Contrast agent dose reduction in computed tomography with deep learning using a conditional generative adversarial network

Abstract

Objectives: To reduce the dose of intravenous iodine-based contrast media (ICM) in CT through virtual contrast-enhanced images using generative adversarial networks.

Methods: Dual-energy CTs in the arterial phase of 85 patients were randomly split into an 80/20 train/test collective. Four different generative adversarial networks (GANs) based on image pairs, which comprised one image with virtually reduced ICM and the original full ICM CT slice, were trained, testing two input formats (2D and 2.5D) and two reduced ICM dose levels (-50% and -80%). The amount of intravenous ICM was reduced by creating virtual non-contrast series using dual-energy and adding the corresponding percentage of the iodine map. The evaluation was based on different scores (L1 loss, SSIM, PSNR, FID), which evaluate the image quality and similarity. Additionally, a visual Turing test (VTT) with three radiologists was used to assess the similarity and pathological consistency.

Results: The -80% models reach an SSIM of > 98%, PSNR of > 48, L1 of between 7.5 and 8, and an FID of between 1.6 and 1.7. In comparison, the -50% models reach a SSIM of > 99%, PSNR of > 51, L1 of between 6.0 and 6.1, and an FID between 0.8 and 0.95. For the crucial question of pathological consistency, only the 50% ICM reduction networks achieved 100% consistency, which is required for clinical use.

Conclusions: The required amount of ICM for CT can be reduced by 50% while maintaining image quality and diagnostic accuracy using GANs. Further phantom studies and animal experiments are required to confirm these initial results.

Key points: • The amount of contrast media required for CT can be reduced by 50% using generative adversarial networks. • Not only the image quality but especially the pathological consistency must be evaluated to assess safety. • A too pronounced contrast media reduction could influence the pathological consistency in our collective at 80%.

Keywords: Contrast media; Image processing, computer-assisted; Tomography, spiral computed.

Fig. 1

Distribution of the CT body region (left), age distribution within sex groups (middle), and frequency distribution for sex groups for the training (blue) and test (red) data (triangle indicates the mean)

Fig. 2

Schematic illustration of the generation of the used image pairs (input, target) through the combination of the VNC and the isolated ICM images

Fig. 3

The schematic network architecture (blue = encoder, red = decoder) and the flow of the input data (2D (A) and 2.5D (B)) through the network, resulting in the final prediction with the enhanced variant of the input image

Fig. 4

Scores and metrics for each model type (blue = −80%, red = −50%) as a boxplot. The calculation of the boxplots is based on the single volume scores from each ensemble model. The triangle within each boxplot symbolizes the mean of the respective input type for the relevant score or metric

Fig. 5

50% 2.5D ICM reduction network. Comparison of input, target, prediction, and difference image used and generated by the 2.5D model (−50%). The diffmap indicates the difference in HU between the target and the output images. The red regions indicate that the model predicted a lower ICM intensity (−50 HU) for a specific region, whereas blue regions indicate a higher ICM intensity (+50 HU) prediction for a region

Fig. 6

80% 2.5D ICM reduction Network. Comparison of input, target, prediction, and difference image used and generated by the 2.5D model (−80%). The diffmap indicates the difference in HU between the target and the output images. The red regions indicate that the model predicted a lower ICM intensity (−50 HU) for a specific region, whereas blue regions indicate a higher ICM intensity (+50 HU) prediction for a region