Summary
International Workshop on Smart Info-Media Systems in Asia
2023
Session Number:RS3
Session:
Number:RS3-7
Thrombus Visualization using Pattern Recognition and Classification Model
Atiwich Chantharamalee, Akdanai Thumchaidecha, Dittapong Songsaeng, Thanaruk Theeramunkong, Natsuda Kaothanthong,
pp.97-101
Publication Date:2023/8/31
Online ISSN:2188-5079
DOI:10.34385/proc.77.RS3-7
PDF download
Summary:
A method for localizing thrombus in a set of noncontrast computer tomography (ncCT) in the middle cerebral artery (MCA) in the brain was proposed. Due to a relatively small thrombus region in an image, many false positive regions were obtained from the classification model. To reduce the false positive, Circle of Willis were found and employed as salient region where thrombus is commonly found in MCA. Then, Hounsfield Units (HU) value of each pixel was clustered to find thrombus candidates. A machine learning model was applied to classified each candidate in order to visualize the thrombus. The proposed method achieved 7.14% precision, 76.22% accuracy, with 85.89% sensitivity and 76.30% specificity combining the probability returned from a support vector machine (SVM) model and an object detection model.