Phase Retrieval from Bright-Field Microscope Images

Abstract: Bright-field microscopy is a very essential tool in medicine and biology. Thereby cells in bright-field microscope images often lack contracts. The reason for that is that they nearly do not change the amplitude but introduce a phase shift. With standard microscopy techniques like phase contrast microscopy this phase shifts can be made visible. However, they only produce qualitative phase images where the phase information is coupled non-linearly and cannot be retrieved quantitatively. With numerical methods the pahse can be obtained quantiatively out of several in-focus and defocused images. An iterative approach to retieve the phase information out of bright-field images of cells is implemented for the first time. in addition, a non-iterative approach and aproaches to improve it are investigated. Till now no quantitative investigations have been performed. In this work a first method is presented to quantitatively evaluate the different algorithms with real bright field microscope images of cells with the help of Mutual Information and phase contrast images. It is possible to calculate phase images without any additional hardware than a standard bright field microscope. This makes these numerical methods very interesting for different applications, since already existing systems can easily be upgraded. For the evaluation nine data stacks were recorded for adherent and suspended L929 cells. The evaluation shows that for adherent cells in some areas the non-iterative approach provides the best result while for suspended cells and cells in areas disturbed by the noise of the non-iterative approach the iterative approach performs better. The presented approaches to improve the non-iterative method were successful.