RGB-D Videos-Based Early Prediction of Infant Cerebral Palsy via General Movements Complexity

Abstract

Early detection and intervention of cerebral palsy can promote neural remodeling in the process of brain development, thus reducing the negative effects of cerebral palsy. In this paper, we proposed a novel method for early prediction of infant cerebral palsy based on General Movements Assessment (GMA) theory with RGB-D videos. Firstly, we explored the human pose recognition in supine position based on RGB-D videos. Then we further apply it to auto-GMA. Specifically, we employ current pose estimation method on RGB images to achieve the infant full body 2D key points. By combining the depth information, the 3D movement of infant in supine position can be obtained. Then the infant's movement complexity index is achieved by extracting the infant's whole-body movement characteristic. In order to verify the effectiveness of the method, we did some experiments on a public dataset consisting 12 real recorded infants' movement RGB-D videos, with 4 of the samples were diagnosed as abnormal infants by a GMA expert. We use expert GMA ratings of these recorded movements as the gold standard. Our method achieved state-of-the-art with sensitivity of 100%, specificity of 87.5%, and accuracy of 91.7%. The results show that the method has great potential in assisting doctors in diagnose infant cerebral palsy.

Towards Explainable Abnormal Infant Movements Identification: A Body-part Based Prediction and Visualisation Framework

“…Next, we present the visualization results as qualitative analysis in Section III-B. We follow the standard protocol as in [11], [12], [18] to conduct a leave-one-subject out crossvalidation to ensure the results presented in this section are obtained base on unseen data during the training process.…”

A Pose-Based Feature Fusion and Classification Framework for the Early Prediction of Cerebral Palsy in Infants

“…However, each typically assesses the effectiveness of the extracted pose rather than the effect this might have upon final classification. In contrast, [33], [34], [38], [51], [56], examine the viability of using pose-based features for the prediction of CP using a feature extraction and classification pipeline.…”

“…In order to assess the effectiveness and robustness of our system, we reimplement several video-based methods from the literature to serve as baselines for comparison, including Centroid of Motion and Quantity of Motion [1], [2]; Absolute Motion Distance, Relative Frequency, and Magnitude of Wavelet Coefficients [53]; and Frequency Analysis [45]. We also compare our results with those reported in [56] and [57], as well as conducting our experiments using the source code provided by the authors of [38] and [51].…”

“…To evaluate the generality of the classifiers using both our proposed features and the baseline approaches, a leave-one-subject-out cross validation method is used. Splitting the data in this manner ensures that the classifiers will be evaluated using unseen data, as suggested in previous works [33], [34], [56]. We further evaluate the performance of each type of feature on different classifiers (as explained in Section III-F).…”

Identification of Abnormal Movements in Infants: A Deep Neural Network for Body Part-Based Prediction of Cerebral Palsy

“…0.6667 0.5000 0.7500 CX SD [4] kNN (k=3) 0.6667 0.7500 0.6250 CYmean [4] SVM 0.6667 0.0000 1.0000 CY SD [4] kNN (k=3) 0.7500 0.7500 0.7500 C SD [1] Ensemble 0.7500 0.5000 0.8750 Qmean [1] kNN (k=1) 0.5000 0.2500 0.6250 Q SD [1] kNN (k=3) 0.7500 0.7500 0.7500 CPP [1] Ensemble 0.6667 0.5000 0.7500 AMD [40] 32 Ensemble 1.0000 1.0000 1.0000 Freq. [40] 8 LDA 1.0000 1.0000 1.0000 Wavelet [40] 128 LDA 0.8333 0.7500 0.8750 Wu et al [43] 0.9167 1.0000 0.8750 HOJO2D [30] 8 LDA 1.0000 1.0000 1.0000 HOJO2D [30] 16 LDA 0.7500 0.7500 0.7500 HOJD2D […”