- Featured Events
- Today's Events
- Browse by Type
- Browse by Subject
- Browse by Organization
Optics and Electronics Seminar
Apr 10, 2017
4:15 PM, Spilker 232
http://campus-map.stanford.edu/index.cfm?ID=04-040 - Map
BRINGING MEDICAL IMAGING INTO THE ERA OF BIG DATA
IBM Almaden Research Center
San Jose, CA
The era of big data promises new solutions in many areas. However, it also highlights
the challenges of curating large datasets for training machine learning algorithms.
In medical imaging two major issues have slowed the pace of progress. First, data
is usually not labeled with ground truth for use in training. Second, data from different
practices is often inconsistent in terms of modalities.
In this talk, we present solutions to these problems through novel algorithms in deep learning and in random forest domains. The talk will primarily report technical advances from two recent publications of the speaker, namely “Learning in Data-Limited Multimodal Scenarios: Scandent Decision Forests and Tree-Based Features” and “A Cross-Modality Neural Network Transform for Semi-Automatic Medical Image Annotation.”
Dr. Mehdi Moradi is a research manager at IBM Almaden Research Center where he leads a team of scientists working on developing machine learning solutions in medical image analysis. He is also an adjunct professor at the Dept. of EECE at the Univ. of British Columbia, where he was an assistant professor before joining IBM in 2014. Dr Moradi completed his PhD in computer science at Queen's University at Kingston, Canada and a research fellowship at Harvard Medical School. He has developed award winning technologies in automatic ultrasound and MRI-based disease detection. He is a senior member of IEEE, an associate editor of Medical Physics and of IEEE Signal Processing Letters, a program committee member of Medical Image Computing and Computer Assisted Interventions (MICCAI), and of SPIE Medical Imaging. He has co-authored over 50 peer-reviewed publications in medical imaging and machine learning. His most recent work addresses difficulties of producing large scale medical imaging datasets and has resulted in algorithms for combining inconsistent datasets, and automatic ground truth generation.