Live 1057

Free Internet Radio Stations
UCSF Radiology Valentina Pedoia – Computer Vision and Machine Learning

UCSF Radiology Valentina Pedoia – Computer Vision and Machine Learning

Hi. My name is Valentina Pedoia. I’m an assistant professor in the Musculoskeletal Quantitative
Imaging Research Group. I am a Ph.D. computer scientist and I graduated
five years ago in Italy and I came to UCSF right after graduation and I’m here since
then. My focus is on applying Computer Vision and
Machine Learning techniques to mostly magnetic resonance images to study musculoskeletal
disorders and osteoporosis. Machine learning finds a set of techniques
in which you teach a machine to solve a problem as a human would do. And that in Medical Imaging we can transforming
an image into information. Those information that can be used by Radiologies
and Clinician for better characterization of the disease and prediction of file cams. Let’s say you are a soccer player. You come in with an ACL Injury at your anterior
cruciate ligament and the part when your MRI exam, we don’t know how extensive is your
injury, if you have a meniscus tear or how the other knee structure looks like. And the Machine Learning Algorithm could analyze
your image while you are on the table and tell us where we should look like and how
we should look at your knee. And that will shape your MRI protocol around
your needs. That is what we call precision medicine. And on top of that, we can take information
that we got in the scanner, put it together with your clinical information, with your
demographic and maybe predict if your are one of those 50 precent of subject that after
an ACL injury will have post-traumatic osteoarthritis. And here we are speaking about the debilitating
knee pain when you are in your forties. And if we can predict the onset of the disease,
that means early intervention, and early intervention can mean prevention or delay the onset of
the disease. We analyze big data sets of 3D volumetric
MRIs, so we need very high computational power. We need machine that handles those kind of
data set. And we also work on images and data that are
sensitive because they are patient data and we need infrastructure that is secure and
database that is secure for transferring an immunization pipeline. Collaboration are very important. I am a technician. I am a computer scientist working in a medical
school. Collaboration with them, this, are very important. The success of machine learning algorithm
is mostly related to domain knowledge. What the problem is complex, posting the hard
question is even more difficult than finding the answer, and clinician help with that. Technology in general and Machine Learning
in specific; they are going to change how we know the radiology. Everything about what we know about how radiology
is performed is going to change in the future. And that is just not for Image Interpretation,
Machine Learning will shape the entire data pipeline; from ordering an image exam, to
process the image, to clinical decision. The acquisition will be faster, and cheaper,
and the interpretation will be a very much deductive science and it’s going to be great.

Leave a Reply

Your email address will not be published. Required fields are marked *