Analytical

Methods
View Article Online
PAPER View Journal | View Issue
Published on 05 May 2023. Downloaded by Indian Institute of Technology Kharagpur on 8/19/2024 6:54:50 PM.

Lipidest: a lipid profile screening test under

Cite this: Anal. Methods, 2023, 15, 2427
extreme point of care settings using a portable
spinning disc and an office scanner†
Victor Pakira,a Rahul Agarwal, b
Subhamoy Chatterjee,c Arghya Mukherjee b

and Suman Chakraborty *b

The demand for lipid profile (the cholesterol and triglyceride elements in the blood) testing outside
resourced diagnostic centers is continuously increasing for personalized and community-based
healthcare to ensure timely disease screening and management; however, it is inevitably challenged by
several bottlenecks in the existing point of care technologies. These deficits include delicate sample pre-
processing steps and device complexity, which give rise to unfavourable cost propositions to safeguard
against compromised test accuracy. To circumvent these bottlenecks, herein, we introduce a new
diagnostic technology, ‘Lipidest’, that integrates a portable spinning disc, a spin box, and an office
scanner to reliably quantify the complete lipid panel from finger-prick blood. Our design facilitates the
direct miniature adaptation of the established gold standard procedures as against any indirect sensing
technologies that are otherwise common in point-of-care applications introduced commercially. The
test procedure harmoniously connects all the elements of sample-to-answer integration in a single
device, traversing the entire pipeline of the physical separation of plasma from the cellular components
of the whole blood, the automated mixing with the test reagents on the same platform in situ, and
office-scanner-adapted quantitative colorimetric analytics that eliminate any undesirable artefacts on
account of variabilities in the background illumination and camera specifications. The exclusive value of
eliminating sample preparation steps, including the rotationally actuated segregation of the specific
blood constituents without any cross-interference between them, their automated homogeneous
mixing with the respective test reagents, and the simultaneous, yet independent, quantitative readout
without specialized instrumentation, render the test user-friendly and deployable in resource-
constrained settings with a reasonably wide detection window. The extreme simplicity and modular
nature of the device further make it amenable to mass manufacturing without incurring unfavourable
Received 17th March 2023
Accepted 3rd May 2023
costs. Extensive validation with laboratory-benchmark gold standards provide acceptable accuracy and
indicates the value of the first-of-its-kind ultra-low-cost extreme-point-of-care test with a scientific
DOI: 10.1039/d3ay00412k
foundation akin to highly accurate laboratory-centric technologies for cardiovascular health monitoring
rsc.li/methods and beyond.

non-specically linked with several other non-cardiac ailments

1. Introduction such as depressive disorder,7 Smith-Lemli Opitz syndrome,8
The globally increasing instances of cardiovascular diseases diabetes mellitus,9 acute pancreatitis,10 cerebral dysfunction,11
suggest the increasing need for lipid prole testing for early cystic brosis,12 and certain forms of cancers.13 While the lipid
diagnosis, as well as the continuous monitoring of the prole markers were previously perceived as matters of concern
vulnerable.1–6 The lipid prole markers are also known to be only for those who over-consume unhealthy fatty diets, their
implications in human health have by now been established to
be signicantly more far-reaching and they impact the low-to-
a
Advanced Technology Development Centre, Indian Institute of Technology Kharagpur, medium income groups of people as well. This is attributable
Kharagpur, 721302, India to increasing risk factors on account of a compromised holistic
b
Department of Mechanical Engineering, Indian Institute of Technology Kharagpur, lifestyle along with stressful living conditions. A lack of access
Kharagpur, 721302, India. E-mail: [email protected]; Tel: +91-3222-282990
c
to early diagnostics and patient management due to the non-
Department of Electronics and Electrical Communication Engineering, Indian
availability of centralized and resourced clinical settings and
Institute of Technology Kharagpur, Kharagpur, 721302, India
† Electronic supplementary information (ESI) available. See DOI:
trained healthcare professionals, has compounded the impact
https://doi.org/10.1039/d3ay00412k on human health to unbounded proportions, resulting in fatal

This journal is © The Royal Society of Chemistry 2023 Anal. Methods, 2023, 15, 2427–2440 | 2427
 View Article Online

Analytical Methods Paper

events such as massive cardiac arrest occurring, without any cross-reactivity between some test-specic reagents (such as
particular bias towards geographical location or age. cholesterol oxidase, ascorbic acid, hydrogen peroxide), blood
The World Heart Federation outlined a comprehensive plasma constituents (such as bilirubin), resulting in a potential
Cholesterol Roadmap,14 emphasizing the compelling barriers compromise in the test specicity as well as inevitable bottle-
against the implementation of appropriate combating necks in the sample preparation, cost, as well as compromised
measures in low-resource environments. Accordingly, the stability (for instance, instability against denaturation or
Published on 05 May 2023. Downloaded by Indian Institute of Technology Kharagpur on 8/19/2024 6:54:50 PM.

diagnostics and monitoring of the vulnerable population at the protease digestion) unless adapted with safeguarding process-
point of care (POC) in resource-limited settings emerged as inventiveness. Articial enzyme-mimics emerged as alternative
a high-priority consideration for public health management, viable options, resulting in the introduction of novel diagnostic
aligned with a focus target of the United Nations' Sustainable assays based on metal/metal oxide nanoparticles,26,27 nano-
Development Goals and WHO's Global Action Plan.15 This has materials,28 and biomolecules,29,30 having progressively
recently been emphasized as an “indicative example of increasing applications in detecting blood cholesterol levels
evidenced-based innovation(s)”, demanding sustained collec- and beyond.24,31–33 However, despite immense promises, these
tive effort for its mitigation. The POC diagnostics industry has assays have not been extensively deployed on a commercial
been prompt in responding to this need via various innovations basis, which is attributable to the constraints associated with
in lipid prole detection technology that offer some key the sample preparation and non-uniformity in the color
advantages as compared to traditional laboratory-centric rendering.34 These decits could be overcome by alternative
procedures. However, the concern of striking a balance technologies such as nanober-based bead-free elements for
between scientic rigor and accuracy, as compared to the immobilizing enzymes such as cholesterol oxidase and peroxi-
established gold standard benchmarks and evidenced func- dase.21 This approach, however, brought in other inaccuracies
tionality including viability in adapting to resource-limited stemming from the signal-generating system that requires
settings at the same time, continues to drive newer innova- a signicant amount of blood to be sucked into the test strip, as
tions in this diagnostic space.16 There is a progressively well as an operational inconvenience due to several minutes of
increasing need to arrive at technology disruptions that involve waiting time between successive steps.35 These challenges may
no separate sample preparation steps, non-interference with be overcome to some extent by using complementary metal
other blood constituents or signals, desirable stability and user- oxide semiconductor (CMOS)-based detection,36 including
friendliness of the test procedure outside of controlled labora- multi-modal analytical features such as integrated colorimetric,
tory settings, reasonably wide detection range and sample-to- chemiluminescence, surface plasmon resonance and hydrogen
answer turnaround time, as well as favourable features of ion measurements.37
device mass-manufacturing without any unwarranted cost Electrochemical sensing (either in the form of the ampero-
penalties. metric or voltammetric method) as a stand-alone testing tech-
Lipid prole testing traditionally includes the independent nology38 has also progressed over the past years for enzymatic
quantitative assessment of the levels of low-density lipoproteins analysis based on cholesterol oxidase,39 cholesterol dehydroge-
(LDL) and high-density lipoproteins (HDL) as the cholesterol nase,40 and CEH-cholesterol dehydrogenase among others,
elements, as well as triglycerides. Some classical methods for based on their rapid response, low cost, and portability, albeit
cholesterol testing, such as the Sperry-Webb, Zak-Henly and suffering from the constraints resulting from the interfering
Abell-Kendall methods, suffer from well-known shortcomings reducing elements in the blood or incomplete conversion to the
such as their interference with other blood analytes (commonly, nal product.41,42 The traditional high-voltage systems for lipid
hemoglobin and bilirubin) or non-cholesterol sterols and prole testing43 turned out to be prone to interferences from
cholesterol precursors or oxidation products that enforce the other electroactive analytes in the blood samples such as uric
use of environment-unfriendly corrosive reagents.17 Non- acid, ascorbic acid, bilirubin, pyruvate, and glutathione, among
enzymatic procedures, such as gas–liquid chromatography,18 others. The second-generation enzymatic biosensors use elec-
high-performance liquid chromatography (HPLC),19 and spec- tron mediators to overcome this bottleneck.44–51 Further
trophotometric analysis20 emerged to be accurate, albeit improvements in the assay performance may be achieved by
infrastructure-intensive, requiring elaborate and extensive depositing a thin lm on nanostructured screen-printed elec-
sample pre-treatment procedures. trodes.52,53 However, the efficacies of these more sophisticated
Over the past years, enzymatic assays have progressively variants of the detection procedure are commonly based on the
matured from their expensive and laboratory-centric early screen-printed electrode architecture, as well as the features of
formats to more stable and user-friendly variants that are the counter electrode for a three-electrode system,40 and the
adaptive for lipid prole testing at the POC.21 These methods hematocrit correction electrode for a four-electrode system,40
are typically combined with established detection technologies which are by no means trivial. Furthermore, issues such as
such as surface plasmon resonance (SPR),22 uorescence,23 enzyme immobilization, its uncontrolled distribution on a test
colorimetry,24 and electrochemical sensing,25 to result in strip and the sensitive dependence of the surface charge on the
a readout technology that interfaces with the bio-analytical solution pH54 continue to present challenges when imple-
assay. Despite such widespread emergence, the scientic mented without stringent laboratory control. Recently, dispos-
foundation of the enzymatic assays for the lipid prole (such as able detection systems emerged that were integrated with the
the peroxidase-based ones) continues to be challenged by the test strip as a single unied unit, which turned out to be

2428 | Anal. Methods, 2023, 15, 2427–2440 This journal is © The Royal Society of Chemistry 2023
 View Article Online

Paper Analytical Methods

effective in eradicating some of the above issues originating steps of separating plasma from the whole blood via centrifugal
from the lack of laboratory control, albeit resulting in adverse action, the subsequent automated homogeneous mixing with
cost penalties. A randomized controlled trial with 4968 patients the test reagents for stand-alone enzymatic reactions for each
in 53 general practices across Australia established some independent analyte detection without cross-interference, and
obvious merits of such POC tests55 but there was also a higher quantitative colorimetric analytics using standard office utili-
cost per test as compared to standard laboratory procedures, ties were synchronized with the action of a portable compact
Published on 05 May 2023. Downloaded by Indian Institute of Technology Kharagpur on 8/19/2024 6:54:50 PM.

opening up a dilemma as to whether this incremental cost is disc driven by a low-cost spin stand. This enables the cost-
justied in terms of the value proposition envisaged. effective, albeit direct, miniaturization of established
Offering a somewhat more favourable cost proposition than laboratory-based gold standard procedures, resulting in
the above enzymatic sensing technologies, the new-generation a dramatic reduction in the volume of the test reagents as used
colorimetric lipid prole test panels have progressed to be for laboratory-centric analysis. Our test procedure also features
signicantly more advanced in terms of their quantitative a unique adaptation of an office scanner as a high-quality
readout capabilities than ever before, thanks to the other optical detection system, which, unlike smartphone-enabled
parallel technology advancements in the digital world (such as imaging, is not prone to variabilities in the background elimi-
high-performance cameras, scanners, open-source soware and nation or imaging conditions that are inevitable in resource-
image analysis algorithms). However, challenges do remain in constrained settings.
establishing a self-contained modular design of complete We conducted an extensive review but did not come across
sample-to-answer portable diagnostic units by harnessing these any similar affordable technology that can integrate the auto-
advancements, which, at the same time, remain simple to mated sample preparation comprising the plasma extraction
manufacture and scalable, directly take up unprocessed and mixing capabilities of the compact disk technology and the
samples, and are extremely user-friendly to the extent that the robust colorimetric capabilities offered by a routine office
test may be executed by personnel with no specialized training. scanner for lipid proling. Another emphatic aspect of our
Here, we demonstrate the integration of a portable spinning technology is a direct miniature adaptation of the gold standard
disc with an adapted office scanner to obtain a POC-based analytical method. On many occasions, POC devices have
platform technology that can be applied in accurate lipid compromised performance because of adapting indirect
proling from a tiny volume of whole blood with minimal sensing methods that deviated from the gold standard princi-
intermediate human intervention, having the unique value ples, simply because the gold standard protocol appears to
proposition of incurring no cost-penalty on account of device compulsorily demand high-end laboratory-centric resources.
sophistication (Fig. 1). To realize this in effect, the essential From this perspective, implementing a laboratory gold standard

Fig. 1 Details of the spinning disc design. This is a 5-layer structure comprising alternate layers of PMMA (polymethylmethacrylate) and PSA
(pressure sensitive adhesive), a clear double-sided adhesive tape. The thicknesses of PMMA and PSA are 0.4 mm and 45 microns, respectively.
The outer diameter of the disc is 110 mm. R-1 refers to the region in which plasma and reagents are contained. R-2 refers to the region in which
the cellular components of blood are segregated after centrifugation. The details of the dimensions of the individual portions of the layers are
provided in the ESI (see Fig. S1).†

This journal is © The Royal Society of Chemistry 2023 Anal. Methods, 2023, 15, 2427–2440 | 2429
 View Article Online

Analytical Methods Paper

analytical method without laboratory control is by itself optimally designed to prevent the backow of the cells in the
a disruption to the envisaged POC technology, which makes it imaging eld.
clinically more favourable as compared to alternative non- Aer separation, the cells remain in chamber R2, whereas
specic sensing paradigms. the plasma remains in chamber R1. This channel maintains the
unidirectional entrapment of the cellular components in
2. Experimental section chamber R2 of the fourth layer. In the absence of such a junc-
Published on 05 May 2023. Downloaded by Indian Institute of Technology Kharagpur on 8/19/2024 6:54:50 PM.

tion, the separated red blood cells (RBCs) would otherwise ow
2.1. Fabrication of the test disc and its key features back and mix with the reagent substrate complex to form
The test disc is a composite structure made from transparent a diffuse red band at the contact point (as the reagents hemo-
plastic (PMMA) and double-sided pressure-sensitive adhesive lyze the RBCs), interfering with the intended colorimetric
(PSA) (see Fig. 1). The top layer (layer-1) contains the loading sensing.
and pressure release holes. It also acts as the top surface for the
microchannels. The next three layers (i.e., layers-2, 3 and 4)
contain the side walls of the microchannels. The bottom layer 2.2. Development of a portable detection system
(layer-5) provides structural support to the disc and acts as the The portable system comprises two major components, viz. A
base of the microchannels. Patterns were carved in the PMMA rotational drive and a modied office scanner.
and PSA with a vinyl cutter (Graphtec CE 6000-60 Plus) with 2.2.1. Rotational drive. The rotational drive, made in-
different settings of the cutting blade, pressure, and speed. All house, consists of a 12 V DC motor controlled by an
the layers were combined and then pressed together in a cold ATmega328 development board-based Arduino nano unit
roller press to seal against leakage. The integrated disc contains (Fig. 2b). The rotational box (Fig. 2a) includes three buttons.
three independent microchannels. These are utilized to The rst button S1 spins the disc at 2000 rpm (u1) for 5 minutes
measure the three components of the lipid prole, namely, for the extraction of plasma. The second button S2 spins the
high-density lipoproteins (HDL), low-density lipoproteins (LDL) disc in a spin prole sequence u2 (Fig. 2c) for 52 s. This spin
and triglycerides (TG). For ease of tracking, the microchannels prole is utilized for mixing. The third button S3 allows for
corresponding to the respective analytes are indexed by separate manual control of the spin to aid in the loading of the sample
barcodes throughout the data collection and processing stages. and reagents. The electronic components are enclosed in a 3D-
This marking need not be unique for a given device. Any order printed enclosure with a vertical motor alignment to minimize
of the test channels can be adopted locally and suitable changes the vibrations due to rotation. The whole setup has a power
may be reected in the barcode mapping of the test readouts consumption of 10.8 W and weighs about 470 g.
without needing any change in the device hardware. 2.2.2. Office scanner adaptation. The imaging device
The test disc platform has several key features in its design, exemplied here is a commercially available scanner Cano Scan
rendering merits for the test objective addressed herein. These LiDE 300 (Fig. 2d). The distinct advantage offered by such
are summarized below. a scanner, in addition to its low cost and portability, is uniform
2.1.1. Bubble-free loading and mixing. The loading and illumination in the background over the entire surface of the
pressure release holes for each of the microchannels were scanning bed. Additionally, the entire imaging system is
placed 7.12 mm apart (see Fig. S1 of ESI†) in rapid and optimal enclosed, which eliminates the noise in colorimetric measure-
sample and reagent loading. When the channel is partially lled ments due to ambient factors such as light, temperature,
with a liquid (the test sample or Reagent-1), the liquid that is humidity, dust, etc. Our cassette mask for highlighting the
being loaded subsequently moves alongside the adjacent side target image location is made from 25 mm × 36 cm plywood of
wall of R1 and starts to ll it from the bottom until the meniscus 2 mm thickness. This allows the identical positioning of the
reaches the radial level of the two holes. The proximity of the disc onto the scanner bed each time and further reduces the
holes and the tapered chamber (R1) restricts the bubble experimental noise. It also accounts for the thickness of the disc
formation to a minimum during the mixing step, and the (1.5 mm) on the scanner bed and shields the imaging from stray
meniscus of the colored product does not extend beyond the light. The cassette mask has a circular cut-out region of 110 mm
region of interest during imaging. diameter for accommodating the disc. The circular gap is
2.1.2. Blood plasma separation and the capture of the covered by a colored paper mask containing a negative cut-out
blood cells. The cells are centrifugally separated and collected portion of the regions of interest such as R1. This excludes
in chamber R2 in the fourth layer, while the separated plasma the regions of the disc that are not meant to be imaged, thereby
lies in chamber R1. Chamber R2 has a height of 45 microns and enabling a scan of only the desired reaction spot regions. Such
is shaped in a teardrop fashion, which aids in the expulsion of design adaptation further obviates the need for the undesirable
air from inside the chamber. This results in a clear separation of complicated and obscured In-Camera Pipeline (ICP) features
the cellular components, which obviates the requirement of that are otherwise imperative for the different kinds of imaging
imposing additional pressure. The details are provided in S3 of devices currently used in commercial practice for colorimetric
the ESI.† analysis, noises due to various sizes of the CMOS sensors,
2.1.3. Prevention of backow of cells in the imaging eld. A unavoidable human error, such as the blurring effect during
junction, 0.7 mm long, having a width of 1.20 mm and a height manual handling of the device, inadequate background illu-
of 45 microns connects the R1 and R2 chambers and was mination, texture generation within an image due to different

2430 | Anal. Methods, 2023, 15, 2427–2440 This journal is © The Royal Society of Chemistry 2023
 View Article Online

Paper Analytical Methods

Published on 05 May 2023. Downloaded by Indian Institute of Technology Kharagpur on 8/19/2024 6:54:50 PM.

Fig. 2 (a) The low-cost spin station fabricated in-house. It has 3 programmable modes of rotation and an on/off switch and is powered by a 12 V
DC power source. The box has dimensions of 10 cm × 10 cm × 4.5 cm. (b) Circuit schematic of the portable setup. The speed and direction are
controlled by L298 and AT mega chipset. (c) Spin speed profile for mixing the reagents and samples (plasma and pure analytes); (the x-axis
represents time in seconds, the y-axis represents the rotational speed in rpm). (d) Commercial flatbed scanner with cassette mask of size 25 mm
× 36 cm.

angles of incidence of the source light source, among various 2.3.1. HDL. The HDL cholesterol assay was carried out by
inevitable sources of aberration. Thus, non-iterative, and a direct enzymatic method. This method is dependent on the
computationally simple noise estimation and de-noising algo- property of the detergent in the reagent solution L1 to prefer-
rithms turn out to be adequate for the present scanner sensor- entially solubilize HDL, release the HDL cholesterol, and inhibit
based detection technology, as opposed to the more computa- non-HDL lipoproteins like LDL, VLDL, and chylomicrons by
tionally involved ones for the other alternative optical sensing surface absorption. The free cholesterol formed sequentially
tools. reacts with cholesterol esterase, cholesterol peroxidases and
chromogens present in reagent solution L2 to give a colored
compound.
2.3.2. LDL. The levels of serum LDL cholesterol were
2.3. Reagents and biochemical analysis
directly estimated in a two-step enzymatic process, starting with
All the reagents were purchased from Coral Clinical Systems (a the elimination of non-LDL cholesterol by the reagent solution
division of Tulip Diagnostics Pvt Ltd, India), and were used as L1, followed by the measurement of LDL cholesterol by the
per the manufacturer's protocol (see Table S1 of the ESI†). The reagent solution L2. In the rst step (L1-mediated), non-LDL
test reaction chemistry for each of the lipid prole components cholesterol is selectively degraded by cholesterol esterase and
thus examined is as follows.

This journal is © The Royal Society of Chemistry 2023 Anal. Methods, 2023, 15, 2427–2440 | 2431
 View Article Online

Analytical Methods Paper

cholesterol oxidase to produce H2O2, which then reacts with components of the analytes. No interference was observed with
4AP to produce a colorless end product. In the next step (L2- ascorbic acid up to 50 mg dl−1, hemoglobin up to 500 mg dl−1 or
mediated), H2O2 formed from the LDL cholesterol by the bilirubin up to 30 mg dl−1, complying with the interference
action of cholesterol esterase and cholesterol oxidase undergoes assessment requirements of the gold-standard test reagents.
a peroxidase-catalyzed reaction, forming a colored quinone- However, while these considerations indicated the appropriate
imine complex. working of the underlying biochemical principles for the test
Published on 05 May 2023. Downloaded by Indian Institute of Technology Kharagpur on 8/19/2024 6:54:50 PM.

2.3.3. Triglycerides. The main ingredients of L1 are lipo- workow, there could be other valuable specic indicators
protein lipase and glycerol kinase. Triglycerides are hydrolyzed based on the medical history, lifestyle and other pre-existing
by lipoprotein lipase to glycerol and free fatty acids. Glycerol is non-communicable disease burdens such as diabetes and
then phosphorylated by adenosine-5′-triphosphate (ATP) form- hypertension, which might cumulatively contribute to the risk
ing glycerol-1-phosphate (G-1-P) and adenosine-5′-diphosphate severity and probability of coronary heart diseases, trans-
(ADP) in the reaction catalyzed by glycerol kinase (GK). L2 forming the outcome of the analyte testing alone to the broader
contains glycerol phosphate oxidase (GPO) and 4-amino- perspective of disease diagnosis as per benchmarked clinical
antipyrine (4-AAP), phenol and peroxidase. G-1-P is then inference. For the latter, one needs to combine the diagnostic
oxidized by GPO to dihydroxyacetone phosphate (DAP) and test results with the relevant metadata of the patient along with
hydrogen peroxide (H2O2). Peroxidase catalyzes the coupling of other readily measurable body vitals (such as blood pressure
H2O2 with 4-AAP and sodium N-ethyl-N-(3-sulfopropyl) m-ani- and oxygen saturation) for a holistic and comprehensive
sidine (ESPA) to produce a quinone-imine dye, producing patient-centric assessment. However, the present study focused
a colorimetric signal that is proportional to the triglyceride only on the analytical performance of the test outcome and not
concentration of the sample. on its broader implication towards clinical decision-making
By adhering to the gold standard reaction protocols delin- from such an inclusive viewpoint.
eated above, our test methodology ensures a complete adapta- The benchmark data for the validation studies were
tion of the established gold-standard laboratory recipe, albeit in produced by a commercial biochemistry analyzer (EM360,
a miniaturized format. For each of the lipid prole components, Transasia, India). A double-blind comparison was conducted
the reagent solutions (L1 and L2) of volume 100 ml each are between the results of the present setup and the benchmark
used. The color calibration is done using standard solutions as data. Signicantly, while the commercial biochemistry analyzer
provided by the manufacturer, across a series of different is efficient in handling multiple samples (deploying high-
concentrations for each of the components diluted with phos- quality efficient pumps to handle all the uidic operations
phate buffered saline (PBS; pH 7.4). These benchmarking within the machine), it requires dedicated laboratory space and
reactions are conducted on individual standalone disks expert personnel. In addition, there is a compulsory waiting
following a mixing protocol. The reagents are added, and the period of about 15–20 minutes to allow for the separation of
test products are allowed to settle for 10 minutes. Thereaer, plasma from the blood. In addition, in a commercial setting,
the calibration charts are prepared from those scanned images. the machine is run only when a threshold number of blood
samples are collected and made available for the test as against
responding to individual patient-centric needs, thereby
2.4. Study approval, sample collection and processing increasing the sample-to-answer time.
Following a review of the detailed study plan by the Institute's
Ethical Committee, ethical clearance was granted for the
present study (IEC No. IITKGP/IEC/2022). Blood samples (n = 2.5. Operational steps
92) were collected with the support of the B.C. Roy Technology 2.5.1. Plasma separation. Here, 20 ml of blood sample was
Hospital, Indian Institute of Technology, Kharagpur. Each loaded in each of the three microchannels in the CD. The CD
specimen was collected aer obtaining informed consent from was then mounted on the rotational drive and spun at 2000 rpm
donors voluntarily participating in the study. (u1) for 5 minutes to separate plasma from the whole blood
The samples were collected over ve months from the (Fig. 3ii). At the end of the rotation, the plasma was distinctly
voluntary participants. The venous samples were collected in separated from the cells. The cells lie in R-2 and the plasma lies
EDTA vacutainers as per hospital requirements; 60 microliters in R-1 (see Fig. 1).
of the collected blood were used for the present assay. The 2.5.2. Mixing. In the next step, 100 ml of reagent solution L1
capillary blood was collected separately using a regular pen was loaded into the disc, followed by 100 ml of L2. The separated
lancet and was immediately used for the assay. There was no plasma and the reagents were mixed as per the spin prole u2
need for sample storage, and in situ testing at the point of (Fig. 2c). At the end of the rotation, a uniform-colored product
collection was possible, which greatly reduced the sample-to- was formed in each of the four microchambers (Fig. 3vii). The
answer time. Appropriate biosafety practices were used for the spin prole for u2 is a continuous increase and decrease in the
sample collection. The reagents used for lipid proling were speed of the rotor, bounded within 1300 rpm (clockwise) to
procured from accredited commercial sources to meet the −1300 rpm (counterclockwise). This cyclic variation in the
laboratory standards and thus, to rule out any deviations from direction of rotation facilitates the ‘complete mixing’ of the
the quality control for the reagent preparation. Nevertheless, we separated plasma and the two reagents. The ‘complete mixing’
further examined the reagent cross-reactions with various here implies that once the rotation stopped, the color of the

2432 | Anal. Methods, 2023, 15, 2427–2440 This journal is © The Royal Society of Chemistry 2023
 View Article Online

Paper Analytical Methods

Published on 05 May 2023. Downloaded by Indian Institute of Technology Kharagpur on 8/19/2024 6:54:50 PM.

Fig. 3 Schematic of the experimental process. This depicts the process flow starting with the loading of blood in the microchannel and ending
with the data acquisition in a computer interface. It may be noted that colors displayed in (vii) do not represent the colors of the actual reaction
products but merely convey the fact that the reaction products have colors that are distinct and easily distinguishable with the naked eye.

reaction products does not change with time. Every rotation in this background region are assumed to be a result of noise
cycle lasts 14 seconds and consists of clockwise acceleration during the processing. The overall pre-processing step consists
followed by anticlockwise deceleration. Overall, three cycles of the identication of a background, selecting random patches
totaling 52 seconds turned out to be optimal for satisfactory from the background, and nally, noise estimation and de-
colour generation aer the uid mixing, striking a balance noising as shown schematically in Fig. 4ii. The histogram
between the power consumption and the intended purpose. plots of the spatial variation of the uniform background for the
2.5.3. Detection. The disc was allowed to rest for 10 min at scanner sensor-based images are shown in Fig. 4i.
room temperature aer the mixing step. It was then disengaged The histogram plot shows that the spatial noise or variation
from the rotational drive and aligned with the mask (Fig. 3viii) present in the scanner sensor can be modelled well by
and kept inside the adapted scanner. The image was taken a Gaussian probability density function (pdf), and the noise
against a white background and may be saved in any scanner- variance can be estimated from the approximated Gaussian
compatible unit such as a tablet, laptop or desktop computer. characteristics. A non-iterative Gaussian local smoothing is
adequate for smoothening the Additive White Gaussian Noise
2.6. Image analytics (AWGN), and signicantly improving the Signal-to-Noise ratio
2.6.1. Denoising. To capture the spatial noise characteris- (SNR). The measured pixel intensity, f(x,y), at a pixel location
tics for the scanner sensors, we utilized the uniform and (x,y), can be perceived as the sum of the true pixel intensity
homogeneous background around the region of interest. g(x,y) and the AWGN with zero mean and an estimated variance
Specically, any variations in the pixel intensities within a patch as given by the following:

This journal is © The Royal Society of Chemistry 2023 Anal. Methods, 2023, 15, 2427–2440 | 2433
 View Article Online

Analytical Methods Paper

Published on 05 May 2023. Downloaded by Indian Institute of Technology Kharagpur on 8/19/2024 6:54:50 PM.

Fig. 4 (i) A histogram plot of spatial variations of the scanner device. (ii) Block diagram showing the noise removal from the image. (a) Input of the
raw image, (b) identification of the background by selecting random patches from the corresponding black and white image, (c) subsequent
noise estimation by the denoising algorithm. (iii) Flowchart of the image processing operations. Four chambers illustrate that it is possible to
increase the number of samples being tested on a single disposable disc by suitably arranging the microchannels on the disc: (a) image of the disc
after mixing, (b) segmentation of the background from the foreground, (c) extraction of individual chambers (ROIs) for further processing, (d)
fine-tuning, (e) extraction of color space features and correlating with benchmark data to develop a regression model, (f) developing the final
calibration equations from regression analysis.

f(x,y) = g(x,y) + h(x,y) (1) E(f(x,y)) = E(g(x,y)) (2)

The expected or mean value of the measured pixel intensity

(averaged over a selected random patch) should be equal to the 2.6.2. Bubble detection and other pre-processing steps.
true pixel intensity with the assumption of proper estimation of Bubbles may be unwarrantedly introduced into the visual eld
noise variance. of the region of interest (ROI), due to improper reagent loading

2434 | Anal. Methods, 2023, 15, 2427–2440 This journal is © The Royal Society of Chemistry 2023
 View Article Online

Paper Analytical Methods

within the chambers, which is unavoidable in uncontrolled interfaces with the image analytics soware to predict all the
extreme POC settings. The presence of such bubble spots and components of the lipid prole, i.e., HDL, LDL, and TG, and
other undesired regions in the ROI may deviate the extracted displays the same on the screen.
feature values from the desired color feature. To circumvent
this, a contour detection algorithm was implemented for
detecting any inner contours within the outer contour 3. Results and discussion
Published on 05 May 2023. Downloaded by Indian Institute of Technology Kharagpur on 8/19/2024 6:54:50 PM.

boundary. Further, to strengthen the bubble spot detection, the 3.1. Calibration of the proposed device
saturation plane of the extracted ROI was taken as a threshold
The device was rst calibrated with standard sample solutions
to distinguish the colorless bubble region from the colorful
of the different analytes (see S2 of ESI†). The readings were
ROI. A series of morphological operations, namely dilation,
taken for three separate experiments on three separate discs,
erosion, opening and closing were performed to lter out any
which depicted the strong correlation and discrimination
noise in the segmented region of interest, resulting in the nal
ability of the selected features for the different standard solu-
test image as illustrated in Fig. 4.
tion concentrations used for the in vitro analysis. This suggests
2.6.3. Outlier detection, feature selection and cross-vali-
that the averages of the RGB components of the images are
dation. There is always the possibility of confronting outlier
signicantly correlated with the respective analyte concentra-
data due to human error, instrument error or simply due to
tions. The selected features from the standard sample solutions
natural personalized variabilities within the test populations.
were then used to detect the outliers and to nally perform the
For this reason, ltering out the outliers from the training set is
cross-validation operation on the training set for evaluating the
of paramount importance to rule out any bias in calibration due
performance of the algorithm on real-time venous and capillary
to outlier samples. Here, we implemented an error residual-
human blood samples.
based Random Sample Consensus (RANSAC) algorithm to
detect and remove the outlier samples from the training set.
To realize this, the different color space features like RGB, 3.2. Limit of blank (LoB)
HSV, YCbCr, and lab average features were extracted from the LoB measures the analytical signal that may otherwise be
segmented ROI. The correlation coefficients of these features present in a sample with an ultra-low low concentration of any
with the true analyte concentrations were calculated. Based on of the comparative analytes of HDL, LDL and TG. To assess the
this correlation coefficient, the best image features were same, DI water devoid of any analyte was used for the blank
selected. The average red, green and blue features of the RGB measurement. Here, 200 ml DI water was loaded in each of the
color space yielded the best correlation with the 3-analyte four chambers of a test disc and then scanned. Four such
concentrations as evidenced by the training and the test data readings were taken for 10 chambers. The blank was calculated
sets. Further, to eliminate any bias in the calibration curve, a 10- from the RGB features. Statistically, it corresponds to the center
fold cross-validation method was implemented. The training of the assumed Gaussian distribution of the raw analytical
set was randomly divided into 10 almost equal-sized subsets. signals from the blank samples. The LoB conforms to the region
The cross-validation operation was repeated 10 times; each time where 90% of the observed values of the samples containing no
using a single fold as the testing set and the other 9 folds as the analyte will lie, quantied as follows: LoB = Meanblank + 1.645
training set. Finally, the obtained calibration curve thus ob- SDblank. The other 5% of the readings is the probable response
tained was averaged to remove any bias in the training set. of a very low concentration of an analyte. The LoB values turned
2.6.4. Graphical user interface (GUI) development. An in- out to be 5.46, 4.16 and 35.76 mg dl−1 for HDL, LDL and TG,
house GUI was developed based on OpenCV 4.0.1. The Scikit respectively.
Image 0.18.1 libraries of Python were further used to process the
scanned images. Scikit Learn 0.24.1 was used for the regression
analysis and development of calibration equations. The scan- 3.3. Limit of detection (LoD)
ned image was processed with a Python code. First, the LoD is a measure of the smallest analyte concentration that
segmentation allows the extraction of the foreground (colored must be present to produce a detectable color change. It was
reaction products) from the background (paper mask) (Fig. 4ii). measured and compared to the colorimetric signal produced by
Second, the ROI was extracted from the segmented image. The DI water. The mean and standard deviation (SD) of the blank
colorimetric features for the R1 regions (Fig. 1) corresponding chamber readings were calculated to provide a conservative
to the lipid prole components then became available for value of 8.10 mg dl−1 for HDL, 5.31 mg dl−1 for LDL and
detection and interpretation. Next, the ROIs were ne-tuned to 45.48 mg dl−1 for TG. These values were compared with the
remove any noise that might be introduced because of improper standard low concentrations of cholesterol of 25 mg dl−1,
mixing and non-uniform color generation at the edges of the 12.67 mg dl−1 of HDL, 38 mg dl−1 of LDL and 66.67 mg dl−1 of
microchannels or due to bubble formation (Fig. 5). Next, the TG. The LoD was calculated as follows: LoD = LoB + 1.645 SD.
color space features were extracted for each of the four cham- The resulting values were 8.10, 5.31 and 45.48 mg dl−1, for HDL,
bers and were correlated with the benchmark data. Processing LDL and TG, respectively. The LoD and LoB values indicated
and analysis of the data generated from the image analysis were that this setup may be reliably used to differentiate the presence
done using MS Excel and MATLAB. Finally, the calibration or absence of lipid proling parameters, and is, therefore,
equations were generated by regression analysis. The GUI specic to the intended analytical purpose.

This journal is © The Royal Society of Chemistry 2023 Anal. Methods, 2023, 15, 2427–2440 | 2435
 View Article Online

Analytical Methods Paper

Published on 05 May 2023. Downloaded by Indian Institute of Technology Kharagpur on 8/19/2024 6:54:50 PM.

Fig. 5(a) Color noise introduced due to incomplete mixing. (b) The removal of color noise developed at the edge. (c) The presence of bubbles in
the microchannel generated due to sample loading-related experimental error. (d) The detection and removal of bubbles.

3.4. Validation results with venous (V) and capillary (C) benchmark. The efficacy of the present technology in cardio-
blood drawn from human subjects vascular disease screening was subsequently assessed by
comparing the readout from the discs and the laboratory
We analyzed the data based on medically established limits of
standards (see Fig. 6).
all three target analytes: triglycerides, HDL, and LDL. The
The graph shows that our POC estimated data is strongly
thresholds for these are, respectively, 200 mg dl−1, 35 mg dl−1
correlated with the automated biochemistry analyzer bench-
and 160 mg dl−1. Any samples for triglycerides having
mark data for the various samples (venous and capillary) used
a concentration greater than 200 mg dl−1, HDL less than 35 mg
in the in vitro analysis. It implies that our selected feature
dl−1 and LDL more than 160 mg dl−1 were considered risk
vectors, averages of the RGB components of the images, are
positive. Based on the actual (gold standard) and the predicted
values obtained from our method, we quantied the true posi- signicantly correlated with the analyte concentrations and
a quantitative estimation of HDL, LDL and TG can be inferred
tive (TP), true negative (TN), false positive (FP) and false nega-
without compromising quality.
tive (FN) samples. The nal evaluation metrics like accuracy,
According to clinical signicance,56,57 the high values for
positive predictability, etc. were derived from these parameters.
triglycerides and LDL (above the healthy limit) and low values
For evaluating the continuous value prediction of TG, HDL
for HDL (below the healthy limit) were considered as the posi-
and LDL for the unknown venous and capillary samples, the
tive (vulnerable) classes for potential cardiovascular disease
root mean square error (RMSE) performance metric was used.
screening. The four-performance metrics depend on true posi-
For our large training and test sets, the RMSE for the three
analytes TG, HDL and LDL were found to be 7.23, 11.58 and tive, true negative, false positive and false negative values. If an
originally positive class as evaluated by the automated
6.71 mg dl−1 and 7.09, 10.85 and 7.01 mg dl−1 for capillary and
biochemistry analyzer (gold standard) device is also detected as
venous samples, respectively. Further, to validate the device and
positive by our device, it is declared as a true positive; otherwise,
the test method, we used four different performance metrics,
it is classied as a false negative. Similarly, if a negative class as
namely accuracy, sensitivity, specicity and positive predictivity
per the evaluation of the gold standard is also detected as
(see Table 1). First, the gold standard laboratory data obtained
a negative class by our device, it is considered as a true negative;
from the automated biochemistry analyzer was taken as the
otherwise, the outcome is considered as a false positive. The

2436 | Anal. Methods, 2023, 15, 2427–2440 This journal is © The Royal Society of Chemistry 2023
 View Article Online

Paper Analytical Methods

Table 1 Performance characteristics of the proposed device for venous (V) and capillary (C) blood

TG (V) TG (C) HDL (V) HDL (C) LDL (V) LDL (C)

Accuracy 0.892 0.935 0.729 0.794 0.837 0.881

Sensitivity 0.917 0.875 0.738 0.754 0.735 0.824
Specicity 0.883 0.956 0.71 0.871 0.897 0.914
Published on 05 May 2023. Downloaded by Indian Institute of Technology Kharagpur on 8/19/2024 6:54:50 PM.

Positive predictivity 0.733 0.875 0.833 0.92 0.807 0.849

accuracy of the device indicates its efficacy in detecting a posi- determines how many samples are truly positive among all the
tive or negative class correctly and is expressed as the ratio of detected positive samples.
the sum of the true positive and true negative cases detected to The numerical values of all these metrics are high, indicating
the total number of samples. On the other hand, sensitivity is a high-performance quality of the proposed device for both the
the metric for determining how correctly the device detects capillary blood and venous blood. Moreover, the correlation
a positive class and specicity is the metric of how correctly the coefficients for capillary and venous blood samples with the
device detects a negative class. Additionally, positive predictivity benchmark data (see Fig. 6) are similar and close to unity. It was

Fig. 6 Experimental data. The left (a)–(c) and the right (d)–(f) panels show the comparison of the lipid profiles as measured by the proposed
device with the benchmark for capillary blood samples and venous blood samples, respectively for HDL, LDL and TG. The vertical (y) axis
corresponds to the measured values from the portable device, whereas the horizontal (x) axis corresponds to the benchmark data, i.e., as
measured by the commercial biochemistry analyzer. The benchmark data for both the capillary and venous blood samples were generated by
the processing of venous blood samples in the commercial biochemistry analyzer.

This journal is © The Royal Society of Chemistry 2023 Anal. Methods, 2023, 15, 2427–2440 | 2437
 View Article Online

Analytical Methods Paper

Table 2 A comparison of the proposed device Lipidest with other commercial portable devices for measuring the lipid profile

Fully
loaded
Run time weight Cost
Product Manufacturer Chemistry (min) Sample preparation steps (kg) Power supply (USD)a
Published on 05 May 2023. Downloaded by Indian Institute of Technology Kharagpur on 8/19/2024 6:54:50 PM.

Piccolo Xpress® Abbott Colorimetric 12 3 5 15 V DC car DC: 5200,

measurement battery PTC: 11
MS-S600 Ningbo Medical System Colorimetric 13–15 1 4.9 110–240 V AC, NA
(mWafer) Biotech. Co. Ltd measurement 50–60 Hz
Allegro® Nova Biomedical Immunoassay 3–10 5 10.43 Mains supply NA
Cobas b 101 Roche Dry chemical assay, 4–6 3 2 115–230 V 50/ DC: 2360,
system immunoassay 60 Hz PTC: 90
Skyla™ HB1 Skyla Colorimetric 13.5 3 5.5 Mains supply NA
POC analyzer measurement
Skyla Hi analyzer Skyla Combined turbidimetry, 6–13 3 1.7 Mains supply NA
colorimetry and ELISA
SimplexTAS 101 Tascom Immuno-chemistry 10 NA 3.4 100–240 V AC NA
Lipidest Proposed system Colorimetric reectance 10 None for the venous 1.9 12 V DC DC: 25,
developed in-house sample, 1 for the capillary PTC: 0.60
sample
a
There are two types of costs: DC is the device cost that involves everything related to setting up the equipment, whereas PTC is the cost per test that
is related to the consumables required for each new test once the system/equipment is functional.

observed that the spread of the capillary blood data is more devices (see Table 2) establishes the advantages that our device
towards the lower limits in comparison to the venous blood, offers over the state-of-the-art. The weight of the fully loaded
which eventually leads to a slightly weaker correlation of device, run time and power demand of Lipidest is amongst the
capillary blood with benchmark data (Fig. 6c and f). This is lowest. There are practically no sample preparation steps of an
more prominent for triglycerides. This is because the compar- advanced version of the proposed device where blood dilution
ison of capillary blood data from the disc is done with the can be automated with the aid of suitable modications in the
venous blood data from a commercial biochemistry analyzer. spin prole and reagents can be stored on-disc. A rudimentary
This is primarily attributed to the different sources of blood cost comparison shows a nearly 100-fold reduction in both the
being analyzed in the commercial analyzer and the POC device. device cost and per-test costs. This is gained at no loss in
Secondly, the capillary blood was diluted before use in the disc accuracy when compared to the gold standard results.
due to its low sample volume availability. Therefore, it is likely
that the true concentration of the parent sample is not reected
in the measured values. The composition of capillary blood 4. Conclusions
itself differs from venous blood, even if the blood is drawn from
A simple, scalable, and portable device technology has been
the same individual at the same time. This is due to the prev-
developed that reliably measures HDL, LDL and triglycerides
alent unique characteristics of hemodynamics through struc-
from a tiny volume of blood sample by the miniature adaptation
turally different veins and capillaries.
of the respective gold standard protocols via independent
Despite all these limitations, the present device displays
measurements, without using any biased correlations among
excellent agreement between the measured capillary blood lipid
them. The diagnostic device comprises an inexpensive plastic
prole and the benchmark lipid prole of venous blood,
compact disc, a regular office scanner and a laptop (or a tab or
substantiating the fact that while the above-mentioned devia-
smartphone). The scanner was tted with a mask made from
tions in the characteristic analyses with capillary and venous
a wooden plank and a piece of paper and was adapted as the
blood are of fundamental concern, these do not turn out to be
colorimetric detection module. The colorimetric analysis was
the primary inuencing factors in the quantitative detection
done with an image-processing algorithm to produce digitized
methodology adapted herein, where several statistical measures
calibration charts integrated with a mobile application inter-
are already embedded to accommodate inevitable variabilities
face. As a future modication, an even smaller version of such
in the sources of the test sample. This turns out to be very
an imaging device may be custom-made, where the scanner bed
promising for the deployment of the present device at point-of-
size is adjusted to the disc size. This would allow greater
need, either at the settings that allow the drawing of a venous
convenience in transportation and handling during operation.
blood sample for direct use, and at home/low-resource settings
Despite disruptive simplications in the device technology
where nger-prick blood can be the preferred source of the test
as against the automated high-end and expensive devices used
sample.
in established pathological laboratories, no signicant
A comparison of the characteristics of Lipidest with the other
discrepancies were observed between the results obtained from
commercial rapid and/or portable lipid prole measurement
the commercial biochemistry analyzer and our device setup.

2438 | Anal. Methods, 2023, 15, 2427–2440 This journal is © The Royal Society of Chemistry 2023
 View Article Online

Paper Analytical Methods

This could be due to the favourable correlation between the 15 D. Grabowski, J. Aagaard-Hansen, I. Willaing and
results obtained from our device with the gold standard B. B. Jensen, Soc, 2017, 7, 10.
laboratory-centric technology. This promises the adoption of 16 P. St-Louis, Clin. Biochem., 2000, 33, 427–440.
the present device as a rapid, inexpensive, albeit reasonably 17 M. Nakamura, H. Iso, A. Kitamura, H. Imano, M. Kiyama,
accurate, extreme POC technology for low-resource settings S. Yokoyama, Y. Kayamori, I. Koyama, K. Nishimura,
where other more established modes of cardiovascular risk M. Nakai, M. Dasti, H. W. Vesper, T. Teramoto and
Published on 05 May 2023. Downloaded by Indian Institute of Technology Kharagpur on 8/19/2024 6:54:50 PM.

assessment are either scarce or simply non-existent. Y. Miyamoto, Clin. Chim. Acta, 2015, 445, 127–132.
18 N. Chauhan, J. Narang and C. S. Pundir, J. Anal. Chem., 2010,
Conflicts of interest 01, 41–46.
19 A. Ahmadalinezhad, A. K. M. Ka and A. Chen, Electrochem.
There are no conicts to declare. commun., 2009, 11, 2048–2051.
20 T. Ferri, A. Poscia and R. Santucci, Bioelectrochem. Bioenerg.,
Acknowledgements 1998, 44, 177–181.
21 M. Dhawane, A. Deshpande, R. Jain and P. Dandekar, Sens.
The authors acknowledge Mr Rajib Das of BC Roy Hospital at Actuators B Chem., 2019, 281, 72–79.
IIT Kharagpur for helping with the validation studies and 22 S. K. Arya, G. Chornokur, M. Venugopal and S. Bhansali,
supplying biological samples. SC acknowledges the nancial Biosens. Bioelectron., 2010, 25, 2296–2301.
support provided by DST, Government of India, through Sir J. C. 23 H.-C. Chang and J. A. Ho, Anal. Chem., 2015, 87, 10362–
Bose National Fellowship. 10367.
24 N. R. Nirala, S. Abraham, V. Kumar, A. Bansal, A. Srivastava
References and P. S. Saxena, Sens. Actuators B Chem., 2015, 218, 42–50.
25 M. Zarei, Biosens. Bioelectron., 2017, 98, 494–506.
1 J. Stamler, M. L. Daviglus, D. B. Garside, A. R. Dyer, 26 Y. Liu, G. Zhu, J. Yang, A. Yuan and X. Shen, PLoS One, 2014,
P. Greenland and J. D. Neaton, JAMA, J. Am. Med. Assoc., 9, e109158.
2000, 284, 311–318. 27 C.-L. Hsu, C.-W. Lien, C.-W. Wang, S. G. Harroun,
2 D. M. Lloyd-Jones, L. T. Braun, C. E. Ndumele, S. C. Smith, C.-C. Huang and H.-T. Chang, Biosens. Bioelectron., 2016,
L. S. Sperling, S. S. Virani and R. S. Blumenthal, 75, 181–187.
Circulation, 2019, 139, E1162–E1177. 28 H. Wei and E. Wang, Chem. Soc. Rev., 2013, 42, 6060.
3 Y. Zhu, J. M. Lu, Z. Bin Yu, D. Li, M. Y. Wu, P. Shen, H. B. Lin, 29 S. J. Pollack, J. W. Jacobs and P. G. Schultz, Science, 1986,
J. B. Wang and K. Chen, Nutr. Metab. Cardiovasc. Dis., 2019, 234, 1570–1573.
29, 1205–1213. 30 Y. Lu, N. Yeung, N. Sieracki and N. M. Marshall, Nature,
4 T. Bulum, M. Tomić and L. Duvnjak, Arch. Med. Res., 2017, 2009, 460, 855–862.
48, 467–471. 31 Y. Li, Z. Kang, L. Kong, H. Shi, Y. Zhang, M. Cui and
5 E. Olié, M. C. Picot, S. Guillaume, M. Abbar and P. Courtet, J. D.-P. Yang, Mater. Sci. Eng. C, 2019, 104, 110000.
Affect. Disord., 2011, 133, 234–238. 32 V. Sharma and S. M. Mobin, Sens. Actuators B Chem., 2017,
6 R. C. Pasternak, ACC Curr. J. Rev., 2002, 11, 37–45. 240, 338–348.
7 S. L. Kraav, T. Tolmunen, O. Kärkkäinen, A. Ruusunen, 33 R. Li, C. Xiong, Z. Xiao and L. Ling, Anal. Chim. Acta, 2012,
H. Viinamäki, P. Mäntyselkä, H. Koivumaa-Honkanen, 724, 80–85.
M. Valkonen-Korhonen, K. Honkalampi, K. H. Herzig and 34 E. W. Nery, Analysis of Samples of Clinical and Alimentary
S. M. Lehto, Psychiatry Res., 2019, 272, 326–333. Interest with Paper-based Devices, Springer International
8 F. D. Porter and G. E. Herman, J. Lipid Res., 2011, 52, 6–34. Publishing, 2016.
9 M. Cassader, G. Ruiu, R. Gambino, N. Alemanno, F. Veglia 35 D. Surabattula, C. M. Harvey, F. Aghazadeh, J. Rood and
and G. Pagano, Atherosclerosis, 1993, 99, 47–53. A. Darisipudi, Int. J. Ind. Ergon., 2009, 39, 167–173.
10 W. Hong, V. Zimmer, Z. Basharat, M. Zippi, S. Stock, 36 M. A. Al-Rawhani, B. C. Cheah, A. I. Macdonald, C. Martin,
W. Geng, X. Bao, J. Dong, J. Pan and M. Zhou, Clin. Nutr., C. Hu, J. Beeley, L. C. Gouveia, J. P. Grant, G. Campbell,
2020, 39, 250–257. M. P. Barrett and D. R. S. Cumming, IEEE Sens. J., 2017,
11 M. G. Martı́n, F. Pfrieger and C. G. Dotti, EMBO Rep., 2014, 17, 240–247.
15, 1036–1052. 37 M. A. Al-Rawhani, S. Mitra, M. P. Barrett, S. Cochran,
12 X. Yu, T. J. Kelley, H. J. Chiel and J. D. Burgess, J. Electrochem. D. R. S. Cumming, C. Hu, C. Giagkoulovits, V. F. Annese,
Soc., 2016, 163, B453–B455. B. C. Cheah, J. Beeley, S. Velugotla, C. Accarino and
13 L. S. Eberlin, A. L. Dill, A. B. Costa, D. R. Ifa, L. Cheng, J. P. Grant, IEEE. Trans. Biomed. Eng., 2020, 67, 614–623.
T. Masterson, M. Koch, T. L. Ratliff and R. G. Cooks, Anal. 38 Y. Yoneyama, Y. Yonemori, M. Murata, H. Ohnuki, K. Hibi,
Chem., 2010, 82, 3430–3434. T. Hayashi, H. Ren and H. Endo, Talanta, 2009, 80, 909–915.
14 A. Murphy, J. R. Faria-Neto, K. Al-Rasadi, D. Blom, 39 M. Situmorang, Talanta, 1999, 49, 639–649.
A. Catapano, A. Cuevas, F. Lopez-Jimenez, P. Perel, 40 W. C. Shih, M. C. Yang and M. S. Lin, Biosens. Bioelectron.,
R. Santos, A. Sniderman, R. Sy, G. F. Watts, D. Zhao, 2009, 24, 1679–1684.
S. Yusuf and D. Wood, Glob. Heart, 2017, 12, 179–197.e5.

This journal is © The Royal Society of Chemistry 2023 Anal. Methods, 2023, 15, 2427–2440 | 2439
 View Article Online

Analytical Methods Paper

41 Y. Kayamori, H. Hatsuyama, T. Tsujioka, M. Nasu and 52 S. Basu, D. Flood, P. Geldsetzer, M. Theilmann,

Y. Katayama, Clin. Chem., 1999, 45, 2158–2163. M. E. Marcus, C. Ebert, M. Mayige, R. Wong-McClure,
42 T. Quah, S. Abdellaoui, R. D. Milton, D. P. Hickey and F. Farzadfar, S. Saeedi Moghaddam, K. Agoudavi, B. Norov,
S. D. Minteer, J. Electrochem. Soc., 2017, 164, H3024–H3029. C. Houehanou, G. Andall-Brereton, M. Gurung, G. Brian,
43 A. A. Ivanov, N. V. Yudina, E. I. Korotkova and P. Bovet, J. Martins, R. Atun, T. Bärnighausen, S. Vollmer,
O. I. Lomovsky, Solid Fuel Chem., 2008, 42, 68–73. J. Manne-Goehler and J. Davies, Lancet Glob. Health, 2021,
Published on 05 May 2023. Downloaded by Indian Institute of Technology Kharagpur on 8/19/2024 6:54:50 PM.

44 K. V. Derina, E. I. Korotkova, E. V. Dorozhko, O. A. Voronova 9, e1539–e1552.

and D. A. Vishenkova, Procedia Chem., 2014, 10, 513–518. 53 P. Salazar, M. Martı́n and J. L. González-Mora, J. Electroanal.
45 G. Li, J. M. Liao, G. Q. Hu, N. Z. Ma and P. J. Wu, Biosens. Chem., 2019, 837, 191–199.
Bioelectron., 2005, 20, 2140–2144. 54 U. Hanefeld, L. Gardossi and E. Magner, Chem. Soc. Rev.,
46 A. Merkoçi, M. Pumera, X. Llopis, B. Pérez, M. del Valle and 2009, 38, 453–468.
S. Alegret, TrAC, Trends Anal. Chem., 2005, 24, 826–838. 55 T. J. Anderson, J. Grégoire, G. J. Pearson, A. R. Barry,
47 J. J. Gooding, Electrochim. Acta, 2005, 50, 3049–3060. P. Couture, M. Dawes, G. A. Francis, J. Genest, S. Grover,
48 S. Moussa and J. Mauzeroll, J. Electrochem. Soc., 2019, 166, M. Gupta, R. A. Hegele, D. C. Lau, L. A. Leiter, E. Lonn,
G25–G38. G. B. J. Mancini, R. McPherson, D. Ngui, P. Poirier,
49 A. K B, A. Varghese, N. M and L. George, J. Electrochem. Soc., J. L. Sievenpiper, J. A. Stone, G. Thanassoulis and R. Ward,
2019, 166, B1016–B1027. Can. J. Cardiol., 2016, 32, 1263–1282.
50 V. C. Soorya and S. Berchmans, J. Electrochem. Soc., 2017, 56 Expert Panel on Detection, Evaluation, and Treatment of
164, B435–B442. High Blood Cholesterol in Adults, JAMA, J. Am. Med. Assoc.,
51 D. Wang, C. Liu, Y. Liang, Y. Su, Q. Shang and C. Zhang, J. 2001, 285, 2486–2497.
Electrochem. Soc., 2018, 165, B361–B369. 57 K. P. Iyengar, V. K. Jain and R. Vaishya, Postgrad. Med. J.,
DOI: 10.1136/postgradmedj-2020-138496.

2440 | Anal. Methods, 2023, 15, 2427–2440 This journal is © The Royal Society of Chemistry 2023