Development of a novel artificial intelligence algorithm to detect pulmonary nodules on chest radiography

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Lung cancer diagnosis relies on accurate interpretation of chest radiographs.
• Pulmonary nodule detection is crucial for early lung cancer diagnosis.
• Physician interpretation of chest X-rays can be time-consuming and prone to error.

Purpose of the Study:

• To develop a novel artificial intelligence (AI) algorithm for pulmonary nodule detection.
• To enhance the efficiency and accuracy of lung cancer diagnosis using AI-assisted interpretation of chest radiographs.

Main Methods:

• Chest X-ray images from Fukushima and the National Institutes of Health (NIH) Chest X-ray 14 dataset were analyzed.
• Two datasets (Type A: combined, Type B: Fukushima only) were used for algorithm training and validation.
• AI algorithms generated heatmap displays and positive probability scores for pulmonary nodules.

Main Results:

• The AI algorithm achieved a receiver operating characteristic (ROC) area under the curve (AUC) of 0.74 (sensitivity 0.75, specificity 0.60) for the Type A dataset.
• For the Type B dataset, the AI achieved an AUC of 0.79 (sensitivity 0.72, specificity 0.74).
• The AI algorithms demonstrated accuracy superior to radiologists and comparable to previous studies.

Conclusions:

• The developed AI algorithms show comparable accuracy to radiologists in interpreting chest radiographs.
• High-quality AI algorithms can be trained even with smaller datasets (Type B).
• Further research is necessary to enhance and validate the AI algorithm's accuracy for clinical application.