Chest Region Estimation using Chin Landmark Keypoints for Heart Attack Detection

• Published in: IEEE Control System Graduate Research Colloquium (Online) pp. 204 - 209
• Main Authors: Ibrahim, Noraizan, Karim, Rohana Abdul, Arshad, Nurul Wahidah, Samsudin, Wan Nur Azhani Binti W., Yakno, Marlina
• Format: Conference Proceeding
• Language: English
• Published: IEEE 02.08.2025
• Subjects: Accuracy; Cardiac arrest; chest localization; chest pain; Corner detection; Face recognition; Face reference point; harris; heart attack; key point; local minima; Location awareness; Pain; Sensitivity; Thresholding (Imaging); Visualization; X-rays
• ISSN: 2833-1028
• DOI: 10.1109/ICSGRC65918.2025.11159829

This paper proposes a lightweight and noninvasive approach for localizing the external chest area by utilizing facial landmarks, aiming to support early detection of chest pain related to heart attacks. While most existing research focuses on internal chest imaging such as X-rays and MRI, external chest localization remains largely unexplored. The challenge arises because areas around the chest, including the hands, face, and neck, often move slightly or change position, which can make it difficult for the system to consistently and accurately identify the chest location. To address these challenges, our approach focuses on accurately estimating the chin as a stable reference point using two techniques: Local Minima Detection based on grayscale intensity changes and Harris Corner Detection, known for its robustness in identifying geometric features. We evaluated both methods on six test samples, each consisting of 200 images captured under controlled conditions. The results show that Harris Corner Detection achieves a higher peak accuracy of 85 \%, outperforming the Local Minima method at 80 \%. This improved performance is mainly due to Harris Corner Detection's ability to reliably detect the chin-neck junction even in the presence of visual noise and variations in subject posture.