Artificial Intelligence Could Predict Chest Tube Drainage Necessity for Spontaneous Pneumothorax

Area of Science:

• Medical Imaging and Artificial Intelligence
• Thoracic Surgery and Critical Care

Background:

• Artificial intelligence (AI) is primarily used for medical diagnostics, with demonstrated success in detecting pneumothorax on chest X-rays (CXR).
• However, AI has not yet been applied to guide treatment decisions for pneumothorax, such as the necessity of chest tube drainage (CTD).

Purpose of the Study:

• To develop and evaluate an AI-based clinical decision support system for predicting the need for CTD in spontaneous pneumothorax patients.
• The system integrates CXR analysis and patient clinical data to enhance treatment decision-making accuracy and efficiency.

Main Methods:

• A two-stage AI model was created: deep learning for pneumothorax segmentation/quantification from CXR, and machine learning for predicting CTD necessity.
• The models were trained on data from 163 pneumothorax patients, incorporating pneumothorax size and clinical parameters.
• Performance was evaluated using AUROC, sensitivity, specificity, and inference time.

Main Results:

• The AI model achieved high pneumothorax segmentation accuracy (DSC: 76.95%, MAE: 5.06%).
• The integrated AI model demonstrated a superior AUROC of 89.68% for predicting CTD necessity, outperforming models using only pneumothorax ratio or clinical data alone.
• The optimized model achieved 80.95% sensitivity and 100% specificity with a mean inference time of 0.75 seconds.

Conclusions:

• An AI-based clinical decision support system was successfully developed to predict CTD requirements in spontaneous pneumothorax.
• Integrating AI-driven pneumothorax quantification with clinical data significantly improves the accuracy and efficiency of treatment decision-making.
• Further validation with larger datasets and prospective studies are recommended to refine and confirm the model's clinical utility.