Signal to Noise Ratio as a Cross-Platform Metric for Intraoperative Fluorescence Imaging.

Real-time molecular imaging to guide curative cancer surgeries is critical to ensure removal of all tumor cells; however, visualization of microscopic tumor foci remains challenging. Wide variation in both imager instrumentation and molecular labeling agents demands a common metric conveying the ability of a system to identify tumor cells. Microscopic disease, comprised of a small number of tumor cells, has a signal on par with the background, making the use of signal (or tumor) to background ratio inapplicable in this critical regime. Therefore, a metric that incorporates the ability to subtract out background, evaluating the signal itself relative to the sources of uncertainty, or noise is required. Here we introduce the signal to noise ratio (SNR) to characterize the ultimate sensitivity of an imaging system and optimize factors such as pixel size. Variation in the background (noise) is due to electronic sources, optical sources, and spatial sources (heterogeneity in tumor marker expression, fluorophore binding, and diffusion). Here, we investigate the impact of these noise sources and ways to limit its effect on SNR. We use empirical tumor and noise measurements to procedurally generate tumor images and run a Monte Carlo simulation of microscopic disease imaging to optimize parameters such as pixel size.

3D Reconstruction of cellular images from microfabricated imagers using fully-adaptive deep neural networks.

Millimeter-scale multi-cellular level imagers enable various applications, ranging from intraoperative surgical navigation to implantable sensors. However, the tradeoffs for miniaturization compromise resolution, making extracting 3D cell locations challenging-critical for tumor margin assessment and therapy monitoring. This work presents three machine-learning-based modules that extract spatial information from single image acquisitions using custom-made millimeter-scale imagers. The neural networks were trained on synthetically-generated (using Perlin noise) cell images. The first network is a convolutional neural network estimating the depth of a single layer of cells, the second is a deblurring module correcting for the point spread function (PSF). The final module extracts spatial information from a single image acquisition of a 3D specimen and reconstructs cross-sections, by providing a layered "map" of cell locations. The maximum depth error of the first module is 100 µm, with 87% test accuracy. The second module's PSF correction achieves a least-square-error of only 4%. The third module generates a binary "cell" or "no cell" per-pixel labeling with an accuracy ranging from 89% to 85%. This work demonstrates the synergy between ultra-small silicon-based imagers that enable in vivo imaging but face a trade-off in spatial resolution, and the processing power of neural networks to achieve enhancements beyond conventional linear optimization techniques.

Microfluidic Packaging Integration with Electronic-Photonic Biosensors Using 3D Printed Transfer Molding.

Multiplexed sensing in integrated silicon electronic-photonic platforms requires microfluidics with both high density micro-scale channels and meso-scale features to accommodate for optical, electrical, and fluidic coupling in small, millimeter-scale areas. Three-dimensional (3D) printed transfer molding offers a facile and rapid method to create both micro and meso-scale features in complex multilayer microfluidics in order to integrate with monolithic electronic-photonic system-on-chips with multiplexed rows of 5 µm radius micro-ring resonators (MRRs), allowing for simultaneous optical, electrical, and microfluidic coupling on chip. Here, we demonstrate this microfluidic packaging strategy on an integrated silicon photonic biosensor, setting the basis for highly multiplexed molecular sensing on-chip.