| K-number | K243350 |
| Device name | Rapid Neuro3D |
| Applicant | Ischemaview, Inc. |
| Product code | QIH |
| Device class | Class II |
| Decision date | Jan 22, 2025 |
| Decision | Substantially Equivalent |
| Regulation | 892.2050 |
Rapid Neuro3D (RN3D) is image analysis software that processes CT angiography (CTA) scans of the head and neck to visualize and segment arterial vessels supplying blood to the brain. It removes bone, tissue, and venous vessels, producing 3D and 2D visualizations including maximum intensity projections (MIPS), volume renders (VR), and curved planar reformations (CPR). The software is designed to support physicians in confirming lesions, documentation, and treatment planning but is not intended as a stand-alone basis for clinical decision-making.
Rapid Neuro3D is a Software as a Medical Device using pre-trained AI/machine-learning models that operate on DICOM CTA images and return processed DICOM output. The predicate Rapid device processes broader imaging modalities (CT, CTP, CTA, C-arm CT, MRI) with mixed traditional and AI/ML implementation. Both are SaMD platforms that interface with PACS systems on standard off-the-shelf hardware (on-premises or cloud-hybrid), but RN3D is narrower in scope, focusing specifically on CTA neurological imaging with automated bone/tissue/vein removal and vessel segmentation.
ISO 14971:2019 (risk management), IEC 62304:2015 (software lifecycle), IEC 62366-1:2015 (usability), NEMA PS 3.1-3.20 (DICOM), and ISO 15223-1:2021 (labeling symbols). Device was designed using vulnerability assessments, software bill of materials (SBOM), and penetration testing for cybersecurity.
RN3D shares the same intended use as the predicate Rapid device (automated radiological image processing for neurological imaging analysis) and contains analogous CTA outputs (2D/3D vessel visualization, MIPS, automated segmentation). Both are SaMD packages interfacing with PACS on standard hardware. Performance validation testing demonstrated RN3D achieves 99.8% clinical accuracy for MIP images, 100% for SSE and CPR outputs, with segmentation accuracy (Dice 0.89–0.97) and labeling 100% accurate, demonstrating it reliably processes CTA for neurovasculature visualization equivalent to the predicate's established performance.
View the full FDA submission: accessdata.fda.gov