SmartChest

SmartChest is an AI-powered software designed to analyze adult and transitional adolescent frontal chest X-rays for suspected pneumothorax and pleural effusion, providing passive notifications to PACS for worklist prioritization. Its methodology relies on a trained convolutional neural network to identify critical findings, with performance validated through standalone studies demonstrating high sensitivity and specificity, aiming to assist radiologists by flagging potentially urgent cases without altering the standard diagnostic review process.

SmartChest is a radiological computer-assisted triage and notification software designed to analyze frontal chest X-ray images (PA or AP) from transitional adolescents (18-21 years) and adults (22 years and older). It employs an artificial intelligence algorithm to detect features suggestive of suspected pleural effusion and/or pneumothorax. The software integrates with existing PACS or other DICOM storage platforms, receiving images and then passively sending case-level notifications back to the worklist. These notifications, conveyed via a new DICOM series and tags, categorize studies as "SUSPECTED FINDING," "CASE PROCESSED," or "NOT PROCESSED," and specify the type of suspected finding if present, aiding in the prioritization of potentially critical cases for review by qualified medical specialists. The AI model was trained on 9,560 X-ray images from French institutions. Standalone performance was validated in two studies using US-representative datasets (300 cases each), independent of the training data, with ground truth established by three board-certified radiologists. For pneumothorax detection, SmartChest achieved an ROC AUC of 0.989, sensitivity of 92.7%, and specificity of 97.3%. For pleural effusion, it demonstrated an ROC AUC of 0.975, sensitivity of 93.3%, and specificity of 90.0%. Mean execution times were approximately 2.3 seconds on a local server and 28 seconds on a cloud server. Importantly, SmartChest is a prioritization tool only, does not highlight specific image regions, and its results are not intended for standalone diagnostic decision-making; all images, including unflagged ones, still require full interpretation by a specialist.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. What were the specific criteria or thresholds used by the AI algorithm to flag a case as "SUSPECTED FINDING" for pleural effusion and pneumothorax?; How does the performance (sensitivity, specificity, AUC) of SmartChest vary across different X-ray machine manufacturers or imaging protocols, beyond what was generally described for the diverse dataset?; What is the expected impact on radiologist workflow efficiency and time-to-diagnosis for critical findings in a real-world clinical setting using SmartChest for prioritization?; Were there any subgroups within the test population (e.g., based on age, sex, specific comorbidities, or subtlety of findings) where the algorithm's performance significantly differed from the overall reported metrics?; Given the training data was from French institutions, what measures were taken during the US-representative validation to ensure the model's robustness to potential variations in image acquisition, patient demographics, and disease presentation in the US population?