HealthPNX

HealthPNX is an artificial intelligence-driven software tool designed to expedite the detection and prioritization of suspected pneumothorax cases from adult chest X-rays within a medical care environment. By analyzing images for suggestive findings and passively notifying the PACS/workstation, it aims to optimize radiologist workflow and reduce triage times for potentially critical conditions without providing a standalone diagnosis or altering the original medical images.

HealthPNX, developed by Zebra Medical Vision Ltd., is a radiological computer-assisted triage and notification software system designed to aid clinicians in the assessment of adult chest X-rays for features suggestive of pneumothorax. The system automatically analyzes PA/AP chest X-rays using an artificial intelligence algorithm. Key software modules include data input and validation to ensure image compatibility, the HealthPNX algorithm for identifying suspected pneumothorax findings, an IMA integration feature for notifying the PACS/workstation to facilitate worklist prioritization, and an error coding system for failed analyses. The device provides a passive notification to the radiologist through the PACS/workstation, indicating cases that may benefit from earlier review, without marking, highlighting, or directing attention to specific locations on the original X-ray image, thus aiding in prioritization and triage only. The performance of HealthPNX was evaluated in a pivotal study involving 588 anonymized chest X-ray cases, including 146 pneumothorax-positive cases, with ground truth established by three US Board Certified Radiologists. The software demonstrated an Area Under the Curve (AUC) of 98.3%, an overall agreement with ground truth of 93.03%, a sensitivity of 93.15%, and a specificity of 92.99%. A significant outcome was the reduction in triage time; HealthPNX enabled review in a mean time of 8.05 minutes compared to 68.98 minutes with standard 'First-in-First-out' care. The average time for the device to analyze a study and send a notification to the PACS worklist was 22.1 seconds, showcasing its efficiency for integration into clinical workflows.

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 HealthPNX algorithm differentiate between true pneumothorax and various confounding imaging factors or artifacts that might mimic its appearance on a chest X-ray?; What was the range of pneumothorax sizes (e.g., small, moderate, large, tension) included in the validation dataset, and does the device's sensitivity vary significantly with the subtlety or size of the pneumothorax?; Beyond triage time reduction, was there any assessment during the pivotal study of the HealthPNX system's impact on overall diagnostic accuracy or inter-reader variability among the radiologists reviewing the prioritized worklists?; How robust is the HealthPNX algorithm to variations in image quality, patient positioning, and different X-ray acquisition parameters commonly encountered across diverse clinical settings and equipment?; What are the specific technical requirements for integrating HealthPNX with various existing PACS/workstation environments, and what measures are in place to ensure data security and privacy during image transfer and analysis?