Cerebral Palsy Risk Identification System

点击翻译按钮获取中文摘要

PROJECT SUMMARY AND ABSTRACT [ Pediatric specialists are often required to identify infants who are likely to suffer poor neurodevelopmental outcome, including Cerebral Palsy (CP). CP is the most common developmental disability among children in the United States and results from several factors, including low weight for gestational age, premature birth, and stroke. Although MRI and cranial ultrasound (cUS) provide valuable structural information in the preterm period, they have moderate sensitivity to CP and require transportation of the infant. Over the past 20 years, numerous studies have validated the clinical potential of General Movement Assessment (GMA) for CP risk identification. During the early period, (23 weeks to 36 weeks gestational age), the presence of Cramped Synchronized General Movements (CSGMs), has demonstrated very high sensitivity and specificity for CP, conjointly ranging from 80%-98%. CSGMs are assessed while preterm infants are still in an acute care facility (NICU) and can inform the clinician independently, and in combination with cUS and MRI. Despite its potential, GMA is available in only a few clinical centers, as adoption and routine application depend on lengthy, cost-intensive observation and availability of specially trained raters. A Cerebral Palsy Risk Identification System (CPRIS) is proposed that will automate GMA for bedside evaluations in both preterm and postterm periods. The CPRIS constitutes a key enabling technology not only for routine risk identification, but also for establishing disease trajectory and potentially differentiating CP subtypes and assessing efficacy of emerging treatments along the early developmental continuum. Preliminary studies at UC Irvine have demonstrated that GMA analysis for CSGMs can be automated by quantifying infant limb movement using highly miniaturized, 3-axis wireless accelerometers and classifying CSGMs using a patented Markov-type approach that merges an application-specific Erlang-Cox state transition model with a Dynamic Bayesian Network (“EC-DBN”), treating instantaneous machine learning classification values as observations and explicitly modeling CSGM (and non-CSGM) duration and interval. In Phase I, this approach will be utilized in a comparative evaluation of two movement measurement modalities to determine which has the best overall performance and clinical utility at three leading NICU centers. Infant movement data will be concurrently acquired using an advanced, second generation prototype wireless accelerometer system (CPRIS-A) and a high definition 3D (infrared) optical camera (CPRIS-O). The optical modality offers significant potential advantages as it requires no infant contact and can monitor unattended, intermittently, over weeks or months. However, its potential for GMA automation must be systematically evaluated. Classifier results from both modalities will be compared to expert rater consensus in 80 preterm infants. The primary outcome will be CSGM identification accuracy, as determined by ROC-AUC analyses, with a threshold for success of 0.85. Additional comparative performance measures include reliability and practicability in the NICU environment. An Advisory Committee of experts in the fields of neonatology, pediatrics and cerebral palsy will evaluate project results and advise on the clinical potential of each modality. ]