Red Dot

The red dot™ device is an artificial intelligence-powered software designed to analyze adult chest X-rays for features suggestive of pneumothorax, providing case-level notifications to PACS/workstations for workflow prioritization. Its methodology relies on a visual recognition algorithm processing validated and anonymized DICOM images, with performance studies indicating high diagnostic accuracy (AUROC 0.975, Sensitivity 94.65%) and rapid processing times to aid in the timely triage of potential pneumothorax cases.

The red dot™ device by Behold.AI Technologies is a software workflow tool designed to assist in the clinical assessment of adult chest X-rays by identifying features suggestive of pneumothorax. It employs an artificial intelligence algorithm, specifically a visual recognition algorithm, to analyze PA/AP chest X-ray images received in DICOM format. The system integrates into the existing radiology workflow by first validating image readability and anonymizing patient identifiable information, then processing them through its AI model. If potential pneumothorax findings are detected, it sends a case-level notification via a standard JSON message to the Picture Archiving and Communication System (PACS) or Radiology Information System (RIS) workstation, enabling prioritization or triage of these cases for earlier review by radiologists; it does not mark images, direct attention to specific locations, or serve as a standalone diagnostic tool. Performance of the red dot™ device was evaluated in a multi-center, retrospective study using 888 anonymized adult chest X-ray images, including 299 pneumothorax positive cases and 589 negative cases, with ground truth established by at least two board-certified radiologists. The study demonstrated strong diagnostic performance for pneumothorax detection, achieving an Area Under the Receiver Operating Characteristic curve (AUROC) of 0.975. The accuracy was 90.20%, with a sensitivity of 94.65% and specificity of 87.95%. The system also exhibited efficient timing, with a mean image processing time of 13.8 seconds and an average notification transit time of 15.5 seconds, resulting in a total red dot™ performance time of approximately 29.3 seconds from image receipt to notification display on the worklist, indicating its potential to expedite the review of critical cases.

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. How does the red dot™ algorithm's performance vary with different types and sizes of pneumothoraces (e.g., small apical vs. large tension pneumothorax)?; What specific confounding imaging features were included in the non-normal control group, and how did they impact the algorithm's specificity?; What is the typical range of improvement in radiologist review time for pneumothorax-positive cases when using the red dot™ prioritization system in a real-world clinical setting, beyond the device's processing and notification time?; Beyond the reported AUROC, sensitivity, and specificity, what are the positive and negative predictive values of the red dot™ device given the prevalence of pneumothorax in the study population (299/888 cases)?; How does the image input, validation, and anonymization feature handle images with poor diagnostic quality or corrupted metadata, and are there any failure modes that could prevent an image from being processed or lead to misinterpretation?