We propose a fully automatic technique to obtain aberration free quantitative phase imaging in digital holographic microscopy (DHM) based on deep learning. The traditional DHM solves the phase aberration compensation problem by manually detecting the background for quantitative measurement. This would be a drawback in real time implementation and for dynamic processes such as cell migration phenomena. A recent automatic aberration compensation approach using principle component analysis (PCA) in DHM avoids human intervention regardless of the cells’ motion. However, it corrects spherical/elliptical aberration only and disregards the higher order aberrations. Traditional image segmentation techniques can be employed to spatially detect cell locations. Ideally, automatic image segmentation techniques make real time measurement possible. However, existing automatic unsupervised segmentation techniques have poor performance when applied to DHM phase images because of aberrations and speckle noise. In this paper, we propose a novel method that combines a supervised deep learning technique with convolutional neural network (CNN) and Zernike polynomial fitting (ZPF). The deep learning CNN is implemented to perform automatic background region detection that allows for ZPF to compute the self-conjugated phase to compensate for most aberrations.

Cellularized collagen gels are a common model in tissue engineering, but the relationship between the microstructure and bulk mechanical properties is only partially understood. Multiphoton microscopy (MPM) is an ideal non-invasive tool for examining collagen microstructure, cellularity and crosslink content in these gels. In order to identify robust image parameters that characterize microstructural determinants of the bulk elastic modulus, we performed serial MPM and mechanical tests on acellular and cellularized (normal human lung fibroblasts) collagen hydrogels, before and after glutaraldehyde crosslinking. Following gel contraction over 16 days, cellularized collagen gel content approached that of native connective tissues (∼200 mg ml⁻¹). Young's modulus (E) measurements from acellular collagen gels (range 0.5-12 kPa) exhibited a power-law concentration dependence (range 3-9 mg ml⁻¹) with exponents from 2.1 to 2.2, similar to other semiflexible biopolymer networks such as fibrin and actin. In contrast, cellularized collagen gel stiffness (range 0.5-27 kPa) produced concentration-dependent exponents of 0.7 uncrosslinked and 1.1 crosslinked (range ∼5-200 mg ml⁻¹). The variation in E of cellularized collagen hydrogels can be explained by a power-law dependence on robust image parameters: either the second harmonic generation (SHG) and two-photon fluorescence (TPF) (matrix component) skewness (R²=0.75, exponents of -1.0 and -0.6, respectively); or alternatively the SHG and TPF (matrix component) speckle contrast (R²=0.83, exponents of -0.7 and -1.8, respectively). Image parameters based on the cellular component of TPF signal did not improve the fits. The concentration dependence of E suggests enhanced stress relaxation in cellularized vs. acellular gels. SHG and TPF image skewness and speckle contrast from cellularized collagen gels can predict E by capturing mechanically relevant information on collagen fiber, cell and crosslink density.

The human oral mucosa hosts a diverse microbiome and is exposed to potentially toxic biomaterials from dental restoratives. Mucosal health is partly determined by cell and tissue responses to challenges such as dental materials and pathogenic bacteria. An in vitro model to rapidly determine potential layer-specific responses would lead to a better understanding of mucosal homeostasis and pathology. Therefore, this study aimed to develop a co-cultured microfluidic mucosal model on-a-chip to rapidly assess mucosal remodeling and the responses of epithelial and subepithelial layers to challenges typically found in the oral environment. A gingival fibroblast-laden collagen hydrogel was assembled in the central channel of a three-channel microfluidic chamber with interconnecting pores, followed by a keratinocyte layer attached to the collagen exposed in the pores. This configuration produced apical and subepithelial side channels capable of sustaining flow. Keratinocyte, fibroblast, and collagen densities were optimized to create a co-culture tissue-like construct stable over one week. Cells were stained and imaged with epifluorescence microscopy to confirm layer characteristics. As proof-of-concept, the mucosal construct was exposed separately to a dental monomer, 2-hydroxylethyl methacrylate (HEMA), and the oral bacteria Streptococcus mutans. Exposure to HEMA lowered mucosal cell viability, while exposure to the bacteria lowered trans-epithelial electrical resistance. These findings suggest that the oral mucosa-on-a-chip is useful for studying oral mucosal interactions with bacteria and biomaterials with a histology-like view of the tissue layers.