ScreenDx

ScreenDx is a software-only device that analyzes lung CT imaging data using deep learning to identify interstitial lung findings compatible with interstitial lung disease. It provides a binary positive/negative output to flag patients for potential specialist referral and further workup, supplementing but not replacing standard clinical diagnosis methods.

ScreenDx uses a 3-D deep learning algorithm trained on over 3,000 CT cases from multiple sources and manufacturers, whereas the predicate Fibresolve uses machine learning pattern recognition. Both process DICOM-compliant lung CT scans and output qualitative classification results without image alteration, segmentation, or localization. The key difference is ScreenDx screens a broad population for ILD findings, while Fibresolve differentiates within suspected ILD cases specifically for IPF classification.

Software Verification and Validation testing per IEC 62304 was performed to demonstrate safety based on current industry standards.

Both devices use machine learning pattern recognition on DICOM CT scans to provide qualitative imaging findings output as an adjunct to standard workflow for lung disease assessment. Both have the same intended user population (lung disease clinicians), same anatomical focus (chest), same input (DICOM lung CT), and same output format (qualitative classification without image alteration). Clinical performance testing demonstrated sensitivity of 91.4% and specificity of 95.2%, exceeding the prespecified 80% thresholds, with consistent performance across demographic subgroups and CT protocols. The algorithmic differences between screening a general population for ILD versus differentiating within confirmed ILD cases do not create new safety or effectiveness concerns, as both operate in parallel to standard care workflow without diagnostic intent.