Lung AI (LAI001)

Lung AI is a computer-aided detection (CADe) software tool that analyzes lung ultrasound cine clips to assist healthcare professionals in detecting consolidation/atelectasis and pleural effusion. It is an adjunctive tool that alerts users to regions of interest (ROIs) on ultrasound images from the PLAPS point using the BLUE protocol. The device is intended for use by trained emergency department physicians and does not provide diagnosis or replace standard clinical care.

Lung AI uses supervised deep learning with convolutional neural networks for segmentation, landmark detection, and classification of ultrasound images. Unlike the predicate device (Lung-CAD), which analyzes digital X-ray images, Lung AI processes multi-frame ultrasound cine clips. Both devices use artificial intelligence as a concurrent reading aid for trained professionals on adult patients, but Lung AI is designed specifically for ultrasound modality in emergency departments rather than chest radiography.

IEC 62304:2006/AC:2015 (Medical device software – Software life cycle processes). The FDA guidances cited include Content of Premarket Submissions for Device Software Functions, Computer-Assisted Detection Devices Applied to Radiology Images and Radiology Device Data, Clinical Performance Assessment for CAD Devices, and Cybersecurity in Medical Devices.

Lung AI is substantially equivalent because it shares the same intended use (adjunctive CADe tool for trained healthcare professionals in adult point-of-care settings) and the same principle of operation (supervised deep learning artificial intelligence) as the predicate Lung-CAD. Although the modality differs (ultrasound vs. X-ray), both devices perform similar detection functions with comparable performance metrics. Bench testing on 465 lung scans achieved sensitivity of 0.97 and specificity of 0.91-0.94, and MRMC clinical validation with six emergency physicians demonstrated statistically significant improvements in reader performance (AUC improvements of 0.028-0.035), meeting the acceptance criteria. The device does not raise new safety or effectiveness questions when used as intended.