ABSTRACT
Refractive index mismatch between the specimen and the objective immersion oil results in spherical aberration, which causes distortion and spreading of the point spread function, as well as incorrect readings of the axial coordinates. These effects have to be taken into account when performing three-dimensional restoration of wide-field fluorescence images. By using objects with well-defined geometry (fluorescently stained Escherichia coli or actin filaments) separated from a cover slip by a layer of oil with known refractive index, we investigated the accuracy of three-dimensional shape restoration by the commercial programs Huygens and Autoquant. Aberration correction available in the software dramatically reduced the axial blur compared to deconvolution that ignored the refractive index mismatch. At the same time, it failed to completely recover the cylindrical symmetry of bacteria or of actin fibres, which showed up to a three to five times larger width along the optical axis compared to the lateral plane. The quality of restoration was only moderately sensitive to the exact values of the specimen refractive index but in some cases improved significantly by assuming a reduced NA of the objective. Because image restoration depends on the knowledge of the vertical scale, we also performed detailed measurements of the axial scaling factor and concluded (in agreement with some previous authors) that scaling is adequately described by the simple paraxial formula, even when high-NA oil-immersion objectives are used.
