A low-cost, open-source centrifuge adaptor for separating large volume clinical blood samples

Abstract

Blood plasma separation is a prerequisite in numerous biomedical assays involving low abundance plasma-borne biomarkers and thus is the fundamental step before many bioanalytical steps. High-capacity refrigerated centrifuges, which have the advantage of handling large volumes of blood samples, are widely utilized, but they are bulky, non-transportable, and prohibitively expensive for low-resource settings, with prices starting at $1,500. On the other hand, there are low-cost commercial and open-source micro-centrifuges available, but they are incapable of handling typical clinical amounts of blood samples (2-10mL). There is currently no low-cost CE marked centrifuge that can process large volumes of clinical blood samples on the market. As a solution, we customised the rotor of a commercially available low-cost micro-centrifuge (~$125) using 3D printing to enable centrifugation of large clinical blood samples in resource poor-settings. Our custom adaptor ($15) can hold two 9 mL S-Monovette tubes and achieve the same separation performance (yield, cell count, hemolysis, albumin levels) as the control benchtop refrigerated centrifuge, and even outperformed the control in platelet separation by at least four times. This low-cost open-source centrifugation system capable of processing clinical blood tubes could be valuable to low-resource settings where centrifugation is required immediately after blood withdrawal for further testing.

Fig 1

(A) Sample processing volume vs price of available commercial and open source academic microcentrifuges, (B) Illustration of the different sized tubes used with the available microcentrifuges along with S-Monovette tube used in this project, (C) Adaptor concept: C.i) Commercial SciSpin MINI Microfuge, model: SQ-6050 with its original rotors C.ii) CAD schematic of the designed three-part rotor adaptor C.iii) Final 3D printed rotor adaptor mounted on the commercial microcentrifuge base C.iv) Top cover removed from the adaptor, showing inside part of the rotor adaptor that holds two standard 9mL S-Monovette tubes.

Fig 2

(A) Schematic illustration of the tubes on a fixed-angle adaptor and the radial dimensions (B) The influential parameters on the applied RCF on the sample in the tube mounted on the adaptor.

Fig 3. Evolution of design and basic characterisation.

Fig 4

Influence of design shapes and sizes on speed (A) Schematic diagram showing an arbitrary shape design provides higher aerodynamic drag force than the truncated cone shape design by resisting much of the surrounding airflow. For simplicity, airflow is showed only in one direction instead of all sides. (B) Radial distance of all designs vs their rotational speed (C) Lateral surface area of the truncated cone-shaped designs vs their rotational speed. NB: Both the panel B and C share the same legend.

Fig 5

(A) Deflection measurement from the recorded video of conical Design C. A.i) Snapshot of one of the video recordings of Design C A.ii) Edge capturing from one of Design C videos using Canny, an edge detection operator in Python OpenCV module. Further details about the deflection measurements are available in (S1 Fig in S1 File) A.iii) Measured deflection of both acceleration and decelerating phase from the recorded video using python OpenCV module for Design C. All recordings are available from (S2 Fig in S1 File) (B) Critical speed measurement in Ansys workbench B.i) Meshing of Design C (Nodes: 6749, Elements: 39228) C B.ii) Mode shape during natural frequency B.iii) Simulated (model) vs experimental critical speed of each conical shape design showing a good agreement between them with low percentage difference (1–5%) (C) Measured deflection of each conical shape design showing all designs experiencing deflection from 4 to 7 mm (D) Duration of deflection higher than 1.5 mm was decreasing the lowering of critical speed except design E5 which experienced higher deflection for longer times.

Fig 6

(A) Images of the initial and post centrifugated (3,6,10 minutes) S-Monovette tubes showing that larger centrifugation times resulted in a higher volume of plasma separated from 9 mL sample. (B) Measured separation yield of different designs compared with the control centrifugation performance. Statistics: standard unpaired t-test between each column and the control. Design E2 and E5 were able to separate the same amount of plasma as the control at 6 and 10 minutes with no statistically significant difference.

Fig 7

Blood count at 3, 6 and 10 minutes (A) The red blood cell counts of pre (initial) and post (3,6, 10 minutes) separated blood show the integrity of the sample after centrifugation. The statistical indications relate to a non-paired t-test between the RBC counts between 0, 3, 6, and 10 minutes (B) The red blood cell (RBC) count on plasma after centrifugation shows RBC concentration with design E2 was significantly lower than the control in all periods. On the contrary, design E5, although performing well at 3 minutes was hampered by vibrations reversing the separation process. Here the statistics relate to a non-paired t-test between each of the designs and the control, for each time point (C) Platelet count (PLT) on separated plasma. Design E2 was able to separate almost 4 times higher platelet than the control within 10 minutes. Here the statistics relate to a non-paired t-test between each of the designs and the control for each time point (D) Relative RBC and Platelet counts after 10 minutes centrifugation from two different pools of blood. Here the statistics relate to a non-paired t-test between Pool 1 and Pool 2 for each experiment. No significant difference was observed between different pools of blood.

Fig 8

(A) Percentage of hemolysis in separated plasma from E0, E2 and control. (B) Albumin concentration in E0, E2 and control (C) Cell-free DNA levels in E0, E2, and control after single centrifugation or with additional second separate centrifugation.