Research Article: Using machine-learning to optimize phase contrast in a low-cost cellphone microscope

Date Published: March 1, 2018

Publisher: Public Library of Science

Author(s): Benedict Diederich, Rolf Wartmann, Harald Schadwinkel, Rainer Heintzmann, Chulmin Joo.


Cellphones equipped with high-quality cameras and powerful CPUs as well as GPUs are widespread. This opens new prospects to use such existing computational and imaging resources to perform medical diagnosis in developing countries at a very low cost. Many relevant samples, like biological cells or waterborn parasites, are almost fully transparent. As they do not exhibit absorption, but alter the light’s phase only, they are almost invisible in brightfield microscopy. Expensive equipment and procedures for microscopic contrasting or sample staining often are not available. Dedicated illumination approaches, tailored to the sample under investigation help to boost the contrast. This is achieved by a programmable illumination source, which also allows to measure the phase gradient using the differential phase contrast (DPC) [1, 2] or even the quantitative phase using the derived qDPC approach [3]. By applying machine-learning techniques, such as a convolutional neural network (CNN), it is possible to learn a relationship between samples to be examined and its optimal light source shapes, in order to increase e.g. phase contrast, from a given dataset to enable real-time applications. For the experimental setup, we developed a 3D-printed smartphone microscope for less than 100 $ using off-the-shelf components only such as a low-cost video projector. The fully automated system assures true Koehler illumination with an LCD as the condenser aperture and a reversed smartphone lens as the microscope objective. We show that the effect of a varied light source shape, using the pre-trained CNN, does not only improve the phase contrast, but also the impression of an improvement in optical resolution without adding any special optics, as demonstrated by measurements.

Partial Text

In recent years the field of smart microscopy tried to enhance the user-friendliness as well as the image quaility of a standard microscope. Since then, the final output of the instrument thus can be more than what the user sees through the eyepiece. Taking a series of images and extract the phase information using the transport of intensity equation (TIE) [4], extracting the amplitude and phase from a hologram [5, 6] or capture multi-mode images such as darkfield, brightfield and qDPC [3, 7, 8] at the same time are just some examples.

According to the illumination principle introduced by Siedentopf [22] it is best practise in order to enhance the contrast in the object plane, by leaving out the illumination direction which do not contribute to the image formation.

Even after optimizing the code by replacing the 4D-TCC with only two convolution-kernels following Eq (16) to simulate one intensity image, the process of finding an optimized set of the parameters takes about 20 s on a quadcore computer, which is not reasonable for biologists or high throughput-applications, such as drug screening or industrial metrology, who needs results in real-time.

For proving, that the numerical optimization enhances the phase-contrast of a brightfield microscope, a standard upright research microscope ZEISS Axio Examiner.Z1 equiped with home-made SLM using a high-resolution smartphone display (iPhone 4S, Apple, USA) in the condenser plane, was used for the first tests. In comparison to LED-condensers, as presented in [37], a ZEMAX simulation of the koehler-illumination using an LCD in the aperture plane shows two-times better light-efficiency, which also improves the SNR and allows lower exposure times of the camera.

The proposed methods to improve the phase contrast by manipulating the condenser aperture was first tested on a standard research microscope (ZEISS Axio Examiner.Z1) with a magnification M = 20×, NA = 0.5 in air (ZEISS Plan-Neofluar) at a center-wavelength of λ = 530nm ± 20nm to give a proof-of-principle before the method was evaluated on the cellphone microscope.

In most cases the computational and imaging potential of a cellphone is not fully exhausted. They serve an integrated framework with an already existing infrastructure for image acquisition and hardware synchronization, as well as for rapid development of user-defined image processing applications.




0 0 vote
Article Rating
Notify of
Inline Feedbacks
View all comments