Detecting cells rotations for increasing the robustness of cell sizing by impedance measurements, with or without machine learning

Abstract

The Coulter principle is a widespread technique for sizing red blood cells (RBCs) in hematological analyzers. It is based on the monitoring of the electrical perturbations generated by cells passing through a micro‐orifice, in which a concentrated electrical field is imposed by two electrodes. However, artifacts associated with near‐wall passages in the sensing region are known to skew the statistics for RBCs sizing. This study presents numerical results that emphasize the link between the cell flow‐induced rotation in the detection area and the error in its measured volume. Based on these observations, two methods are developed to identify and reject pulses impaired by cell rotation. In the first strategy, the filtering is allowed by a metric computed directly from the waveform. In the second, a numerical database is employed to train a neural network capable of detecting if the cell has experienced a rotation, given its electrical pulse. Detecting and rejecting rotation‐associated pulses are shown to provide results comparable to hydrodynamical focusing, which enforces cells to flow in the center of the orifice, the gold standard implementation of the Coulter principle.

Red blood cell rheology during a complete blood count: A proof of concept

“…The blood cells having a conductivity significantly lower than the suspending medium, their passage through the micro-aperture results in an increase of resistivity, which in turn leads to a tension pulse measurable at the electrodes, as Fig 1 illustrates. As simultaneous passages are rare due to the large dilution rate employed, the pulses count directly leads the cell count in the samples, while for each pulse, the amplitude ΔU is assumed to be proportional to the cell volume V p [43]:…”

“…This is not the case for near-wall trajectories of RBCs. First, the electrical field is known to be highly heterogeneous near the orifice edges [42,46,47], as illustrated in Fig 1 . Besides, next to the aperture walls, high shear stresses result in rotations [42][43][44] and deformations [42] of RBCs. These dynamical edge-effects lead to variations of f s while the RBC crosses the detection area, thus invalidating the linearity assumption between ΔU and V p .…”

“…These dynamical edge-effects lead to variations of f s while the RBC crosses the detection area, thus invalidating the linearity assumption between ΔU and V p . This explains why many efforts have been made for detecting and rejecting pulses impaired by the edge-effects [43,45,48], or for designing Coulter counters with hydrofocusing [49] to prevent RBC from passing next to the orifice wall.…”

“…On the other hand, simulating the RBCs dynamics in the entire domain would induce prohibitive computational costs. Therefore, a specific series of simulations, described in Fig 2, is required for handling the calculation of the RBC dynamics [42,43]. Details and validation of this simulation pipeline have been presented by Taraconat et al [42].…”

“…To do so, blood is first diluted, then flows through a sensor that involves a cylindrical micro-orifice (or aperture) and two electrodes located across the orifice (see Fig 1). The micro-orifices are typically 75 μm long and 50 μm in diameter [40][41][42][43][44]. The electrodes, powered with a constant intensity current, generate an electrical field in the flowing suspension.…”