qXR-LN

qXR-LN is a deep learning-driven computer-aided detection software developed to aid physicians by identifying and localizing suspected pulmonary nodules (6-30mm) on adult frontal chest radiographs, functioning as an adjunctive second reader. Its methodology, validated through extensive clinical studies using a diverse dataset and multiple reader types, demonstrates enhanced diagnostic accuracy, evidenced by significant improvements in nodule detection sensitivity, reader agreement (AFROC), and a slight reduction in false positives when integrated into the radiological review process for incidental findings.

qXR-LN is a computer-aided detection (CADe) software designed to assist physicians in reviewing frontal (AP/PA) chest radiographs from adult patients. Its primary function is to identify and mark regions suspected of containing pulmonary nodules, sized 6 to 30 mm, in the incidental adult population, acting as a second reader to provide adjunctive information without replacing the original image or primary clinical assessment; the system utilizes a deep learning algorithm trained on a diverse global dataset of 2.5 million scans from over 25 X-ray equipment manufacturers, processing DICOM input images and producing DICOM output with highlighted potential nodules for review in standard viewers or PACS. The device's efficacy was established through comprehensive testing, including a standalone study and a multi-reader multi-case (MRMC) study: the standalone study, on a diverse U.S. dataset with CT-correlated ground truth, yielded a nodule-level sensitivity of 84.1%, an overall false positive per image (FPPI) of 0.18, and a scan-level AUC of 94.51%; the MRMC study, involving radiologists, pulmonologists, and ER physicians, demonstrated that using qXR-LN as a second reader significantly improved diagnostic performance, with the Alternative Free-response Receiver Operating Characteristic (AFROC) estimate increasing by 0.0762, reader AUC improving by 0.0697, nodule-level sensitivity rising by 11.96%, and the FPPI slightly decreasing by 0.0078 with software aid, underscoring its utility in identifying incidental findings including smaller nodules (6-10mm).

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 qXR-LN's performance in detecting nodules between 6-10mm specifically influence early-stage lung cancer detection pathways compared to systems with a 10mm minimum detection size?; What is the quantifiable impact of qXR-LN on reducing inter-reader variability among radiologists, emergency room physicians, and pulmonologists with varying experience levels when interpreting chest X-rays for incidental pulmonary nodules?; Considering the training dataset included 2.5 million scans from 5 countries, are there specific patient populations or imaging conditions where qXR-LN's performance (sensitivity/FPPI) shows notable deviation, and what are the underlying reasons?; What is the typical processing time per image for qXR-LN, and how does this integrate into high-volume clinical workflows without causing significant delays?; Beyond the reported AFROC and sensitivity improvements, what is the effect of qXR-LN on the localization accuracy of detected nodules, particularly for those near complex anatomical structures or in cases with confounding conditions?