TrueFidelity DLIR (GE Deep-Learning CT Reconstruction)
GE HealthCare's deep-learning image reconstruction stack for CT — the FDA-cleared (2019) trained-neural-network reconstruction algorithm replacing or supplementing iterative reconstruction (ASiR-V) in routine clinical use. TrueFidelity is offered as a license-tier feature across the current Revolution Apex / Revolution CT / Revolution Ascend platforms and is propagated retroactively to selected legacy systems where compute hardware allows.
The deep-learning-reconstruction (DLIR) category as a whole is one of the principal differentiators in current-generation CT. Each major OEM has a competing implementation: GE TrueFidelity, Canon AiCE, Siemens Quantum Iterative Reconstruction (QIR) + Deep Resolve (Deep Resolve is currently MRI-side; Siemens CT uses ADMIRE / QIR). Philips Precise Image (the analog on the Incisive / Incisive CV platforms). Architecturally these are all neural-network-based reconstructions trained on paired low-dose / high-dose data — the differentiation is in training-data scale, network architecture, and validation methodology rather than fundamental physics.
Why buyers care
- Dose reduction at equivalent image quality — typical claims 30–50% dose reduction vs filtered back projection at equivalent task-based image quality. Clinical-protocol re-engineering required to capture the benefit.
- Image-quality improvement at fixed dose — alternative use case at the same dose, image is sharper and lower-noise.
- Workflow — TrueFidelity reconstruction is fast on appropriately-sized reconstruction hardware (real-time or near-real-time on premium platforms).
- License-tier dependent — TrueFidelity is not a standard feature; it's a license that adds materially to capital-deal pricing and to refurb / used-system inheritance.
Why engineers care
- Reconstruction-engine hardware dependence — DLIR requires substantial GPU / accelerator compute. Older reconstruction engines may not support TrueFidelity even on system platforms that nominally allow it. Refurb-CT due-diligence on TrueFidelity-equipped systems must verify reconstruction-hardware compatibility, not just the license.
- Smart Subscription — GE has shifted DLIR licensing to a subscription model (Smart Subscription) that decouples ongoing AI / app updates from the hardware refresh cycle. Sites buying refurb CT must verify subscription state, not just initial-license inheritance.
- Validation / acceptance — clinical adoption of DLIR requires task-based image-quality validation; sites doing this rigorously distinguish DLIR-reconstructed images that maintain clinically-relevant features from images that are merely lower-noise.
- Model-update implications — neural-network reconstructions can update with retraining; OEM model-update events are part of the ongoing service relationship.
Refurb implications
- License + subscription state inheritance is critical — a refurb buyer expecting TrueFidelity must verify both the license is transferable and the Smart Subscription is current.
- Reconstruction-hardware verification — older reconstruction engines may not support TrueFidelity even with the license.
- Comparable competitor licenses (AiCE, Precise Image, ADMIRE / Deep Resolve) face analogous refurb due-diligence patterns.
Related platforms
- GE Revolution Apex
- GE Revolution CT
- GE Revolution HD
- GE Revolution Ascend
- Discovery CT750 HD (legacy — TrueFidelity availability platform-dependent)