AIR Recon DL
GE HealthCare
Reconstructs MRI images from raw k-space data using a trained network, with the stated aim of shorter scan times at equivalent image quality.
At A Glance
- Manufacturer
- GE HealthCare
- Medical specialty
- Radiology
- Clearance route
- 510(k)
- Year authorized
- 2020
- Modality
- MRI
- Role in workflow
- Image reconstruction
Reconstruction AI is the quietest and arguably most consequential category on this list. It does not interpret anything. It changes what the image is, upstream of every other model and every human reader, which makes it a dependency for everything downstream.
Where It Is Used
- Scan time reduction across MRI protocols
- Signal-to-noise improvement at fixed acquisition time
- Improved throughput in constrained MRI capacity
Limitations Worth Knowing
Learned reconstruction can produce images that look clean while suppressing or, less commonly, fabricating fine detail. The relevant failure mode is not obvious artifact but plausible-looking output. Any downstream model validated on conventionally reconstructed images is operating outside its validation set once reconstruction changes.
Regulatory status, indications, and labeling change over time. Confirm current status on the FDA's AI/ML-Enabled Medical Devices list and in the manufacturer's current instructions for use before relying on anything stated here.