The following is a summary of “Automated Real-time Detection of Lung Sliding Using Artificial Intelligence: A Prospective Diagnostic Accuracy Study,” published in the February 2024 issue of Pulmonology by Fiedler et al.
In the realm of clinical urgency, the prompt assessment of pneumothorax (PTX) stands as a prevalent necessity. While lung ultrasound (LUS) is a standard tool for PTX evaluation, its diagnostic efficacy can fluctuate depending on patient and provider factors. Artificial intelligence (AI) aided imaging methodologies have emerged to bolster LUS performance in detecting pulmonary pathology. Nonetheless, the precise diagnostic accuracy of AI-assisted LUS (AI-LUS) in real-time applications for PTX diagnosis remains uncertain.
This study aimed to ascertain the real-time diagnostic precision of AI-LUS in identifying the absence of lung sliding in patients suspected of PTX. Employing a prospective design, we conducted an AI-assisted diagnostic accuracy investigation involving a convenience sample of individuals with suspected pneumothorax. Following the calibration of model parameters and imaging settings for bedside utility, we assessed the diagnostic accuracy of AI-LUS for lung sliding absence against an expert consensus reference standard.
The analysis, based on 241 lung sliding evaluations derived from 62 patients, revealed AI-LUS to exhibit a sensitivity of 0.921 (95% CI 0.792, 0.973), specificity of 0.802 (95% CI 0.735 – 0.856), an area under the curve of the receiver operating characteristic (AUC) of 0.885 (95% CI 0.828, 0.956), and an accuracy of 0.824 (95% CI 0.766 – 0.870) in diagnosing absent lung sliding.
Source: sciencedirect.com/science/article/abs/pii/S0012369224001570