Most of known light-scattering technologies, which allow one to separate spherical from non-spherical single particles, utilize either analysis of 2D light-scattering pattern or depolarization of light scattered. Both approaches force one to use high-sensitive detectors to provide a suitable signal to noise ratio for two-dimensionless photo matrix or for optical system with crossed polarizers. In this study, we introduce the method for discrimination of spherical and non-spherical single particles. The approach is based on measurement of leading, most intensive, element S11 of light-scattering matrix. To provide maximal signal to noise ratio we specified the light-scattering profile (LSP) in terms of integrated over azimuthal angle S11 as a function of polar scattering angle. The shape-sensitive vector-invariant for individual spherical particles was constructed from the parameters of LSP spectrum. The vector-invariant plays a role of the numerical criterion to identify spherical particles from LSPs. It can be applied to find a sphere with characteristics ranging from 16.5 to 70 and from 0.5 to 7.0 for size and phase-shift parameters respectively (size parameter α = πdn0/λ, where d – sphere diameter, λ – wavelength of the incident light, and n0 – medium refractive index, RI, phase-shift parameter ρ = 2α(m − 1), where relative RI m = n/n0 and n is the sphere RI). These ranges cover all possible characteristics of blood cells within the visible region of wavelengths. The ability of the vector-invariant to recognize spherical cells among non-spherical ones was tested theoretically by LSP databases of optical models of platelets and mature red blood cells. Moreover, experimentally the vector-invariant demonstrated good performance in searching of near-perfect spheres among milk fat globules, isolated nuclei of mononuclear cells, and completely spherized cells in a course of red blood cell lysis.