One-shot phase-recovery using a cellphone RGB camera on a Jamin-Lebedeff microscope.
Benedict DiederichBarbora MarsikovaBrad AmosRainer HeintzmannPublished in: PloS one (2019)
Jamin-Lebedeff (JL) polarization interference microscopy is a classical method for determining the change in the optical path of transparent tissues. Whilst a differential interference contrast (DIC) microscopy interferes an image with itself shifted by half a point spread function, the shear between the object and reference image in a JL-microscope is about half the field of view. The optical path difference (OPD) between the sample and reference region (assumed to be empty) is encoded into a color by white-light interference. From a color-table, the Michel-Levy chart, the OPD can be deduced. In cytology JL-imaging can be used as a way to determine the OPD which closely corresponds to the dry mass per area of cells in a single image. Like in other interference microscopy methods (e.g. holography), we present a phase retrieval method relying on single-shot measurements only, thus allowing real-time quantitative phase measurements. This is achieved by adding several customized 3D-printed parts (e.g. rotational polarization-filter holders) and a modern cellphone with an RGB-camera to the Jamin-Lebedeff setup, thus bringing an old microscope back to life. The algorithm is calibrated using a reference image of a known phase object (e.g. optical fiber). A gradient-descent based inverse problem generates an inverse look-up-table (LUT) which is used to convert the measured RGB signal of a phase-sample into an OPD. To account for possible ambiguities in the phase-map or phase-unwrapping artifacts we introduce a total-variation based regularization. We present results from fixed and living biological samples as well as reference samples for comparison.
Keyphrases
- high resolution
- high speed
- deep learning
- single molecule
- high throughput
- magnetic resonance
- gene expression
- optical coherence tomography
- machine learning
- mass spectrometry
- working memory
- convolutional neural network
- label free
- computed tomography
- cell death
- magnetic resonance imaging
- cell proliferation
- single cell
- contrast enhanced