Per partum fetal asphyxia constitutes the third most common cause for neonatal mor- bidity and is responsible for a third of neonatal encephalopathies in term neonates. It is consecutive to a major perturbation of the fetal oxygenation process and lead to acidosis and then to brain lesions. To avoid fetal brain lesions, when asphyxia is suspected, operative deliveries are performed, which consist of either caesarean or instrumental extraction. Such operative deliveries significantly increase risks for maternal and infant trauma, infectious and thrombotic risk and complications for subsequent pregnancies (as uterine rupture with fetal death for the next pregnancy or severe haemorrhage due to placenta praevia and accrete) and increase maternal and neonatal mortality. Therefore, per partum fetal surveillance aiming at relevant detection of fetal asphyxia, while limiting the number of unnecessary (incorrectly decided) operative deliveries, constitutes an important challenge, both for public health and economical reasons, which justifies the significant research efforts devoted to it. Currently, clinical per partum fetal surveillance relies mostly on visual analysis of real-time on-line recordings of fetal heart rate. Decisions for operative deliveries are based on criteria elaborated by the International Federation of Gynecology and Obstetrics, yet still depend heavily on the obstetrician subjectivity. This analysis enables a reliable detection of acidosis where missed detections (False Negatives) must be avoided, at the price, though, of a high rate of false positive detections, leading to unneeded operative deliveries. Optimizing the False Negative versus False Positive trade-off is one of the key challenges of the FETUSES project. Fetal heart rate analysis provides a valuable index describing subject's health status and autonomous nervous system dynamics, and hence is a major research topic of world-wide interest. As an example, the American project PHYSIONET devotes a significant amount of international research effort to adult heart beat variability; the European STAN project elaborates on fetal monitoring performance impro- vements (the medical team involved in FETUSES is participating in it); and, two French teams study cardiac variability for prematures or animal fetus. FETUSES concentrates more specifically on the analysis of per partum fetal heart rate, in the aim to detect fetal acidosis. The FETUSES project relies deeply and crucially on the construction of a truly interdisciplinary research program, gathering skills and knowledges of both medical and statistical signal processing teams, which are recognized as world-leading experts in their fields. FETUSES will combine in a single research program a significant amount of knowledge on fetal heart rate and acidosis detection with a solid expertise in modern statistical signal processing tools. These interactions will be eased by the geographical vicinity, within the city of Lyon, of the research teams involved. These teams have already undertaken the drawing of a common research framework, whose formalization results in the present ANR proposal.

The goals of the FETUSES project consist of the developments of statistical signal processing tools dedicated to per partum fetal heart rate characterization and acidosis detection, and are organized as follows :

• Construction of a large database of per partum fetal heart rate recordings, which is well documented and of significant clinical value ;
• Developments of adaptive (i.e., data driven) algorithms to separate data into trend (decelerations induced by contractions) and fluctuation (cardiac variability) components and algorithms to characterize the non stationary and multifractal properties of per partum fetal heart rate ;
• Acidosis detection and assessment using the large database ;
• Algorithm implementations for performing tests in real clinical situations.