Introducing a new era
of image reconstruction.

deep learning image reconstruction


TrueFidelity images on a BMI 62 patient (400 lbs, 1.73 m)

Where deep learning
does its learning matters.

A deep learning image reconstruction application is only as good as the training it receives. GE Healthcare trained it's reconstruction engine using a library of thousands of low noise, filtered back projection (FBP) images considered the gold standard of image quality.

  • Designing

    Creating layers of mathematical equations, a Deep Neural Network (DNN) that can handle millions of parameters.

  • Training

    Inputting a high noise sinogram through the DNN and comparing the output image to a low noise version of the same image. These two images are compared across multiple parameters such as image noise, low contrast resolution, low contrast detectability, noise texture etc. The output image reports the differences to the network via backpropagation which trains and strengthens the DNN based on the desired output.

  • Verifying

    The network is required to reconstruct clinical and phantom cases it has never seen before, including extremely rare cases designed to push the network to its limits, confirming its robustness.

Confidence. Not compromise.

Compared with even the most sophisticated Model-Based Iterative Reconstruction, TrueFidelity CT Images are scanning taken to another level. Contrast visualization is maintained, noise and artifacts are minimized, edges are maintained—just enough—so there’s remarkable clarity and none of the compromise that comes with unfamiliar noise texture.1


See For Yourself

For current Revolution CT users: Contact your GE Healthcare representative to see your own images reconstructed using TrueFidelity.
It's time to get a closer look at a better way of seeing. Contact your GE Healthcare representative to learn more about GE Healthcare's TrueFidelity Images.

Resources

1. As demostrated in a clinical evalution consisting of 60 cases and 9 physicians, where each case was reconstructed with both DLIR and ASiR-V and evaluated by 3 of the physicians. In 100% of the reads, DLIR's image sharpness was rated the same as or better than ASiR-V's. In 91% of the reads, DLIR's noise texture was rated better than ASiR-V's. This rating was based on each individual reader's preference.

JB00138KO