Recently, the biomedical tomographic diagnosis using near-infrared fluorescence, which is based on the inverse analysis, has been being highly improved to apply molecular biological techniques to DDS. In this study, a novel photo-dynamic diagnosis system was constructed as "Near-Infrared Photo-Dynamic Computed Diagnosis" (NIR-PDCT). Two excitation lights and a CCD detector are symmetrically located at the same side to object, i.e. reflection-type illumination, taking account of clinical applications. This system provides fluorescence tomographic images from multi-directionally captured intensity distributions emitted from object surface, based on convolution back-projection. In this system, proposed was a 2-dimensional deconvolution technique using directional-splitting shift-variant point spread function (PSF), in order to restore the tomographic image blurred by the light scattering effect having biological tissue. Experimental verification was carried out quantitatively using a bio-phantom embedding single tube of DDS agent AIPcS. Consequently, it was found to be easier to identify a fluorophore, suppressing light scattering degradation by PDCT, and recover fluorophore information, e.g. position and concentration.