Meeting Banner
Abstract #3395

Improving Lesion Classification Using an Empirical Knowledge of False Classifications in Multiple Sclerosis

Sushmita Datta1, Xiaojun Sun1, Ponnada A. Narayana1

1Department of Diagnostic and Interventional Imaging, The University of Texas Health Science Center at Houston, Houston, TX, United States

Automated classification of lesions in multiple sclerosis (MS) is often hindered by the presence of false classifications (FCs). These FCs occur due to presence of some regions mimicking lesions. We have developed and implemented a false classification probability (FCP) map for improved lesion classification using the knowledge of false classifications obtained from automated segmented and validated lesion classifications. The application of FCP map significantly improved the lesion classification in 57 MS subjects as assessed by the Dice similarity indices.