Frequency-based local content adaptive filtering algorithm for automated photoreceptor cell density quantification

Photoreceptor cells in the human eye play a vital role in vision. Certain retinal diseases cause the photoreceptor cells to degenerate and may lead to vision loss. Quantification of photoreceptor cell density from adaptive optics (AO) retinal images can provide valuable information and aid in the screening, diagnosis, and follow-up of retinal diseases. In this paper we describe an image model using a windowed two-dimensional (2D) lattice of pulses representing the cells and characterize the frequency content as decaying frequency domain pulses on the reciprocal lattice. Based on this model we propose a novel method for detection of cone photoreceptor cells by analyzing the discrete-space Fourier transform (DSFT) of AO retinal images. This method uses a small-extent block-based 2D discrete Fourier transform (DFT) to determine cell frequency content in order to obtain parameters of an adaptive circularly symmetric band-pass filter that is applied to the image. The filter extracts the underlying cellular structure and removes high-frequency noise as well as very low frequency contamination manifested as slow variations in the image. Subsequent detection yields an automated cell count that compares well with actual and manual counts on test and retinal images and demonstrates the accuracy of the method.