AI methods for PET images

It has always been in the interest of Tecnologie Avanzate to create a network between various Italian centers of excellence, in order to promote integration, collaboration and sharing. Ideas, cutting-edge studies, innovative algorithms are thus made available to everyone and not confined to a single center.

It is with this spirit that in the past years TA supported a multicenter study to evaluate the robustness and reliability of an adaptive threshold algorithm for BTV estimation working on PET images, described in [Brambilla et al., Med. Phys. 35 (4), 1207-1213, 2008]. Subsequently, we developed a 3D version of the algorithm, which was implemented in iTA DATA.

Thanks to the consolidated collaborations with the DET department of the Politecnico of Turin and with the Azienda Ospedaliera Universitaria of Novara, we are studying a deep learning system for the automatic segmentation of 3D lesions on PET images.

The algorithm, after a test and validation phase, will be made available in the software developed by TA, equipped with indispensable tools to make it usable (DICOM RT) in the clinical research field.