Semisupervised AI approach for automated categorization of medical images
Who is this presentation for?
- Healthcare CIOs and CDOs, informaticists, data scientists, and radiologists
In recent years the healthcare industry has adopted AI methodologies. Radiology, in particular, has seen successful applications in image classification and segmentation with supervised ML approaches. The potential of automatic neural network processing of images, for example, automated screening for pathological image markers has far-reaching implications for diagnostic radiology. However, feature variation in medical images and the large number of labeled images required for successful neural network training present challenges for supervised learning.
Stephan Erberich, Kalvin Ogbuefi, and Long Ho share a semisupervised learning approach to address automated image classification toward labeling all radiological images that the department acquired. The approach is composed of cascading self-organizing maps (SOM) and convolutional neural networks (CNNs) to categorize radiological images based on image content. This provides an opportunity to create an automated AI-based system for labeling any radiographic data and queuing the categorized images downstream to specialized ML algorithms for pathology-model inference. You’ll discover the approach, methodologies, algorithm orchestration, and performance.
- A basic understanding of machine learning pipelines and deep learning algorithms
What you'll learn
- Learn about the unsupervised method for labeling medical images
- Discover actable AI implementation for production
- Understand the uses of AI in radiology and healthcare
University of Southern California
Stephan Erberich is a chief data officer at the Children’s Hospital Los Angeles and a professor of research radiology at the University of Southern California. He’s a computer scientist, specialized in medical informatics. He’s an AI practitioner in healthcare with a focus on radiological image processing and computer vision ML.
Children's Hospital Los Angeles
Kalvin Ogbuefi is a data scientist at the Children’s Hospital Los Angeles (CHLA). Previously, he was a project assistant in the USC Stevens Neuroimaging and Informatics Institute, Marina del Rey, on radiology image analysis. His extensive research experience comprises projects in deep learning, statistical modeling, and computer simulations at Lawrence Livermore National Laboratories and other major research institutions. He earned an MS in applied statistics from California State University, Long Beach and a BS in applied mathematics from University of California, Merced.
Children's Hospital Los Angeles
Long Van Ho is a data scientist at Children’s Hospital Los Angeles with over five years of experience in applying advanced machine learning techniques in healthcare and defense applications. His work includes developing the machine learning framework to enable data science at the Virtual Pediatric Intensive Care Unit and researching applications of artificial intelligence to improve care in the ICUs. His background includes research in particle beam physics at UCLA, and Stanford University has provided a strong research background in his career as a data scientist. His interests and goal is to bridge the potential of machine learning with practical applications to health.
Leave a Comment or Question
Help us make this conference the best it can be for you. Have questions you'd like this speaker to address? Suggestions for issues that deserve extra attention? Feedback that you'd like to share with the speaker and other attendees?
Join the conversation here (requires login)
Premier Diamond Sponsors
Premier Exhibitor Plus
For conference registration information and customer service
For more information on community discounts and trade opportunities with O’Reilly conferences
For information on exhibiting or sponsoring a conference
For media/analyst press inquires