Hybrid Artificial Intelligence Solution Combining Convolutional Neural Network and Analytical Approach Showed Higher Accuracy in A-lines Detection on Lung Ultrasound in Thoracic Surgery Patients Compared with Radiology Resident.


  Vol. 45 (3) 2024 Neuro endocrinology letters Journal Article   2024; 45(3): 229-237 PubMed PMID:  39146568    Citation

OBJECTIVES: Lung ultrasound reduces the number of chest X-rays after thoracic surgery and thus the radiation. COVID-19 pandemic has accelerated research in lung ultrasound artifacts detection using artificial intelligence. This study evaluates the accuracy of artificial intelligence in A-lines detection in thoracic surgery patients using a novel hybrid solution that combines convolutional neural networks and analytical approach and compares it with a radiology resident and radiology experts' results. DESIGN: Prospective observational study. MATERIAL AND METHODS: Single-center study evaluates the accuracy of artificial intelligence and a radiology resident in A-line detection on lung ultrasound footages compared with the consensual opinion of two expert radiologists as the reference. After resident's first reading, the artificial intelligence results were presented to the resident and he was asked to revise the results based on artificial intelligence. RESULTS: 82 consecutive patients underwent 82 ultrasound examinations. 328 ultrasound recordings were evaluated. Accuracy, sensitivity, specificity, positive and negative predictive values of artificial inelligence in A-line detection were 0.866, 0.928, 0.834, 0.741 and 0.958 respectively. The resident's values were 0.558, 0.973, 0.346, 0.432 and 0.962 respectively. The resident's values after correction based on artificial intelligence results were 0.854, 0.991, 0.783, 0.701 and 0.994 respectively. CONCLUSION: Artificial intelligence showed high accuracy in A-line detection in thoracic surgery patients and was more accurate compared to a resident. Artificial intelligence could play important role in lung ultrasound artifact detection in thoracic surgery patients and in residents' education.


 Full text PDF