Dr. Olivier Rukundo
Medical University of Vienna, Austria
Speech Title: Revolutionary Processor for Deep Learning Discoveries and Applications
Abstract: I will give an overview of the importance of deep learning in bio/medical image analysis, and the current challenges or limitations in this field. I will talk about two of our recent works on traditional GPU cluster-enabled bio/medical image analysis with convolutional neural networks. The first is about effects of LGE MRI image size on the performance of GPU-enabled deep learning algorithms in quantifying cardiac functions . We used that opportunity to introduce a novel approach for interpolation mask handling which has significant implications for accuracy and efficiency of GPU-enabled deep learning models. The second focuses on testing the automation of mini- Transmission Electron Microscopy (TEM) image analysis systems with GPU-enabled convolutional neural networks. We resort to patching techniques due to the huge size of TEM images used as input to our deep learning architecture of interest. We proposed a semi-automatic annotation tool and later software tool for automatic detection of intact adenovirus in Mini-TEM Images . In the end, I will talk about the solutions brought by the Cerebras System Wafer-Scale Engine and the potential to revolutionize deep learning research and applications.
Short bio: Olivier Rukundo is a
Researcher at the University Clinic of
Dentistry, Medical University of Vienna
. He received his Master and Ph.D.
degrees in Communication and Information
Systems from the Department of
Electronics and Information Engineering
at Huazhong University of Science and
Technology, Wuhan, China. His research
interests include artificial
intelligence for image analysis and
visual computing. He received a research
grant from the Research Council of
Norway for Upscaling Based Image
Enhancement for Video Capsule Endoscopy
(project No. 260175). He has authored
and reviewed many research papers in top
scientific journals and international
conferences . He is currently developing
a deep learning-based decision support
system for classification of oral
dysplasia grades.
.