Biography: Dr Hu has been working in the research area of neural engineering and clinical electrophysiology for more than 30 years. He has conducted more than 30 research projects with granted from Hong Kong RGC, S.K. Yee Medical Foundation, NSFC etc. He published 10 book chapters, and more than 150 journal paper with SCI citation, with published more than 50 in journals of top rank JCR.
He serves as vice president of basic science sub-committee, Spine and spinal cord committee of Chinese Rehabilitation Medicine Society, Council committee member in Chinese Sub-society of Biomedical Sensor Technology and Chinese Sub-society of Medical Neural Engineering. He is also a senior member of IEEE and Chinese BME society.
Dr. Hu was awarded several international prizes for his academic contributions. His work has been selected as the top 5 best papers in biomedical signal processing of the The Yearbook of Medical Informatics 2003. In 2006 and 2007, he was awarded Macnab/Larocca Research Fellowship twice from the International Society for the Study of the Lumbar Spine (ISSLS). In 2013, he was awarded Universitas 21 Fellowship. His papers were awarded the best paper for 2008 in Anaesth Intensive Care, 2011 awarded Oldendorf Award from American Society of Neuroimaging. And Cum Laude Best Poster Award in SPIE Medical Imaging conference 2011.
Speech Title: A Machine Learning Approach to Explore Cervical Myelopathic Location from Somatosensory Evoked Potentials
Abstract: Somatosensory evoked potentials (SEPs) has been demonstrated to be able to provide useful information of the spinal cord injury (SCI) location by using time-frequency analysis. A better understanding of the mechanisms involved in SEP time-frequency components is essential for the development of more reliable algorithms for level diagnosis of cervical myelopathy. In this study, we establish a compressive SCI rat model at various locations (C4, C5, and C6) to simulate cervical myelopathy. Waveforms of SEPs after compressive injuries were collected and decomposed into time-frequency components(TFCs). A multi-stage Support Vector Machines (SVM) is employed to analyze the dataset from 3 groups of SCI rats. The performance of this classifier was evaluated by 10-fold cross-validation. Results showed that the averaged classification accuracy of the constrained 10-fold cross-validation was 81.3±1.2% ranging from 79.6% to 83.6% (in the ten iterations). Features of SEP TFCs contain the location message of cervical spinal cord injury, which demonstrated the potential use of SEP TFCs in level diagnosis of cervical myelopathy.