Lens-free holographic microscopy is a promising diagnostic approach because it is cost-effective, compact, and suitable for point-of-care applications, while providing high resolution together with an ultra-large field-of-view. It has been applied to biomedical sensing, where larger targets like eukaryotic cells, bacteria, or viruses can be directly imaged without labels, and smaller targets like proteins or DNA strands can be detected via scattering labels like micro- or nano-spheres. Automated image processing routines can count objects and infer target concentrations. In these sensing applications, sensitivity and specificity are critically affected by image resolution and signal-to-noise ratio (SNR). Pixel super-resolution approaches have been shown to boost resolution and SNR by synthesizing a high-resolution image from multiple, partially redundant, low-resolution images. However, there are several computational methods that can be used to synthesize the high-resolution image, and previously, it has been unclear which methods work best for the particular case of small-particle sensing. Here, we quantify the SNR achieved in small-particle sensing using regularized gradient-descent optimization method, where the regularization is based on cardinal-neighbor differences, Bayer-pattern noise reduction, or sparsity in the image. In particular, we find that gradient-descent with sparsity-based regularization works best for small-particle sensing. These computational approaches were evaluated on images acquired using a lens-free microscope that we assembled from an off-the-shelf LED array and color image sensor. Compared to other lens-free imaging systems, our hardware integration, calibration, and sample preparation are particularly simple. We believe our results will help to enable the best performance in lens-free holographic sensing.
Detecting, counting, and sizing nanoparticles is a key problem in biomedical, environmental, and materials synthesis fields. Here we demonstrate a cost-effective and high-performance approach that uses wide-field microscopy enabled by the combination of inline lensfree holography, pixel super-resolution, and vapor-condensed nano-scale lenses (nanolenses). These nanolenses are composed of liquid polyethylene glycol (PEG) that self-assembles in situ around particles of interest. A nanolens around each particle generates a more substantial phase shift than the native object alone, making it more easily detectible in the imaging system. This latest generation of lensfree holographic microscope incorporates more precise temperature control and utilizes a hermetically sealed chamber allowing for a controlled, repeatable environment for simultaneous hologram measurements and nanolens formation. To further enhance the sensitivity of our system, we have compared the performance of two different pixel super-resolution algorithms: shiftand- add and gradient descent. It was found that the gradient descent approach provides the highest resolution. Detection and localization results for 1 μm, 400 nm, and 100 nm particles are presented.