Automatic detection of supporting device positioning in intensive care unit radiography

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Radiography

Background:

• Intensive care unit (ICU) patients often require endotracheal, feeding, or nasogastric tubes.
• Accurate positioning of these tubes is critical for patient safety and effective treatment.
• Current methods for verifying tube placement can be time-consuming and prone to error.

Purpose of the Study:

• To develop and evaluate a computer-aided method for automatic detection and identification of endotracheal, feeding, and nasogastric tubes in ICU radiography.
• To enhance the ease and accuracy of tube tip detection for clinicians.
• To potentially improve the quality of patient care in the ICU setting.

Main Methods:

• A dataset of 107 portable X-ray images from 20 patients was utilized.
• A Kodak computed radiography system was employed for image acquisition.
• An experienced chest radiologist provided ground truth for tube presence, type, and position.
• The proposed automatic detection method's performance was evaluated against the radiologist's assessment.

Main Results:

• The computer-aided technique achieved a 94% detection rate for endotracheal tubes with 0.6 false positives per image.
• Detection rates for feeding and nasogastric tubes were 82%, with 0.4 and 0.5 false positives per image, respectively.
• Preliminary findings indicate high sensitivity and an acceptable false positive rate.

Conclusions:

• The developed technique demonstrates accurate detection of tubes in ICU chest radiographs.
• The method achieves a high sensitivity level for tube identification.
• The computer-aided approach offers an acceptable false positive rate, supporting its clinical utility.