Particle Classification through the Analysis of the Forward Scattered Signal in Optical Tweezers

Abstract

The ability to select, isolate, and manipulate micron-sized particles or small clusters has made optical tweezers one of the emergent tools for modern biotechnology. In conventional setups, the classification of the trapped specimen is usually achieved through the acquired image, the scattered signal, or additional information such as Raman spectroscopy. In this work, we propose a solution that uses the temporal data signal from the scattering process of the trapping laser, acquired with a quadrant photodetector. Our methodology rests on a pre-processing strategy that combines Fourier transform and principal component analysis to reduce the dimension of the data and perform relevant feature extraction. Testing a wide range of standard machine learning algorithms, it is shown that this methodology allows achieving accuracy performances around 90%, validating the concept of using the temporal dynamics of the scattering signal for the classification task. Achieved with 500 millisecond signals and leveraging on methods of low computational footprint, the results presented pave the way for the deployment of alternative and faster classification methodologies in optical trapping technologies.

Autonomous and intelligent optical tweezers for improving the reliability and throughput of single particle analysis

“…For this, the forward scattered signal of the trapping beam is collimated with a 10x condenser lens before being steered into a quadrant photodetector (PDQ80A-Thorlabs), using a relay lens to image the back focal plane of the condenser. In specific, for each particle, we obtain a set of three signals (X, Y, and SUM, see figure 1(C)) which can be related to the positions of the particle upon calibration [5]. Nevertheless, even without the calibration factor, the signals already contain information that relates to the physical properties of each individual particle as discussed before, namely size and refractive index.…”

“…OT systems can also be adapted for indirect monitoring of physical and chemical processes. One such method involves studying the scattered radiation from the optically trapped specimen [4], which has motivated the development of identification and classification tools based on optical trapping [4,5]. On the other hand, these setups have also been combined with different probing techniques for a more comprehensive study of the trapped particles.…”

“…One of the most important factors is the reproducibility of the results, which may vary due to external conditions or operating procedures. For example, laser power or temperature fluctuations of the sample due to laser heating can strongly affect the characteristics of the scattered signal, which means that an automatic and controlled procedure is a mandatory requirement in this kind of experiment [5]. From a distinct perspective, the time duration of the procedure shall also be taken into consideration, as most of the applications may require the analysis of hundreds to thousands of particles for statistical significance, thus requiring faster acquisition procedures.…”

Automation strategies and machine learning algorithms towards real-time identification of optically trapped particles

“…Being sensitive to the displacement of the trapped particle in the transverse directions, the acquired signal from the quadrant photodetector allows to probe the Brownian motion of the trapped beads which contains information regarding the physical properties of the particle [1]. In the recent years, our team explored a few strategies for single particle/cell classification using optical trapping systems and scattered radiation, both using the back-scattered [2] and forward scattered signal [3]. Focusing on the latter for the purpose of this work, we reported the classification of trapped particles by introducing an algorithm that consists on performing a dimensional reduction with Principal Component Analysis performed in the Fourier space before the application of common classification strategies such as K-nearest neighbors.…”

Intelligent Optical Tweezers with deep neural network classifiers

“…Furthermore, when integrated with a quadrant-photodetector, see figure 1, OT can also probe the dynamical properties of a trapped particle by tracking the position using the forward scattered light. Recently, it has been shown that these dynamics can be used to identify and classify particles through a careful analysis of the position timeseries and machine learning algorithms, which can be fundamental for future integration into intelligent microfluidic devices [3].…”