Scholarlink

<p>The ST-segment elevation in ECG is a signal that is essential in the case of myocardial infarction and has to be recognized correctly in time. This paper involved training and testing of two lightweight convolutional object detection models (YOLOv10s and YOLOv11s) on a dataset of ECG images. A comparative study indicated that YOLOv11s performed better and was considered for further refinement. The improvement has been done on feature representation of YOLOv11s by adding a BiFPN-based multi-scale fusion neck and other 1x1 and 3x3 convolutional layers to enhance the capability of this model to capture and align the subtle ST elevation patterns in ECG images. Our model achieves 95.1% precision, 97.8% recall, and 98.2% mAP@50, outperforming baseline YOLOv10 and YOLOv11s. Lastly, the trained model was deployed as a user-friendly Flask web application to detect the ST-elevation in real time and assist clinicians in making faster and more reliable decisions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A modified YOLO-based deep learning model for accurate ST-elevation myocardial infarction detection in ECG images

  • 期刊论文

  • 发表日期：

    2026年8月3日

查看全文
  • Roaa Albasrawi,
  • Muhammad Ilyas,
  • Oğuz Bayat

摘要

The ST-segment elevation in ECG is a signal that is essential in the case of myocardial infarction and has to be recognized correctly in time. This paper involved training and testing of two lightweight convolutional object detection models (YOLOv10s and YOLOv11s) on a dataset of ECG images. A comparative study indicated that YOLOv11s performed better and was considered for further refinement. The improvement has been done on feature representation of YOLOv11s by adding a BiFPN-based multi-scale fusion neck and other 1x1 and 3x3 convolutional layers to enhance the capability of this model to capture and align the subtle ST elevation patterns in ECG images. Our model achieves 95.1% precision, 97.8% recall, and 98.2% mAP@50, outperforming baseline YOLOv10 and YOLOv11s. Lastly, the trained model was deployed as a user-friendly Flask web application to detect the ST-elevation in real time and assist clinicians in making faster and more reliable decisions.