Video-Based Neurological Deficit Analysis
Zhuang, Yan, Computer Engineering - School of Engineering and Applied Science, University of Virginia
Rohde, Gustavo, EN-Biomed Engr Dept, University of Virginia
The configuration and movement of the face can indicate the presence or absence of various neurological diseases. A central brain lesion such as a stroke will cause pathological, asymmetric weakness on the lower facial muscles on the contralateral side. However, recognizing facial weakness in existing pre-hospital settings still remains a challenge, largely due to variability in training and experience of non-neurologist providers. The proposed research develops an automated, accurate, and quantitative video-based digital screening tool for facial weakness analysis that can enable fast patient triage and augment standard pre-hospital stroke care. The proposed approach not only achieves equivalent performance to paramedics, but also provides visualizable and interpretable results. In addition, to increase the model’s robustness to illumination changes, we leverage patch-wise (local) image gradient distributions and transport-based metric for illumination-invariant face analysis. The experiment results demonstrate that the proposed method outperforms other alternatives in several face analysis tasks with challenging illumination conditions.
PHD (Doctor of Philosophy)
Neurological deficit, facial weakness, computer vision, image analysis, video classification, Illumination-invariant face recognition, sliced-Wasserstein
All rights reserved (no additional license for public reuse)