When
Monday, April 6, 2026, at 12:00 p.m.
Location: Keating 103 | Zoom link
Hosts: Swarna Ganesh and Kellen Chen
Jocelyn Reynolds
PhD Candidate, Yoon lab
"Smartphone-Based Autofluorescence Imaging and Flow Velocity Analysis for Bacterial Classification Using Machine Learning"
Abstract: Microbiome composition plays a critical role in human health, yet current methods for analyzing microbial communities rely on laboratory-based techniques that are often slow and resource-intensive. These limitations restrict their use in rapid or point-of-care applications. In particular, bioreceptor-based and sequencing approaches often require prior knowledge, specialized instrumentation, and multi-step processing, which can limit their use for rapid analysis of complex bacterial mixtures. As an alternative, this body of work investigates approaches that use non-specific reagents and natural bacterial responses, including autofluorescence and flow behavior, as measurable signals. By leveraging smartphone-based multispectral autofluorescence imaging and microfluidic flow assays, bacterial samples can be characterized without the need for target-specific labeling or complex preparation. These platforms measure optical signals of bacteria or flow behavior changes due to viscosity and surface tension arising from bacterial interactions. These signals are combined with machine learning models to identify patterns associated with different bacterial compositions to evaluate performance in mixed populations. This work explores the detection of pathogenic bacteria within mixed samples using convolutional neural networks applied to autofluorescence images with high sensitivity, the use of flow velocity signatures to accurately classify bacterial species within the same genus and mixtures relevant to skin microbiome analysis, and the extension of flow-based profiling toward gut microbiome applications geared towards colorectal cancer monitoring.
Rafael Romero
PhD Candidate, Kang lab
"High-Resolution Sidestream Dark Field Microscopy for Visualizing Individual Melanocytic Cells in 3D"
Abstract: Dermatoscopy is the current gold standard for the noninvasive evaluation of suspicious pigmented lesions. Dermatoscopes provide 10x magnification of surface-level patterns, structures and colors to inform a diagnosis. However, diagnostic accuracy varies widely among physicians, leading to high rates of unnecessary biopsies.
We present a High Resolution Sidestream Darkfield Microscope (HR-SDF) for evaluating pigmented melanocytic lesions in vivo. This method uses oblique illumination and multiply scattered photons, acting as back illumination within the tissue, to enhance the contrast of individual pigmented cells. This approach enables visualization of the pagetoid distribution (upward spread of melanocytic cells), a histological diagnostic feature of melanoma. We achieved a lateral resolution of 1.4 µm over a 300 µm scanning depth using a high-numerical-aperture water-immersion objective lens. We demonstrated HR-SDF capability in a clinical study, capturing volumetric en face image stacks of melanocytic lesions. These image stacks enable visualization of melanin and vasculature features at specific depths, revealing the spatial distribution of melanocytic cells.
In summary, HR-SDF provides three-dimensional imaging with a large field of view for the evaluation of pigmented lesions. In the future, this technology may enable quantitative analysis of melanocytic cellular morphology and depth-dependent features, thereby improving diagnosis and reducing unnecessary biopsies.