Comparative Study between Horn-Schunk and Lucas Kenade Optical Flow Method for Lung Lesion Identification in Computed Tomography Images

• Published in: Proceedings - 13th IEEE International Conference on Control System, Computing and Engineering, ICCSCE 2023
• Main Author: Abdullah M.F.; Sulaiman S.N.B.; Osman M.K.; Karim N.K.A.; Ani A.I.C.; Damit D.S.A.
• Format: Conference paper
• Language: English
• Published: Institute of Electrical and Electronics Engineers Inc. 2023
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85172864792&doi=10.1109%2fICCSCE58721.2023.10237099&partnerID=40&md5=e99936996a4d91ca88cabec388ead629

This study compares the Horn-schunk and Lucas Kenade optical flow methods for lung lesion identification in computed tomography (CT) images. Mostly, the detected lesions were based only on a single CT image. However, detecting lesions from a single image does not align with the Radiologist's practice based on sequential images. Therefore, there is a need to consider the before-and-after images for identifying the moving object in the lung corresponding to the lesions' detection. Hence, this project aims to identify the lung lesion using two optical flow methods. The method that analyses pair of images called the optical flow method was chosen since the object stores the direction and speed of a moving object from one image, or video frame to another. The optical flow-based approach uses a pair of frames as data to measure optical flow based on four different mode selections. A new method has been proposed to find the possible images of lung lesions based on the standard deviation from the threshold value. As the image pair selected, the Horn-Schunck and Lukas Kenade optical flow method was picked in this study. Based on two optical flow approaches, Horn-Schunk was the most efficient way of locating lesions using mode 0 with a standard deviation of 0.99%. The finding was revealed that the threshold value based on the length of the optical flow vector gives a significant output for predicting the lesions and non-lesion CT images range. © 2023 IEEE.