Synthetic CT

The consolidated use of Cone Beam CT (CBCT) and the continuous evolution towards increasingly advanced imaging have led to the development of an algorithm for optimizing low-resolution images which guarantees a quality standard equivalent to CT while maintaining all the advantages of a CBCT acquisition.

Thanks to the collaboration with the Politecnico of Turin and the constant support offered by doctors and medical physicists in various reference centers, a conversion algorithm has been developed which, starting from the CBCT of the male pelvis district, generates a synthetic image that replicates the acquired anatomy of CBCT with a reconstruction of the tissues in terms of HU corresponding to a traditional CT.

The algorithm implements a Generative Adversarial Network (GAN) pix2pix Deep Learning network, also used for other applications, to create a target image starting from an input image. The algorithm is able to interpret CBCT from any machine (Varian®, Elekta®) and then convert it in a few seconds into a new dataset that can be exported in DICOM format.

To provide better usability, the system was implemented into the platform iTA Data, thus integrating all the functionalities for analysis and advanced evaluation of radiotherapy datasets.
The innovation of this algorithm is its independence from the reference CT.