Enhancing Pulmonary Embolism Detection in COVID-19 Patients Through Advanced Deep Learning Techniques.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pulmonology

Background:

• COVID-19 has increased the incidence of pulmonary embolism (PE), creating diagnostic challenges.
• Accurate and timely detection of PE is critical for patient outcomes.
• Existing diagnostic methods require enhancement, especially in post-pandemic scenarios.

Purpose of the Study:

• To evaluate the effectiveness of deep learning for enhanced PE detection in post-COVID-19 patients.
• To assess the performance of the Ultralytics YOLOv8 model on Computed Tomography Pulmonary Angiography (CTPA) scans.
• To explore AI's potential in improving diagnostic accuracy for PE.

Main Methods:

• A dataset of 746 anonymized CTPA images from 25 patients was utilized.
• The Ultralytics YOLOv8 object detection model was fine-tuned and trained on 676 images.
• The model was validated on 70 images, with performance metrics including mean Average Precision (mAP) and recall.

Main Results:

• The YOLOv8 model achieved a mean Average Precision (mAP) of 0.683 at IoU 0.50 and 0.246 at IoU 0.50:0.95.
• The model demonstrated high diagnostic proficiency with a maximum precision of 0.859 and recall of 0.815.
• Training was completed in approximately 1.021 hours over 200 epochs.

Conclusions:

• Deep learning models like YOLOv8 show significant potential for accurate PE detection in post-COVID-19 patients.
• AI-powered tools can enhance diagnostic capabilities in medical imaging.
• This research contributes to healthcare innovation and improved patient care through advanced technology.