Recently, artificial intelligence algorithms reducing noise by deep learning have been developed with application to SPECT and PET images.

Many studies have reported the possibility of reducing the recording time in bone scintigraphy by applying artificial intelligence algorithms reducing noise

Only two studies compared images denoised by a Deep Learning algorithm to those denoised by conventional filters (Gaussian and median filters). The first study was conducted only on patients, without phantom analysis and without taking into account the size of the lesions. The second study included an analysis on phantom and patients, but with application to planar images rather than to SPECT images that are increasingly used today

The hypothesis of our study conducted on phantom and patients is that an artificial intelligence algorithm reducing noise could replace the conventional filters usually used in bone SPECT for the denoising of scintigraphic images.

Inclusion Criteria:

Patients who had a whole-body thee dimensions bone scan for rheumatological or oncological indications.

Exclusion Criteria:

Patients opposed to the use of their data

Patients who had a whole-body thee dimensions bone scan for rheumatological or oncological indications and non opposed to th use of their data