qER

The qER device is a radiological computer-aided triage and notification software designed to expedite the review of non-contrast head CT scans by identifying potential critical findings such as intracranial hemorrhage, mass effect, midline shift, and cranial fracture. It employs a pre-trained artificial intelligence deep learning algorithm, featuring four independent convolutional neural networks, to analyze DICOM images and provide notifications with non-diagnostic previews to assist medical specialists in prioritizing their worklist, operating in parallel with standard care.

The qER system is a radiological computer-aided triage and notification software designed for the analysis of non-contrast head CT images. Its primary purpose is to assist hospital networks and trained medical specialists in workflow triage by flagging suspected positive findings of intracranial hemorrhage, mass effect, midline shift, and cranial fracture. The qER software, which can be deployed on-premise or on cloud servers, integrates with existing Picture Archiving and Communication Systems (PACS) and worklist systems. It analyzes DICOM files in parallel to the standard of care image interpretation, presenting notifications and non-diagnostic preview images for cases with suspected critical findings; however, it does not alter the original medical images and is not intended for direct diagnostic use, with clinicians remaining responsible for viewing full images. The core of qER's functionality lies in its artificial intelligence, specifically a deep learning analysis module comprising four independent, pre-trained convolutional neural networks (CNNs), each dedicated to detecting one of the target abnormalities. This module preprocesses CT scan series, applies the CNNs for classification, and then post-processes slice-level outputs to generate a scan-level triage result based on predefined thresholds. A retrospective, multicenter clinical study involving 1320 head CT scans demonstrated high sensitivity and specificity across all conditions, such as 96.98% sensitivity and 93.92% specificity for intracranial hemorrhage, and 98.53% sensitivity and 91.22% specificity for the presence of any of the four target abnormalities. The study also quantified clinical benefit, showing qER reduced the mean time-to-notification for critical scans to 2.11 minutes from a standard care mean time-to-open of 65.54 minutes.

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 performance of the qER algorithm vary with different CT scanner models, acquisition protocols, or image quality levels encountered in diverse clinical settings?; What is the real-world false positive notification rate for each of the four detected abnormalities, and what are the most common imaging features or artifacts that might lead to such false positives?; How does the provision of a 'non-diagnostic preview image' influence the subsequent diagnostic interpretation by the medical specialist, and are there any studies on potential cognitive biases, such as anchoring?; What were the specific architectures of the four independent convolutional neural networks, and what was the diversity and size of the datasets used for their pre-training and validation?; Beyond accelerating time-to-review, has any data been collected on the downstream impact of qER on patient outcomes, such as changes in treatment pathways or length of hospital stay for patients with critical findings?