REU WEEK 8 Nancy Zanaty, UCF
Past approach modified Previously: Classifying individual images in a timeseries as “ADHD” or “Non ADHD” as a test of the SVM Able to obtain around 65-75% for individual images
New approach: Majority Classify as “ADHD” or “Non ADHD” based on what the majority of the time series is classified as
SVM Results From past data, 2/3 brains were classified correctly. The brains that were classified incorrectly by only a small margin. Meaning that around 45% of the timeseries was correct but not enough to be majority
Problems Tried to expand data to classify more brains, but was too large for SVM Had to cut down data but sample wasn’t wasn’t broad enough to be accurate Only about 50-60% were classified accurately
Next steps Resizing images or looking to different data sets Train SVM on larger samples and begin classifying more brains