An AI deep learning algorithm for detecting pulmonary nodules on ultra-low-dose CT in an emergency setting: a reader...

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Pulmonary Medicine

Background:

• Ultra-low-dose computed tomography (ULDCT) is increasingly used in emergency departments (ED) for suspected pulmonary disease.
• The OPTIMACT trial investigated the utility of AI for detecting pulmonary nodules on ULDCT.
• Accurate detection of incidental pulmonary nodules is crucial for early diagnosis and management.

Purpose of the Study:

• To evaluate the added value of an artificial intelligence (AI) algorithm in detecting pulmonary nodules on ULDCT.
• To compare the detection rates of pulmonary nodules by AI versus standard radiologist interpretation in the ED setting.
• To assess the trade-off between true positive and false positive findings when using AI for nodule detection.

Main Methods:

• Retrospective analysis of 870 patients from the OPTIMACT trial who underwent ULDCT.
• Prospective reading by ED radiologists followed by post hoc analysis using an AI deep learning algorithm for nodule detection (≥6 mm).
• Independent review by three chest radiologists to establish a true positive reference standard for both prospectively detected nodules and AI marks.

Main Results:

• AI detected 5.8 times more true positive pulmonary nodules requiring follow-up compared to prospective ED radiologist reporting (104 vs. 18).
• The use of AI resulted in a 42.9 times increase in false positive findings (1,758 vs. 41).
• A median of 1 AI mark per ULDCT was observed, with false positives predominantly found in patients with major abnormalities.

Conclusions:

• AI significantly enhances the detection of incidental pulmonary nodules on ULDCT in an emergency setting.
• The increased detection rate comes with a substantial trade-off of a higher false positive rate.
• AI may aid in early pulmonary cancer detection but requires careful consideration of its impact on workflow due to increased false positives.