SEGMENTATION OF INFARCT AND SCAR ON HEART LEFT VENTRICLE BASED ON LGE-CMR IMAGE

• Main Author: Abdurrafi, Alwan
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/51221
• Institution: Institut Teknologi Bandung

Heart failure is the leading cause of death in the world. Amongst all the type, acute myocardial infarction is the deadliest. Acute myocardial infarction caused by the irreversible death of heart muscle cells in the left ventricle called infarct. Infarcted cells create fibrotic tissue around them, called scar. One of the methods to evaluate the position of infarct and scar tissue is to acquire Late Gadolinium Enhancement – Cardio Magnetic Resonance (LGE-CMR) image. To have the correct medical treatment, the image must be first evaluated by observers. Evaluation proses takes a lot of time and the result is prone to inter-variability problem. Based on this problem, automatic algorithm to segment infarct and scar is developed. In this final project, automatic infarct segmentation algorithm was designed using Otsu threshold twice. The designed algorithm was evaluated using 88 LGE-CMR datas with high variance. Metrics used to evaluate the algorithm were Dice Similarity Coefficeint (DSC), detection rate, and bull’seye plot. Evaluation result gives DSC mean score 62,13±23,21% and detection rate score was 86,36%. In this two metrics, the designed algorithm gives the best performance among two other reference algorithms. However, the designed algorithm gives the lowest performance in bull’seye plot evaluation (70,75%). This indicates oversegmentation occurred in the final segmentation.