FDA clears NeuroDR, AI for neurovascular CT scans

Deep Reasoning AI Inc. announced that the FDA granted 510(k) clearance (K262134) for NeuroDR, its AI-enabled software for visualizing the arteries of the head and neck from CT angiography scans, AuntMinnie reported. The software runs automatically once a scan completes, strips out bone, soft tissue and veins, and sends back 3D maximum intensity projections, volume renderings and curved planar reformats of the internal carotid and vertebral arteries — split into anterior and posterior circulation so physicians can review each without overlapping anatomy. A joint announcement from Deep Reasoning AI and Innolitics confirmed the clearance came 105 days after submission, with 3DR Labs — the largest U.S. provider of 3D medical image post-processing — running the performance studies and Innolitics managing the regulatory submission. Clinically, the output is meant to help physicians confirm and document lesions for follow-up and treatment planning, not to make an autonomous diagnosis.

What's notable for anyone building clinical software is the delivery model: NeuroDR doesn't introduce a new viewer or a new login. It writes standard DICOM back into the hospital's existing PACS, so the 3D renderings show up wherever radiologists already look — including the tablet and iPad-native apps that more imaging centers now use for bedside and on-call review, instead of forcing a trip back to a fixed workstation. A zero-click AI step that fattens every study with extra DICOM objects only reaches that bedside workflow if the mobile viewer rendering them is built to stay fast under load.

Why "zero-click" shifts the bottleneck to the viewer

Clearances like NeuroDR, Aview Aorta Metrics for chest CT, and Deep Breathe's pneumothorax-detection AI — all cleared within the same week — share a pattern: the AI step runs in the background and adds more series to every exam, not fewer. For a desktop PACS client with a dedicated GPU, that's a non-issue. For a native iPad or iPhone app pulling the same study over hospital Wi-Fi at 2 a.m., decoding and caching several extra 3D render series without stuttering is a real engineering constraint, not a formality.

What this means for clinics shipping their own mobile viewers

Clinics and imaging vendors building patient- or clinician-facing mobile apps on top of a PACS feed now need to plan for AI-augmented studies as the default, not the exception: progressive DICOM streaming, on-device caching that doesn't balloon with every new AI series, and a UI that lets a radiologist jump straight to the AI-generated reformats instead of scrolling past them. That's a build-time architecture decision for a native iOS app, not something a generic cross-platform wrapper handles well once image volumes climb.