Introducing advanced ClearIQ engine (AiCE) deep learning reconstruction algorithm into a clinical radiotherapy...

This work aimed to introduce Advanced intelligence Clear-IQ Engine (Canon, Japan) (AiCE) reconstruction to the clinical workflow by assessing the image quality and impact on radiotherapy treatment planning. This introduction would then allow for optimization of CT scanning doses.

A CT scan of an electron density phantom was performed using clinical protocols using AIDR3D and AiCE reconstructions at different mAs values. Clinical scans were reconstructed using AiCE and the dose distributions of the clinical plans recalculated. An optimization project was initiated for breast radiotherapy patients after the phantom testing proved the image quality is stable with reducing dose.

In the electron density phantom the difference between the Hounsfield Number of water for AIDR3D and AiCE images was 5-6 HU depending on the mAs. This was within IPEM 91 tolerances for HU variation. When comparing the clinical goals of plans calculated the dose difference was a maximum of 0.45 Gy. The median dose difference was a 0.02 Gy difference. The clinical doses were successfully reduced to the limit of what was tested in phantom measurements with the image quality being clinically acceptable and a 12% reduction in mean DLP.

AiCE has been evaluated to demonstrate HU stability that is suitable for use in Radiotherapy treatment planning. The breast CT protocol was optimized with the introduction of AiCE and the doses were reduced without affecting image quality.

AiCE is a novel reconstruction technique and has been used for clinical treatment planning and for optimizing CT doses.