Artificial Neural Networks for GMR-Based Magnetic Cytometry

Abstract

In this work, we propose an artificial neural network (ANN) for magnetic microcytometry pattern recognition and automated counting. The method is tested for detecting analytes in the 2–3- <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mu \text{m}$ </tex-math></inline-formula> range. The cytometer is composed of a disposable cartridge and an acquisition platform. The disposable cartridge contains microfluidic channels with <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$10\times 100\,\,\mu \text{m}\,\,^{\mathrm{ 2}}$ </tex-math></inline-formula> cross section on top of a substrate with magnetoresistive (MR) sensors. The custom analog signal chain performs with an integrated noise of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$2.99~\mu \text{V}$ </tex-math></inline-formula> textsubscript rms in a 10-kHz bandwidth. To employ the ANN, we synthesize a training dataset based on the magnetic-dipole equation and several dataset expansion methods. The ANN is tested on an experiment with 2.8- <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mu \text{m}$ </tex-math></inline-formula> magnetic particles (MPs) and compared with an improved threshold-based method with reduced false positives. The ANN produces a maximum of 90% detection rate, improving on the 30%–50% detection rates of other single-sensor methods published in the literature.