A new deconvolution method using antidiffusion for high resolution radioisotopic imaging

Radioisotopic investigations are very useful for functional diagnosis. However, visual analysis of images is lowered by blurring due to impulse response size and by their granular aspect due to low statistics. Acquired images result in the convolution between original object and detector spatial impulse response. Classical deconvolution methods used for radioisotopic image restoration fail due to bad signal to noise ratio. A new deconvolution process including two steps is designed. The first stage reduces acquired images statistical noise by applying a second order Butterworth low-pass filter. As the detector response shape may also be viewed as a 2D heat diffusion process, the deconvolution is done by running the diffusion equation into minus time. Tests were performed on physical phantom and on clinical thyroid scintigraphies acquired with IRIS (ARIES) high resolution gamma imaging probe. On physical phantom, cold spherical nodules as small as 4 mm in diameter are clearly enhanced on restored images, and diagnosis is improved.