Neural network ensemble model for prediction of erythrocyte sedimentation rate (ESR) using partial least squares regression.

The erythrocyte sedimentation rate (ESR) is a non-specific blood test for determining inflammatory conditions. However, the long measurement time (60 min) to obtain ESR is an obstacle for a prompt evaluation. In this study, to reduce the measurement time of ESR, deep neural networks (DNNs) were applied to the sedimentation tendency of blood samples. DNNs using multilayer perceptron (MLP), long short-term memory (LSTM), and gated recurrent unit (GRU) were assessed and compared to determine a suitable length of time for the input sequence. To avoid overfitting, a stacking ensemble learning was adopted, which combines multiple models by using a meta model. Four meta models were compared: mean, median, least absolute shrinkage and selection operator, and partial least squares regression (PLSR) schemes. From the empirical results, LSTM and GRU models have better prediction than MLP over sequence lengths of 5 to 20 min. The decrease in [Formula: see text] and [Formula: see text] of GRU and LSTM was attenuated after a sequence length of 15 min, so the input sequence length is determined as 15 min. In terms of the meta model, the statistical comparison suggests that GRU combined with PLSR (GRU-PLSR) is the best case. Then, the GRU-PLSR was tested for prediction of ESR data obtained from periodontitis patients to check its applicability to a specific disease. The Bland-Altman plot shows acceptable agreement between measured and predicted ESR values. Based on the results, the GRU-PLSR can predict ESR with improved performance within 15 min and has potential applicability to ESR data with inflammatory and non-inflammatory conditions.

Quantitative study for control of air-liquid segmented flow in a 3D-printed chip using a vacuum-driven system.

The formation of droplets or bubbles in a microfluidic system is a significant topic requiring device miniaturization and a small volume of samples. Especially, a two-phase segmented flow can be applied to micro-mixing for chemical reactions and the treatment of heat and mass transfer. In this study, a flow of liquid slugs and bubbles was generated in a 3D-printed chip and controlled by a single pump creating a vacuum at the outlet. The pump and chip device were integrated to form a simple and portable system. The size and flow rate of liquid slugs, obtained through image processing techniques, were analyzed considering several parameters related to hydraulic resistance and pressure drop. In addition, the effect of segmentation on mixing was observed by measuring the intensity change using two different colored inks. The hydraulic resistance of air and liquid flows can be controlled by changing the tube length of air flow and the viscosity of liquid flow. Because the total pressure drop along the channel was produced using a single pump at the outlet of the channel, the size and flow rate of the liquid slugs showed a near linear relation depending on the hydraulic resistances. In contrast, as the total pressure varied with the flow rate of the pump, the size of the liquid slugs showed a nonlinear trend. This indicates that the frequency of the liquid slug formation induced by the squeezed bubble may be affected by several forces during the development of the liquid slugs and bubbles. In addition, each volume of liquid slug segmented by the air is within the range of 10-1 to 2 µL for this microfluidic system. The segmentation contributes to mixing efficiency based on the increased homogeneity factor of liquid. This study provides a new insight to better understand the liquid slug or droplet formation and predict the segmented flow based on the relationship between the resistance, flow rate, and pressure drop.

Magnetically actuating implantable pump for the on-demand and needle-free administration of human growth hormone.

A bolus of human growth hormone (hGH) is often prescribed for the treatment of growth hormone deficiency, which requires frequent injections in current clinical settings. This painful needle-involved delivery often results in poor patient compliance, leading to low medication adherence and poor clinical outcomes. Therefore, we propose a magnetically actuating implantable pump (MAP) that can infuse an accurate dose of hGH only at the time of non-invasive magnet application from the skin. The MAP herein could reproducibly infuse 20.6 ±â€¯0.9 µg hGH per actuation without any leak at times without actuation. The infused amount increased proportionally with an increase in the number of actuations. When the MAP was implanted and actuated with a magnet in animals with growth hormone deficiency for 21 days, the profiles of plasma hGH concentration and insulin-like growth factor (IGF)-1, as well as changes in body weight, were similar to those observed in animals treated with conventional subcutaneous hGH injections. Therefore, we anticipate that the MAP fabricated in this study can be a non-invasive alternative to administer hGH without repeated and frequent needle injections.

Electrodynamic Force Derived in-Channel Separation and Detection of Au Nanoparticles Using an Electrochemical AC Microfluidic Channel.

In this study, we have established the separation of Au nanoparticles (AuNPs) using a symmetrical AC electric field applied-electrochemical microfluidic device composed of carbon channel and detection electrodes. The lateral movement of AuNPs in the channel under the AC field was analyzed by simulation using the mathematically derived equations, which were formulated from Newtonian fluid mechanics. It shows that the nanoparticles are precisely separated according to their respective mass or size difference in a short time. The experimental parameters affecting the separation and detection of AuNPs were optimized in terms of applied frequency, amplitude, flow rate, buffer concentration, pH dependency, and temperature. The final separation was performed at 1.0 V amplitude with 8.0 MHz frequency at 0.4 µL/min flow rate for the separation, and the potential of 1.0 V was applied for the amperometric detection of AuNPs in a 0.1 M PBS. Before and after the separation, AuNPs (diameter range: 3-60 nm) were confirmed by UV-visible spectroscopy and transmission electron microscopy. In this case, the separation resolution was 3 nm with an enhanced separation efficiency of up to 597,503 plates/m for the AuNPs. In addition, the amperometric current response of the detection electrode under the AC field application was also enhanced by the sensitivity 5-fold compared with the absence of the AC field.

3D printed microfluidic viscometer based on the co-flowing stream.

Considering the role of viscosity in the dynamics of physical, chemical, and biological systems, accurate measurement of viscosity is essential. Although many conventional viscometers have been widely used, these conventional viscometers suffer from some drawbacks. In this study, a three-dimensional (3D) printed microfluidic viscometer was proposed based on the estimation of the pressure between two fluids to easily measure viscosity with small samples. The 3D printed microfluidic viscometer can be fabricated through amine-epoxy bonding on 3D printed blocks. By separately delivering samples and reference fluids into two inlets, an interfacial line could be induced. Based on the relation between the pressure ratio and the width of the reference flow, the viscosity (µ) of the sample can be estimated by measuring the relative width of the reference flow. The relation between the pressure and interfacial width between test samples and reference flows in the 3D printed microfluidic viscometers was analyzed by experiment and simulation to determine the effects of the mesh-like pattern of the 3D printed viscometers on the pressure estimation. To validate the proposed method, the viscosity values of glycerol mixtures measured by the 3D printed viscometer were compared with those measured by a conventional viscometer. As an application of the 3D printed viscometer, the viscosity curves for blood samples collected from diabetic and non-diabetic patients depending on their shear rates were compared. As expected, a high blood viscosity in the diabetic group was observed. Based on the experimental demonstrations, the 3D printed viscometer has strong potential to develop portable viscometers that can be translated to commercial outcomes.

Variations in pulsatile flow around stenosed microchannel depending on viscosity.

In studying blood flow in the vessels, the characteristics of non-Newtonian fluid are important, considering the role of viscosity in rheology. Stenosis, which is an abnormal narrowing of the vessel, has an influence on flow behavior. Therefore, analysis of blood flow in stenosed vessels is essential. However, most of them exist as simulation outcomes. In this study, non-Newtonian fluid was observed in stenosed microchannels under the pulsatile flow condition. A polydimethylsiloxane channel with 60% stenosis was fabricated by combining an optic fiber and a petri dish, resembling a mold. Three types of samples were prepared by changing the concentrations of xanthan gum, which induces a shear thinning effect (phosphate buffered saline (PBS) solution as the Newtonian fluid and two non-Newtonian fluids mimicking normal blood and highly viscous blood analog). The viscosity of the samples was measured using a Y-shaped microfluidic viscometer. Thereafter, velocity profiles were analyzed under the pulsatile flow condition using the micro-particle image velocimetry (PIV) method. For the Newtonian fluid, the streamline was skewed more to the wall of the channel. The velocity profile of the non-Newtonian fluid was generally blunter than that of the Newtonian fluid. A highly oscillating wall shear stress (WSS) during the pulsatile phase may be attributed to such a bluntness of flow under the same wall shear rate condition with the Newtonian fluid. In addition, a highly viscous flow contributes to the variation in the WSS after passing through the stenosed structures. A similar tendency was observed in simulation results. Such a variation in the WSS was associated with plaque instability or rupture and damage of the tissue layer. These results, related to the influence on the damage to the endothelium or stenotic lesion, may help clinicians understand relevant mechanisms.

Characteristics of pulsatile flows in curved stenosed channels.

Spatial and temporal variations of the hemodynamic features occur under pulsatile conditions in complex vessel geometry. Wall shear stress affected by the disturbed flow can result in endothelial cell dysfunction, which leads to atherogenesis and thrombosis. Therefore, detailed understanding of the hemodynamic characteristics in a curved stenosed channel is highly important when examining the pathological effects of hemodynamic phenomena on the progression of atherosclerosis. The present study measures the velocity fields of pulsatile flows with three different Reynolds numbers in 3D curved vessel models with stenosis using time-resolved particle image velocimetry (PIV). Three different models were cast in PDMS polymer using models made by a 3D printer with different bend angles of 0°, 10°, and 20° between the longitudinal axes at the upstream and downstream of the stenosis. To investigate the 3D flow structures, a stack of 2D velocity fields was obtained by adjusting the position of the laser sheet along the Z-direction. The structures of flow fields in the stenosed models were analyzed using the distribution of the shearing strain as well as the skewness and full width at half maximum of the velocity profile. To support experiment results, distributions of pressure and 3D vortex in the curved stenosed channels were estimated by conducting the numerical simulation. These results indicate that the curvature of the tube considerably influences the skewness of the flow, and the shear stress is intensified near the outer curvature wall due to centrifugal force. The results would be helpful in understanding the effects of geometrical factors on plaque rupture and severe cardiovascular diseases.