qER-Quant

qER-Quant is an automated software tool designed to quantify critical brain structures like intracranial hyperdensities, lateral ventricles, and midline shift from non-contrast head CT scans using convolutional neural networks. Its methodology involves processing DICOM images to provide volumetric measurements and longitudinal analysis, aiming to support clinical assessment by automating and standardizing these quantifications.

qER-Quant is a software device designed for the automated analysis of non-contrast head CT (NCCT) scans. Its primary purpose is to automatically label, visualize, and quantify segmentable brain structures, thereby automating the traditionally manual process of identifying, labeling, and measuring the volume of these structures. The software analyzes NCCT images taken at a single time point to provide volumetric data. Additionally, it can perform longitudinal comparative analysis if multiple images from the same patient, acquired using the same scanner and protocol at different times, are available. The specific brain structures qER-Quant is indicated to analyze include intracranial hyperdensities, lateral ventricles, and midline shift, providing clinicians with quantitative metrics for these features. The qER-Quant system operates as standalone software, receiving NCCT scans in DICOM format from a Picture Archiving and Communication System (PACS) and returning its analysis results, in both PDF and DICOM formats, back to the PACS or another user-specified system. The core of its analytical capability lies in a set of pre-trained convolutional neural networks (CNNs), which process the DICOM images after a pre-processing stage and before a post-processing stage that generates visual and tabular outputs. Performance testing demonstrated its accuracy against manually labeled ground truth images, with reported mean absolute errors for intracranial hyperdensities at 6.56 ml (Dice score 0.75), midline shift at 1.37 mm, and lateral ventricles (left and right) at 2.09 ml (Dice 0.79) and 2.18 ml (Dice 0.75) respectively, meeting pre-set acceptance criteria.

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 qER-Quant's CNNs vary with different CT scanner manufacturers, models, or image acquisition parameters, given it operates on NCCT images?; What is the clinical impact of the reported mean absolute errors (e.g., 6.56 ml for intracranial hyperdensities) on diagnostic decision-making or patient management for the indicated conditions?; Since the predicate device Icobrain analyzes basal cisterns on NCCT while qER-Quant does not, what was the rationale for excluding this structure, and are there plans to include it in future iterations?; Could the longitudinal analysis feature of qER-Quant be adapted to predict the progression of conditions associated with changes in intracranial hyperdensities, lateral ventricle volume, or midline shift?; How does the system handle images with significant artifacts or atypical anatomy, and what are the failure modes or limitations of the CNN-based segmentation in such scenarios?