to occur in 90% of births worldwide, the PELVITRACK project aims at characterising perineum damages in order to predict them and thus to adapt obstetrical process and to improve perineal rehabilitation. With images and videos, that can be easily collected, predictive analysis based on texture descriptors could help perineal tears prevention. Fractal-based method have proved to be efficient at highlighting time series differences in terms of rugosity or complexity at different time scales. Among fractal methods, the Detrended Fluctuation Analysis (DFA) has mainly been applied to time series but some works have proposed extensions to images and videos. In this paper, 2D-DFA is performed on experimental perineal image sequence with the objective of perineal damage characterisation. At the global and local levels the preliminary results are encouraging and illustrate the suitability of 2D-DFA for movement analysis on images.