General Movement Assessment from videos of computed 3D infant body models is equally effective compared to conventional RGB video rating.

BACKGROUND: General Movement Assessment (GMA) is a powerful tool to predict Cerebral Palsy (CP). Yet, GMA requires substantial training challenging its broad implementation in clinical routine. This inspired a world-wide quest for automated GMA. AIMS: To test whether a low-cost, marker-less system for three-dimensional motion capture from RGB depth sequences using a whole body infant model may serve as the basis for automated GMA. STUDY DESIGN: Clinical case study at an academic neurodevelopmental outpatient clinic. SUBJECTS: Twenty-nine high risk infants were assessed at their clinical follow-up at 2-4 month corrected age (CA). Their neurodevelopmental outcome was assessed regularly up to 12-31 months CA. OUTCOME MEASURES: GMA according to Hadders-Algra by a masked GMA-expert of conventional and computed 3D body model ("SMIL motion") videos of the same GMs. Agreement between both GMAs was tested using dichotomous and graded scaling with Kappa and intraclass correlations, respectively. Sensitivity and specificity to predict CP at ≥12 months CA were assessed. RESULTS: Agreement of the two GMA ratings was moderate-good for GM-complexity (κ = 0.58; ICC = 0.874 [95%CI 0.730; 0.941]) and substantial-good for fidgety movements (FMs; Kappa = 0.78, ICC = 0.926 [95%CI 0.843; 0.965]). Five children were diagnosed with CP (four bilateral, one unilateral CP). The GMs of the child with unilateral CP were twice rated as mildly abnormal with FMs. GM-complexity and somewhat less FMs, of both conventional and SMIL motion videos predicted bilateral CP comparably to published literature. CONCLUSIONS: Our computed infant 3D full body model is an attractive starting point for automated GMA in infants at risk of CP.

Unique 4-DOF Relative Pose Estimation with Six Distances for UWB/V-SLAM-Based Devices.

In this work we introduce a relative localization method that estimates the coordinate frame transformation between two devices based on distance measurements. We present a linear algorithm that calculates the relative pose in 2D or 3D with four degrees of freedom (4-DOF). This algorithm needs a minimum of five or six distance measurements, respectively, to estimate the relative pose uniquely. We use the linear algorithm in conjunction with outlier detection algorithms and as a good initial estimate for iterative least squares refinement. The proposed method outperforms other related linear methods in terms of distance measurements needed and in terms of accuracy. In comparison with a related linear algorithm in 2D, we can reduce 10% of the translation error. In contrast to the more general 6-DOF linear algorithm, our 4-DOF method reduces the minimum distances needed from ten to six and the rotation error by a factor of four at the standard deviation of our ultra-wideband (UWB) transponders. When using the same amount of measurements the orientation error and translation error are approximately reduced to a factor of ten. We validate our method with simulations and an experimental setup, where we integrate ultra-wideband (UWB) technology into simultaneous localization and mapping (SLAM)-based devices. The presented relative pose estimation method is intended for use in augmented reality applications for cooperative localization with head-mounted displays. We foresee practical use cases of this method in cooperative SLAM, where map merging is performed in the most proactive manner.

Body pose estimation in depth images for infant motion analysis.

Motion analysis of infants is used for early detection of movement disorders like cerebral palsy. For the development of automated methods, capturing the infant's pose accurately is crucial. Our system for predicting 3D joint positions is based on a recently introduced pixelwise body part classifier using random ferns, to which we propose multiple enhancements. We apply a feature selection step before training random ferns to avoid the inclusion of redundant features. We introduce a kinematic chain reweighting scheme to identify and to correct misclassified pixels, and we achieve rotation invariance by performing PCA on the input depth image. The proposed methods improve pose estimation accuracy by a large margin on multiple recordings of infants. We demonstrate the suitability of the approach for motion analysis by comparing predicted knee angles to ground truth angles.

Local region descriptors for active contours evolution.

Edge-based and region-based active contours are frequently used in image segmentation. While edges characterize small neighborhoods of pixels, region descriptors characterize entire image regions that may have overlapping probability densities. In this paper, we propose to characterize image regions locally by defining Local Region Descriptors (LRDs). These are essentially feature statistics from pixels located within windows centered on the evolving contour, and they may reduce the overlap between distributions. LRDs are used to define general-form energies based on level sets. In general, a particular energy is associated with an active contour by means of the logarithm of the probability density of features conditioned on the region. In order to reduce the number of local minima of such energies, we introduce two novel functions for constructing the energy functional which are both based on the assumption that local densities are approximately Gaussian. The first uses a similarity measure between features of pixels that involves confidence intervals. The second employs a local Markov Random Field (MRF) model. By minimizing the associated energies, we obtain active contours that can segment objects that have largely overlapping global probability densities. Our experiments show that the proposed method can accurately segment natural large images in very short time when using a fast level-set implementation.