Three-dimensional deconvolution based on axial-scanning model for structured illumination microscopy


Current model-based deconvolution methods for three-dimensional structured illumination microscopy (3D-SIM) assume that the observed image can be modeled as the convolution of the fluorescence emission with the detection point spread function. However, such a model is not suitable when 3D data is acquired using axial scanning of the sample as in the case of commercial microscopes, because the structured illumination (SI) pattern is changing axially. Here, we implement and test an iterative deconvolution approach based on a 3D model that takes into account data-acquisition with axial scanning, by minimizing the mean squared error between the data and the model using a conjugate-gradient method. This method is applicable for 3D-SIM systems in which the SI pattern is separable into axial and lateral functions. Numerical results show improvement in the restoration when a model mismatch is avoided.

Publication Title

Proceedings - International Symposium on Biomedical Imaging