Enhanced phase compensation in digital holographic microscopic imaging flow cytometry using radial basis function neural networks

Abstract

The digital holographic microscopy imaging flow cytometer (DHMIFC) facilitates label-free single-cell analysis of dynamic cell populations through the numerically reconstruction of complex wavefronts. Nevertheless, the recorded wavefronts are frequently affected by primary and higher-order aberrations, along with additional aberrations stemming from the digital holographic microscopy (DHM) optical system and surrounding experimental conditions. This study introduces a phase compensation algorithm leveraging radial basis function (RBF) neural networks to address phase aberrations in the distorted Peak function and holograms of PC3 cells. The performance of this algorithm is systematically compared with principal component analysis (PCA), digital phase mask (DPM), and spectrum centroid method (SCM). Results demonstrate that the proposed algorithm yields superior compensation outcomes. Moreover, the RBF neural network effectively mitigates spatial phase aberrations (SPA) without necessitating prior knowledge of the system’s optical parameters. Additionally, the runtime and interpolation accuracy of the algorithm can be fine-tuned by varying the number of sampling points, highlighting its potential applications in 3D surface reconstruction within DHMIFC.