Loading…
This event has ended. Visit the official site or create your own event on Sched.
THE BIG FESTIVAL ABOUT SMALL CITIES
Tom Tom champions civic innovation, creativity, and entrepreneurship in America’s hometowns.

[Back to Tom Tom Festival]
Back To Schedule
Thursday, April 11 • 2:55pm - 3:25pm
MRI Image Synthesis for the Diagnosis of Parkinson's Disease

Log in to save this to your schedule, view media, leave feedback and see who's attending!

Feedback form is now closed.
Parkinsons disease (PD) is a neurodegenerative disease that affects an estimated 1% of adults over 65. While the disease itself is not fatal, complications related to PD are rated as the 14th largest cause of death in the United States by the Center of Disease Control and Prevention. In spite of technological advances, Parkinson’s diagnosis methods have not changed, and the accuracy of diagnosis has remained at approximately 81% for the past 25 years. These methods include analyzing years of neurological data to determine if the patient has developed the symptoms of Parkinson’s, including limb rigidity and tremors, both of which are common side effects of a number of other diseases. With the rise of automated prediction algorithms paired with the generation of massive amounts of data, the automatic diagnosis of Parkinson’s Disease has not caught up to traditional means. This is commonly attributed to the lack of useful data, as most computational systems require a tremendous amount of medical data that isn’t readily available as gathering the data can be expensive. This study presents PDGAN, a tool to aid pathologists and neurologists in the diagnosis of Parkinson’s Disease. PDGAN uses a series of neural networks to classify Magnetic Resonance Images (MRI Images). PDGAN employs Generative Adversarial Networks (GANs) to synthetically generate medical images which is used to augment the classification efforts. The pair of Convolutional Neural Networks exhibited an testing accuracy of 91.4% without the augment of new data, and combined the total accuracy was 96.6%, a 16% increase com- pared to traditional methods at a fraction of the cost and time. PDGAN demonstrates the feasibility of utilizing GANs to generate unseen data for the improvement of classification accuracy in the medical setting. The biggest challenge in leveraging the power of AI/ML in the healthcare space today is a lack of data access. What if we could artificially generate more data to fill in the gaps where traditional patient-driven sampling techniques don’t meet the data demand? This session presents one use-case for Generative Adversarial Networks and Auto-encoders to augment a data-set, as well as other optimization techniques for classifiers working with small amounts of data. This session will demonstrate these concepts in the scope of an important problem in the field of healthcare: accurately assessing Parkinson’s onset for a quick diagnosis. 

You need this ticket from Eventbrite to sign up: Applied Machine Learning Conference.

Speakers
avatar for Neeyanth Kopparapu

Neeyanth Kopparapu

Student, Thomas Jefferson High School for Science and Technology
Neeyanth Kopparapu is a High School Junior at Thomas Jefferson High School in Alexandria, VA. As a student, he is interested in mathematics, computer science, and physics. He is especially interested in developing computer science (specifically artificial intelligence) models to help... Read More →

Sponsors

Thursday April 11, 2019 2:55pm - 3:25pm EDT
Violet Crown, Theater 1 200 W Main St, Charlottesville, VA 22902, USA