QIR Suite

QIR Suite is a software system designed for the quantitative analysis of adult cardiovascular MR images, utilizing both standard computational algorithms and deep learning techniques for image processing and evaluation. Its purpose is to provide medical professionals with detailed cardiac function assessments, including ejection fraction and 2D flow studies, to aid in diagnostic decision-making without replacing the practitioner's ultimate diagnostic responsibility.

QIR Suite is a software application intended for qualified medical professionals to view, post-process, and perform quantitative evaluations of cardiovascular Magnetic Resonance (MR) images from adult patients. The software operates using the DICOM standard format and integrates QIR-MR for the analysis of MR images, employing a combination of user inputs, standard algorithms, and automated deep learning detection algorithms. Its core functionalities encompass the ability to receive, store, transmit, and manipulate medical MR/CT images, including support for JPEG2000 transfer syntaxes. It also offers client/server capabilities for connection to PACS and HL7 servers, visualization of 2D and 2D + time datasets, segmentation of regions of interest, and the measurement of distances and areas. The system provides comprehensive cardiac function MR analyses for all four chambers, including assessments of ejection fraction, local myocardial mass, diastolic function, myocardial thickness, and thickening, alongside 2D flow studies. Each analysis module generates an automated report, which can be stored on PACS or HL7 servers. Performance testing against predicate devices (Segment CMR and CVI42) demonstrated substantial equivalence, with correlation coefficients for quantitative parameters typically exceeding 0.97 (average >0.99) and absolute mean differences generally well below 10%, often under 5%. The deep learning algorithms achieved Dice coefficients around 0.89-0.91, indicating strong agreement with ground truth. QIR Suite is designed as a diagnostic support tool, with the final diagnosis remaining the responsibility of the qualified practitioner.

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 specific deep learning architectures are utilized in QIR Suite, and what were the key characteristics (e.g., size, diversity, pathologies) of the datasets used for their training and validation?; How does the performance of QIR Suite's deep learning algorithms for segmentation and quantitative analysis vary across different MR scanner manufacturers (Siemens, Philips, GE) and magnetic field strengths (1.5T, 3T) beyond the reported aggregate statistics?; Can you elaborate on the specific quantitative parameters derived from the '2D Flow studies' and how their accuracy compares to established reference methods or the predicate CVI42 in more detail than the <10% mean difference?; What impact does the use of QIR Suite have on the typical workflow and reporting time for cardiovascular MR image analysis in a clinical setting compared to manual methods or the predicate devices?; Given the validation for adult patients only, what are the anticipated challenges or modifications required to potentially extend QIR Suite's applicability to pediatric cardiovascular MR analysis in the future?