Classification of Central Venous Catheter Tip Position on Chest X-ray Using Artificial Intelligence

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Radiology

Background:

• Central venous catheters (CVCs) are crucial for patient care but require accurate placement verification.
• Current methods for assessing CVC tip position on chest radiographs lack definitive automated criteria.
• Deep convolutional neural networks (CNNs) have shown promise in analyzing medical images.

Purpose of the Study:

• To develop and validate a deep CNN algorithm for automatic classification of CVC tip position on chest radiographs.
• To enable precise categorization of CVC placement into shallow, proper, and deep positions.
• To automate the segmentation of the trachea and CVC for accurate positional analysis.

Main Methods:

• A retrospective study using 808 supine anteroposterior chest radiographs.
• A two-stage deep CNN approach: U-net++ for segmentation and EfficientNet B4 for classification.
• Classification based on the vertical distance between the tracheal carina and CVC tip.

Main Results:

• The algorithm achieved an overall accuracy of 0.76 (±0.03) in five-fold cross-validation with segmented images.
• Testing on segmentation-free images yielded an average accuracy of 0.82, precision of 0.73, recall of 0.73, and F1-score of 0.73.
• Highest accuracy (0.91) was observed for shallow CVC positions, and the highest F1-score (0.82) for deep positions.

Conclusions:

• Deep CNNs can effectively perform automatic segmentation and classification of CVC tip position on plain chest radiographs.
• The developed algorithm demonstrates comparable performance to manual assessment, offering potential for improved clinical workflow.
• This AI-driven approach aids in accurate CVC placement verification, crucial for patient safety and treatment efficacy.