detecting RBCs rotation with impedance based systems

The Coulter principle is a widespread technique for the sizing of red blood cells in hematological analyzers. It is based on the monitoring of the electrical perturbations generated by the cells when they go through a polarized aperture. However, artifacts associated with near-wall passages in the sensing region are known to skew the statistics for red blood cells sizing. Numerical results presented in the manuscript show how RBCs deform and rotate in the sensing region, provided they flow near the walls of the measurement orifice. Such dynamics generate spurious electrical pulses responsible for the overestimation of RBC volumes. We propose two methods for detecting and rejecting electrical pulses spoiled by rotation in order to provide volume assessments more accurate. The experiments presented in this work assess the accuracy of the aforementioned pulse-editing methods, by comparison with hydrodynamical focusing, the gold standard implementation of Coulter principle which enforces cells to flow in the center of the orifice.
Conclusion:
The filtering methods are shown to retrieve symmetrical volume distributions of red blood cells, in agreement with results from a hydrofocused system. Moreover, good correlations are obtained with the hydrofocused system in terms of mean corpuscular volume (MCV) and RBCs distribution width (RDW).
Notes:
Four sets of experimental results are presented: 1) Results from a Coulter counter equipped with hydrodynamical focusing 2) Results from a classical Coulter counter 3) Results From a classical Coulter counter when applying a first pulse editing method, based on the calculation of a metric from the electrical pulse. 4) Results from a classical Coulter counter when applying a second pulse editing method, allowed by a Neural Network. Configurations (1), (2), (3) and (4) are referred as HF, noHF, noHF-R and noHF-NN in the manuscript, respectively. Results from configurations (2), (3) and (4) are all based on the same set of data. Configurations (3) and (4) consist in rejecting some of the data of configuration (2) for calculating the relevant statistics of interest in Complete Blood Counts. Files ID_HF_RR.fcs are obtained from the system equipped with hydrodynamical focusing, while files ID_noHF_RR.fcs arise from the system without hydrodynamical focusing. "ID" (in ID_HF_RR.fcs and. ID_noHF_RR.fcs ) refers to the sample, and "RR" is 1 or 2, since acquisitions are performed in duplicate for each systems. Files ID_HF_RR.fcs are made of pulses amplitudes and ID_noHF_RR.fcs contain pulses amplitudes, widths ratios and the associated NN scores. The filters are based on widths ratios and NN scores. The numerical database used for training the neural network is provided in "numerical_database.csv". Calibration: Measurements (electrical pulses) are calibrated in such a way that MCVs of configurations (1) and (3) (see Sec. 1.3 of this checklist) corresponds to the MCV rendered by a P120 (HORIBA Medical). Note that configurations (2), (3) and (4) arise from the same system. Hence, the calibration coefficient calculated for (3) stands for (2) and (4). Check of senseless results: Among the samples presented in the manuscript, MCV and RDW are measured 3 times, in a hydofocused system (configuration 1) and a non-hydrofocused system (configurations 2, 3 and 4) and in a reference machine, a P120 (HORIBA Medical), taken here as a reference. All results from the different machines were consistent. Duplicate tests: Random tests on a small number of samples have shown that duplicated acquisitions provide identical results. Two methods are proposed for the removal of rotation-associated pulses and shown to lead to similar results. This agreement between both methods supports the starting assumption: RBCs flowing in the wall vicinity rotate and induce spurious electrical pulses that skew the volume assessments, in a statistical point of view.