Optical dyeing of sperm cells in sexual assault victim samples using a variable multi-spectral LED array

 Supervised by- Professor Alex Mariakakis and Professor Aaron Wheeler, University of Toronto

Abstract

The forensic analysis of sexual assault samples remains a lengthy and challenging process, significantly delaying justice for victims. Current methods using chemical dyes to detect sperm cells are time-consuming, degrade biological samples, and complicate subsequent DNA analysis. This research aims to develop a novel optical virtual staining method utilizing a cost-effective, multi-spectral LED array to rapidly and non-destructively enhance sperm cell visibility without chemical dyes. The proposed approach leverages hyperspectral imaging techniques using a conventional RGB camera paired with a custom-built PCB array of 31 narrow-band LEDs, each emitting specific wavelengths across the visible spectrum.

The research will unfold in three distinct stages: First, a custom PCB equipped with sequentially firing LEDs will be designed and fabricated, emphasizing sustainability and thermal management. Second, computational modelling and machine learning algorithms will optimize the illumination patterns to maximize visual contrast between sperm and surrounding cells. Third, rigorous validation involving visual inspection and software-based image analysis will assess the system's accuracy and reliability compared to traditional staining methods.

This low-cost, user-friendly innovation aims to reduce processing times, preserve sample integrity, and increase forensic accessibility. By minimizing human intervention and enhancing analysis precision, this method has the potential to significantly ease the emotional and procedural burdens on victims, offering them quicker and more accurate forensic outcomes.

Introduction

Every year, over 30,000 cases of sexual assault are registered in Canada alone [1]. Seeking justice as a victim of sexual assault is a protracted and emotionally taxing process, as a typical sexual assault investigation in Canada can take anywhere from 2 months to 2 years [2]. One of the technological challenges in this process is the slow, manual forensic analysis of biological samples.

This process is typically performed using a phase contrast or optical darkfield microscope along with chemical dyes (e.g., DAPI and Alexa488) that stain samples to highlight sperm cells. However, these dyes can degrade the samples over time, often requiring multiple samples to be prepared and tested if the first does not produce a conclusive result. Further, the dyes can significantly reduce the accuracy of any DNA testing done on the sample [3]. This manual process of staining and verifying the presence of sperm cells is time-consuming and error-prone, thus delaying decision-making and heightening the emotional burden on victims.

Background

Prior research into this problem has primarily followed two approaches, each with its own limitations. One prevalent approach involves Fourier ptychographic microscopy (FPM), which utilizes single-colour light-emitting diodes (LEDs) arranged at varied angles to illuminate the sample. This approach uses refracted light and computational reconstruction to generate high-resolution images [4], achieving excellent clarity in thin samples and enabling rapid and accurate verification of sperm presence. Despite the clarity offered by FPM in idealized conditions, it faces significant limitations when applied to real-world forensic samples. The thick, heterogeneous nature of vaginal cell samples often obscures the visualization of spermatozoa, necessitating the additional use of chemical stains or dyes to effectively differentiate sperm cells from surrounding tissue and debris. This reliance on staining introduces extra preparation steps, increasing complexity, processing time, and potentially compromising the integrity of sensitive biological samples.

An alternative emerging approach uses hyperspectral imaging microscopy, which relies on capturing and analyzing the unique spectral signatures of biological materials [5]. It enables precise identification and differentiation of spermatozoa from vaginal epithelial cells and other cellular components without requiring staining, greatly enhancing forensic efficiency. Despite these advantages, hyperspectral imaging microscopy is only available in forensic laboratories since they are bulky and cost thousands of dollars. Additionally, the sophisticated nature of hyperspectral systems necessitates extensive training and technical proficiency among forensic analysts, thereby posing significant barriers to rapid adoption and integration into existing forensic workflows, particularly in resource-limited or smaller laboratories in disadvantaged regions.

Thus, while these two methodologies have significant advantages over traditional optical microscopy, both still face critical practical limitations—ranging from preparation complexity and applicability constraints with FPM to infrastructure requirements and analyst training challenges inherent in hyperspectral imaging microscopy.

Research Objectives & Questions

I propose a hybrid approach that uses hyperspectral imaging to both visually and digitally enhance biological samples to ease the burden of analysts; this process can be considered a form of “virtual staining”. This system will use a series of LEDs on a custom printed circuit board (PCB) to create optical contrast between sperm cells and other organic matter. Capturing the response with a camera-enabled microscope or a smartphone with a microscopy attachment will allow for visual detection of sperm without any of the cell degradation.

Primary Objectives

• Develop a high-speed panel with multiple narrow-band, high-power LEDs to simulate dyeing cells with light
• Classify visual data for samples collected at various times post-incident to analyse spectral signatures of degenerated sperm

Secondary Objectives

• Utilize sustainable and low-cost materials to ensure wide accessibility and adoption
• Ensure the process is user-friendly, intuitive and requires minimal training costs

Methodology

The proposed research can be divided into three main stages.

This initial stage will involve designing and fabricating a custom PCB incorporating a precisely controlled array of 31 LEDs. Each LED will emit a specific wavelength of visible light, spaced at 10-nanometer intervals, covering a broad and precise spectrum. The array will be designed to illuminate biological samples with LEDs firing sequentially at rapid intervals, allowing for custom variable spectral illumination. Careful consideration will be given to thermal management, precise wavelength calibration, and electronic control mechanisms to ensure stability and repeatability of illumination. The array will also need to be easily integrated into either a camera-enabled microscope or a smartphone with a microscopy attachment.

In the second stage, the focus will shift to optimizing the illumination patterns produced by the LED array. The goal will be to identify specific illumination sequences and intensities that maximize the optical contrast between sperm cells and surrounding biological material such as vaginal cells or other organic debris. Computational modelling, combined with preliminary empirical tests, will guide this optimization. Various algorithms and machine learning techniques will be employed to systematically evaluate and select the optimal illumination configuration.

The final stage involves rigorous testing and validation of the developed multispectral imaging system. Initially, visual inspections will be conducted to qualitatively assess the system’s performance in distinguishing sperm cells clearly from other biological components. Following visual evaluation, a comprehensive software-based quantitative assessment will be performed. This will involve image processing algorithms and statistical analysis to measure the accuracy, reliability, and reproducibility of the system. Results will be compared against traditional dye-based microscopy methods to establish the efficacy and potential advantages of the proposed technology.

Below is a detailed breakdown of my proposed timeline:

Phase 1: Developing the Printed Circuit Board (Week 1 – 2)

1. Design and fabricate a custom PCB
   1. Design a high-speed LED panel with 31 narrowband LEDs and gallium-nitride transistors
   2. Ensure appropriate heat dissipation by using thermal vias and heat sinks
   3. Fabricate the PCB base using recycled fibre-glass boards instead of the conventional FR4
2. Soldering components and preparing the final prototype
   1. Use reflow-soldering to attach the narrowband LEDs and transistors onto the board
   2. Connect the PCB with a high-speed chipset like Raspberry Pi Pico to control the LEDs
3. Test and debug the prototype
   1. Integrate the PCB with an existing microscope setup
   2. Use an oscilloscope and photo-sensitive diode to measure LED brightness and response times

Phase 2: Develop an Optimized Illumination Pattern (Week 3 – 5)

1. Modelling spectral reflectance of samples
   1. Procure samples from the Ontario Centre of Forensic Sciences to best model real forensic cases
   2. Compute the spectral reflectance of sperm against surrounding organic matter using a spectroscope at various times after an assault (1-10 days) to model accuracy at various stages of natural cell degradation
2. Illumination optimization algorithm for the LEDs to optically dye sperm
   1. Utilize an optimization algorithm to simulate contrast and converge on an optimal illumination pattern according to the CIE XYZ colour space

Phase 3: Testing, Optimization and Reporting (Week 5 - 6)

1. Final testing and optimization
   1. Recruit microscopy users to confirm the illumination works as well as traditional staining in terms of system usability and visual contrast
   2. Computationally analyse images captured to confirm high contrast
2. Tabulation of results and reporting
   1. Write a detailed and user-friendly instructional manual for using the device
   2. Prepare a research report for publication

Potential Impact

The project aims to deliver a rapid, cost-effective, and reliable alternative to existing forensic testing methods for sexual assault cases. By introducing a novel optical dyeing approach using narrowband LEDs, this research could significantly lower barriers to justice. Ultimately, this technology has the potential to revolutionize automated testing, offering a viable and more accessible solution compared to expensive hyperspectral sensors. Lower human intervention and higher accuracy would preserve discretion and help mitigate stigma, which would hopefully provide victims with the swift closure they deserve.

References

1. Government of Canada, Statistics (2024, July 25). Incident-based crime statistics, by detailed violations, Canada, provinces, territories, census metropolitan areas and Canadian Forces Military Police. Incident-based crime statistics, by detailed violations, Canada, provinces, territories, Census Metropolitan Areas and Canadian Forces Military Police. https://doi.org/10.25318/3510017701-eng
2. Vilkhov, I. (2023, September 15). How the sexual assault investigation process works in Canada. Vilkhov Law. https://vilkhovlaw.ca/how-the-sexual-assault-investigation-process- works-in-canada/
3. Moreira, B. G., You, Y., Behlke, M. A., & Owczarzy, R. (2005). Effects of fluorescent dyes, quenchers, and dangling ends on DNA duplex Biochemical and Biophysical Research Communications, 327(2), 473–484. https://doi.org/10.1016/j.bbrc.2004.12.035
4. Tian, L., Li, X., Ramchandran, K., & Waller, L. (2014). Multiplexed coded illumination for Fourier ptychography with an LED array Biomedical Optics Express, 5(7), 2376. https://doi.org/10.1364/boe.5.002376
5. Pradeep, A. S., Babu, J., Sudaroli Sandana, J., & Deivalakshmi, S. (2024). Innovations in forensic science: Comprehensive review of hyperspectral imaging for bodily fluid. Forensic Science International, 364, 112227. https://doi.org/10.1016/j.forsciint.2024.112227
6. Dhruv Verma, Ian Ruffolo, David B. Lindell, Kiriakos N. Kutulakos, and Alex Mariakakis. 2024. ChromaFlash: Snapshot Hyperspectral Imaging Using Rolling Shutter Cameras. ACM Interact. Mob. Wearable Ubiquitous Technol. 8, 3, Article 132 (September 2024), 31 pages. https://doi.org/10.1145/3678582
7. Elsayed, , Bodo, L., Gaoiran, C., Keuhnelian, P., Dosajh, A., Luk, V., Schwandt, M., French, J. L., Ghosh, A., Erickson, B., Charlesworth, A. G., Millman, J., & Wheeler, A. R. (2024). Toward analysis at the point of need: A digital microfluidic approach to processing multi‐ source sexual assault samples. Advanced Science, 11(41). https://doi.org/10.1002/advs.202405712