Machine learning at the edge for AI-enabled multiplexed pathogen detection.

Multiplexed detection of biomarkers in real-time is crucial for sensitive and accurate diagnosis at the point of use. This scenario poses tremendous challenges for detection and identification of signals of varying shape and quality at the edge of the signal-to-noise limit. Here, we demonstrate a robust target identification scheme that utilizes a Deep Neural Network (DNN) for multiplex detection of single particles and molecular biomarkers. The model combines fast wavelet particle detection with Short-Time Fourier Transform analysis, followed by DNN identification on an AI-specific edge device (Google Coral Dev board). The approach is validated using multi-spot optical excitation of Klebsiella Pneumoniae bacterial nucleic acids flowing through an optofluidic waveguide chip that produces fluorescence signals of varying amplitude, duration, and quality. Amplification-free 3× multiplexing in real-time is demonstrated with excellent specificity, sensitivity, and a classification accuracy of 99.8%. These results show that a minimalistic DNN design optimized for mobile devices provides a robust framework for accurate pathogen detection using compact, low-cost diagnostic devices.

Adaptive time modulation technique for multiplexed on-chip particle detection across scales.

Integrated optofluidic biosensors have demonstrated ultrasensitivity down to single particle detection and attomolar target concentrations. However, a wide dynamic range is highly desirable in practice and can usually only be achieved by using multiple detection modalities or sacrificing linearity. Here, we demonstrate an analysis technique that uses temporal excitation at two different time scales to simultaneously enable digital and analog detection of fluorescent targets. We demonstrated the seamless detection of nanobeads across eight orders of magnitude from attomolar to nanomolar concentration. Furthermore, a combination of spectrally varying modulation frequencies and a closed-loop feedback system that provides rapid adjustment of excitation laser powers enables multiplex analysis in the presence of vastly different concentrations. We demonstrated this ability to detect across scales via an analysis of a mixture of fluorescent nanobeads at femtomolar and picomolar concentrations. This technique advances the performance and versatility of integrated biosensors, especially toward point-of-use applications.

Fast custom wavelet analysis technique for single molecule detection and identification.

Many sensors operate by detecting and identifying individual events in a time-dependent signal which is challenging if signals are weak and background noise is present. We introduce a powerful, fast, and robust signal analysis technique based on a massively parallel continuous wavelet transform (CWT) algorithm. The superiority of this approach is demonstrated with fluorescence signals from a chip-based, optofluidic single particle sensor. The technique is more accurate than simple peak-finding algorithms and several orders of magnitude faster than existing CWT methods, allowing for real-time data analysis during sensing for the first time. Performance is further increased by applying a custom wavelet to multi-peak signals as demonstrated using amplification-free detection of single bacterial DNAs. A 4x increase in detection rate, a 6x improved error rate, and the ability for extraction of experimental parameters are demonstrated. This cluster-based CWT analysis will enable high-performance, real-time sensing when signal-to-noise is hardware limited, for instance with low-cost sensors in point of care environments.

Free-Space Excitation of Optofluidic Devices for Pattern-Based Single Particle Detection.

Optofluidic sensors have enabled single molecule sensing using planar, waveguide dependent multi-spot fluorescence excitation. Here, we demonstrate a new approach to single-particle fluorescence sensing using free-space, top-down illumination of liquid-core antiresonant reflecting optical waveguide (ARROW) devices using two different multi-spot excitation techniques. First, the liquid core ARROW waveguide is excited with a focused beam through a slit pattern milled into an opaque aluminum film, showing comparable performance for single bead fluorescence detection as in-plane, multi-mode interference waveguide based excitation. The second top-down illumination technique images the spot pattern from a Y-splitter SiO2 waveguide chip directly onto the detection device for efficient power utilization and circumventing the need for an opaque cover, producing a further 2.7x improvement in signal-to-noise ratio. The two top-down approaches open up new possibilities for chip-based optical particle sensing with relaxed alignment tolerances.

3D Hydrodynamic Focusing in Microscale Optofluidic Channels Formed with a Single Sacrificial Layer.

Optofluidic devices are capable of detecting single molecules, but greater sensitivity and specificity is desired through hydrodynamic focusing (HDF). Three-dimensional (3D) hydrodynamic focusing was implemented in 10-µm scale microchannel cross-sections made with a single sacrificial layer. HDF is achieved using buffer fluid to sheath the sample fluid, requiring four fluid ports to operate by pressure driven flow. A low-pressure chamber, or pit, formed by etching into a substrate, enables volumetric flow ratio-induced focusing at a low flow velocity. The single layer design simplifies surface micromachining and improves device yield by 1.56 times over previous work. The focusing design was integrated with optical waveguides and used in order to analyze fluorescent signals from beads in fluid flow. The implementation of the focusing scheme was found to narrow the distribution of bead velocity and fluorescent signal, giving rise to 33% more consistent signal. Reservoir effects were observed at low operational vacuum pressures and a balance between optofluidic signal variance and intensity was achieved. The implementation of the design in optofluidic sensors will enable higher detection sensitivity and sample specificity.

3D hydrodynamic focusing in microscale channels formed with two photoresist layers.

3D hydrodynamic focusing was implemented with channel cross-section dimensions smaller than 10 µm. Microchannels were formed using sacrificial etching of two photoresist layers on a silicon wafer. The photoresist forms a plus-shaped prismatic focusing fluid junction which was coated with plasma-enhanced chemical-vapor-deposited oxide. Buffer fluid carried to the focusing junction envelopes an intersecting sample fluid, resulting in 3D focusing of the sample stream. The design requires four fluid ports and operates across a wide range of fluid velocities through pressure-driven flow. The focusing design was integrated with optical waveguides to interrogate fluorescing particles and confirm 3D focusing. Particle diffusion away from a focused stream was characterized.

Multi-channel velocity multiplexing of single virus detection on an optofluidic chip.

Liquid-core waveguide-based optofluidic devices have proven to be valuable tools for analysis of biological samples in fluid. They have enabled single bioparticle sensitivity while maintaining in-plane detection via light-induced fluorescence. The incorporation of multi-spot excitation with multimode interference (MMI) waveguides has enabled spatially and spectrally multiplexed detection of single viruses on an oxide-based optofluidic platform. Here, we introduce a new way of MMI-based multiplexing where multiple analysis channels are placed within a single multi-spot pattern. This stacked channel design enables both velocity and spectral multiplexing of single particles. The principle is demonstrated with differentiated detection of single H3N2 and H1N1 viruses on a polydimethylsiloxane platform.

Optimized ARROW-Based MMI Waveguides for High Fidelity Excitation Patterns for Optofluidic Multiplexing.

Multimode interference (MMI) waveguides can be used for multiplexing and de-multiplexing optical signals. High fidelity, wavelength dependent multi-spot patterns from MMI waveguides are useful for sensitive and simultaneous identification of multiple targets in multiplexed fluorescence optofluidic biosensors. Through experiments and simulation, this paper explores design parameters for an MMI rib anti-resonant reflecting optical waveguide (ARROW) in order to produce high fidelity spot patterns at the liquid core biomarker excitation region. Width and etch depth of the single excitation rib waveguide used to excite the MMI waveguide are especially critical because they determine the size of the input optical mode which is imaged at the MMI waveguide's output. To increase optical throughput into the MMI waveguide when light is coupled in from an optical fiber, tapers in the waveguide width can be used for better mode matching.

Buried Rib SiO2 Multimode Interference Waveguides for Optofluidic Multiplexing.

Multimode interference (MMI) waveguides can be used to create wavelength-dependent spot patterns which enables simultaneous analyte detection on a single optofluidic chip, useful for disease diagnostics. The fidelity of such multi-spot patterns is important for high sensitivity and accurate target identification. Buried rib structures have been incorporated into these SiO2-based waveguides to improve environmental stability. Through experiments and simulation, this letter explores design parameters for a buried MMI rib waveguide based on anti-resonant reflecting optical waveguides in order to produce high-fidelity spot patterns. Optimal rib heights and widths are reported in the context of available microfabrication etch technology and performance for an optimized biosensor is shown.

Optofluidic Lab-on-a-Chip Fluorescence Sensor Using Integrated Buried ARROW (bARROW) Waveguides.

Optofluidic, lab-on-a-chip fluorescence sensors were fabricated using buried anti-resonant reflecting optical waveguides (bARROWs). The bARROWs are impervious to the negative water absorption effects that typically occur in waveguides made using hygroscopic, plasma-enhanced chemical vapor deposition (PECVD) oxides. These sensors were used to detect fluorescent microbeads and had an average signal-to-noise ratio (SNR) that was 81.3% higher than that of single-oxide ARROW fluorescence sensors. While the single-oxide ARROW sensors were annealed at 300 °C to drive moisture out of the waveguides, the bARROW sensors required no annealing process to obtain a high SNR.