Patent Application
20240337630
The present application relates to a method and system for detection and analysis of chemical compounds in a sample subjected to a chromatographic separation on a layered separating medium.
Analytical and chemical sensing tools encompass a wide range of instrumentation whose principal purpose is to analyse chemical makeup of a sample and quantity of each component within a sample1. These tools have been going endless expansion in monitoring and control mechanisms in a wide variety of uses in pharmaceutical, chemical, oil refineries, clinical, and food processing laboratories. As of today, these tools exhibit a varying degree of complexity, from comprehensive laboratory-based instrumentation to portable handheld sensor systems. New methods empowered by new materials, physics principles, biochemistry designs, or data science abound, exhibiting a constant improvement of these former established methods.
As of today, choice of sensing tool depends on intended goal of analysis, the compound(s) being targeted, and the setting mode. On the one side, large laboratory-based or even compact spectrometry or spectroscopy offers comprehensive analysis, quantitation, and identification of wide range of chemical compounds, and are highly suited to discovery studies of small inorganic and/or organic molecules. These tools are bulky and demands extensive sample pre-preparation and trained people, therefore, that are not suitable for onsite and/or direct sample analysis. On the other side of scale, low-cost sensors, whether as selective/lock-and-key devices or cross-reactive arrays in conjugation with pattern recognition methods can be tuned to respond to specific substances and/or as chemical fingerprints of mixtures. Nevertheless, current sensing technologies provide almost one single superposed response pattern, disregarding whether targeted samples contain single compound or a complex mixture of compounds. For the latter, countless possible combination of each component in a mixture, which need to be tested and trained for a reliable prediction in real sample analysis, complicates and decreases reliable prediction in real sample analysis.
For each of two sensing categories, detection of intriguing chemical molecules, such as chiral molecules, is rather difficult mission to achieve. This could be due to insufficient chiral atmosphere at host device, which, sometimes, is not enough to resolve a racemic/enantiomeric mixture easily. Use of helical geometry or chiral centre at sensor side could provide opposite directional mode of change of sensor resistance for pure enantiomeric compound. However, it remains challenging to analyse a mixture (racemic/enantiomeric) as the change of sensor resistance suppress each other due to superposition of two opposite effect.
It is also an issue of circular dichroism spectroscopic chiral recognition when polarisation of light (left/right) provides no significant rotation in presence of racemic mixture due to similar compensation effect. It is also very common that sensor exhibit same directional change with different sensitivity for two opposite handed molecules, thereby a highly challenging task to recognize wide range mixed enantiomer. Studies have tried to overcome this challenge by preceding the sensors with a chromatography column, viz. a stationary phase that separate gas-mixture to individual components and releasing them one-by-one to sensor. Nevertheless, this approach still suffers from similar drawbacks to those of conventional spectrometry. Altogether, these inherent problems limit scope of sensors for application in broader perspective and could be the reason for not achieving same success rate as spectrometry.
For high-impact applications that could be adopted by businesses, government, and individuals there is a need for a disruptive miniaturised or wearable spectrometry approach that could be suspended as stickers everywhere in our daily life (indoor, outdoor, building, human body, plants, etc.) to analyse widest spectrum of chemical compounds in a local environment, exactly as lab spectrometer do. Such system would bring remarkable proliferation both as individual systems as well as network of systems that communicate between them and probably other mobile devices over large surface areas to identify health risks, optimize agricultural production, and manage supply chains.
With present technological advancement, unmet need of multifunctional sensor in every sphere of life demands extensive research motivation. In fact, sensors and detectors are becoming essential part from human physiological monitoring to Mars exploration. Especially wearable and flexible sensors and electronics are gaining much attention as potential next generation electronic skin device for direct monitoring of various physiological parameters (such as, volatile organic compounds (VOC) emission from skin to detect various diseases, heart-beat monitoring, assessment of various liquid biomarkers from sweat/saliva/tears), epidermal vibration monitoring for sound recognition from throat, physical stimuli monitoring of limbs (temperature, humidity, strain, pressure, texture, shape sensation).
In living organism, the detection capability of various VOCs using olfactory receptor is far superior to present day electronic noses. Some recent works suggest the integration of these smell sensitive organic receptors with the artificial electronic materials could be an effective way to construct bio-hybrid sensing network. However, there are issues with long term operation, reliability, reproducibility, strict bio-friendly environment, and poor understanding about the direct replicating a whole biological smell system. Till now gas chromatography, nuclear and mass spectrometry are proven to be useful to discriminate and detect complex VOC mixture, however, it is bulky, time-consuming and demands extensive sample pre-preparation and trained people. Therefore, there is a strong and long-felt need to introduce methods capable of handling any amount of generated spectrometry data in a fast and efficient way without substantial expenses.
While the gas chromatography, nuclear and mass spectrometry are not suitable for onsite sample analysis directly from a patients' skin, breath, or other physiological samples, the room temperature operable electronic sensors arrays using carbon nanotube, 2D materials and noble metal nanoparticles can do. They are small, low power and ideal for both onsite analysis and wearable settings.
Non-invasive monitoring and early diagnosis of various diseases by wearable flexible devices opens new frontiers for distant and real time healthcare monitoring in non-clinical settings. Human skin bears a ton of information about overall physical condition of the body through skin volatile organic compounds, heartbeat, sweat components [1-2]. Therefore, monitoring of these physiological parameters with wearable/skin implantable format would be a game changer for early diagnosis of cancer, kidney/liver diseases, diabetes. Although gas chromatography is the gold standard for correctly estimating the chemical/organic compound from patients' sample, however, it is impractical to use as wearable format due to large device size and complex procedure.
Recently 2D layered material-based printed electronics (such as, graphene, transition metals, dichalcogenide) brings a lot of research and industrial attention due to their excellent mechanical flexibility in conjunction with tuneable electronic properties. However, significant industrialisations of these sensor materials are still a challenge due to the poor selectivity that causes inaccurate estimations of complex VOC mixture, greater device to device variation that restricts large scale fabrications, stringent clean room environment that increases the cost, inadequate similarity with traditional silicon electronics that restricts integration capability.
In conventional VOC sensor array measurements, each sensor in the array exhibits a certain exponential transient reaction kinetics which is the superposition of the response profile of multiple VOCs present in the mix condition, thus, it is challenging for analysing the VOC contents when the number of VOCs become higher. In particular, for real human physiological samples from breath, blood, urine and skin, the confounding and background VOCs are very large in number. Therefore, a cost effective, robust, scalable, sustainable, wearable, highly reliable technological advancement and maturity is needed for continuous monitoring of various volatile chemical/organic compounds released by human body to assess various diseases in early stage or monitoring the disease progression/remission during treatment completely non-invasively. This is also highly important for recent SARS-COV-2 virus detection and rapid isolation in cross-border and airports using breath-biopsy-chromatography analyser which could be much faster the conventional nucleic acid amplification tests (NAATs), antigen test and serology test.
The present invention relates to a method for detection and analysis of chemical compounds in a sample, wherein said method comprises:
In one embodiment, said layered separating medium is a layered cellulose paper or layered nitrocellulose film. The layered separating medium may be optionally patterned. Non-limiting examples of the layered separating medium are origami or kirigami folded layer-by-layer paper or film with embedded electronics. In a specific embodiment, the origami or kirigami paper comprises a quality paper in a flexible and foldable origami or kirigami format.
In a certain embodiment, the sensors are screen-printed or inject-printed on each layer of said separating medium. In some embodiments, the physico-chemical separation is a chromatographic separation.
Non-limiting examples of said detector are a micro-gas chromatograph, miniaturised dispersive optical spectrometer, fibre-coupled optical spectrometer, micro-electromechanical system (MEMS)-based spectrometer, plasmon-enhanced Raman spectrometer, on-chip plasmonic spectrometer, piezoelectric crystal detector, and a spin-induced mass spectrometer. Said detector may further comprise one or more microfabricated components. The microfabricated components are selected from capillary or chip-based microcapillary separation columns, a source of carrier gas, pre-concentrator-injector, micro- and/or nano-optical components, MEMS components, microfluidics components, pumps, filters, and valves. The detector may yet further comprise hardware and software for instrument control, data acquisition, and analysis.
Non-limiting examples of the sensors are:
In a particular embodiment, said sensors further comprise at least one chemical or biomolecular layer immobilised on top of said sensors and capable of binding or adsorbing said compounds from the sample. Said at least one chemical layer may comprise chemical functional groups selected from amines, alkenes, alkynes, phosphines, azides, cycloalkenes, cycloalkynes, cyclopropanes, isonitriles, vinyl boronic acid, tetrazine, maleimide, alcohols, thiols, conjugated dienes, copper acetylide, nitrones, aldehydes, ketones, alkoxyamines, hydroxylamine, hydrazine, hydrazide, isothiocyanate, carbodiimide, and carboxylic acids or derivative thereof, such as esters, anhydrides, N-hydrosuccinimide (NHS), tosyl and acyl halides. Said at least one chemical or biomolecular layer may specifically comprise cyclodextrin, 2,2,3,3-tetrafluoropropyloxy-substituted phthalocyanine or derivatives thereof, or said chemical or biomolecular layer comprises capturing biological molecules, such as primary, secondary antibodies or fragments thereof against certain proteins to be detected, or their corresponding antigens, enzymes or their substrates, short peptides, specific polynucleotide sequences, which are complimentary to the sequences of DNA to be detected, aptamers, receptor proteins or molecularly imprinted polymers. Selectivity of the sensors can be altered by changing the chemical identity of said at least one chemical or biomolecular layer.
In a specific embodiment, the compounds contained in the sample are selected from:
In a further embodiment, the external memory is a mobile device, wearable gadget, smartphone, smartwatch, desktop computer, server, remote storage, internet storage or internet cloud. The external memory may also comprise a processor, a microcontroller or a memory-storing controller suitable for storing executable instructions, which when executed by the processor cause the processor to perform the machine-learning method on the measurement results.
In yet further embodiment, said string or said array of measurements from each said sensor on each said layer of the layered separating medium are generated from a spectrogram or chromatogram of the sample.
In another embodiment, the machine-learning method is suitable for employing a neural network selected from a fully connected neural network, a convolutional neural network, a recurrent neural network, a ResNet neural network and a neural network with attention heads. The machine-learning method may further comprise training of a neural network, on which said method is employed. In still another embodiment, the input is further treated with a wavelet transform with the different number of hidden long short-term memory layers (LSTM) of a neural network.
In another aspect of the present invention, a method for inkjet-printing of electronics and sensors on each layer of said layered separating medium comprises applying inkjet-print grade graphene oxide heterostructures to the layers of the layered separating medium and reducing said graphene oxide heterostructures with dopamine to obtain reduced graphene oxide (rGO) heterostructures. This method does not use any aggressive chemicals or reducing agents. The reducing agent in this method is dopamine, which forms poly-dopamine through self-polymerisation during the reduction reaction of GO and is thereby also suitable for use as a capping agent. Catechol groups of poly-dopamine-coated rGO are suitable for oxidation to a quinone form in weak alkaline condition and further for functionalisation with thiol or amine containing compounds either by Michel addition or Schiff-based mechanism.
The present invention describes specific embodiments of a method for processing micro-gas chromatography data of a sample containing compounds subjected to separation on an origami or kirigami paper, said data generated with a micro-gas chromatograph comprising an array of chemical sensors inkjet-printed on each layer of said origami paper, and providing information on the presence and properties of said compounds in the sample, wherein said method comprises:
In a further aspect, a detector for processing spectrometry data of a sample containing compounds subjected to a chromatographic separation on a layered separating medium, comprises an array of sensors printed on each layer of said layered separating medium, and providing information on the presence and properties of said compounds in the sample. The detector is suitable for generating chemical and physical data on detection and separation of tested chemical compounds from a sample containing said compounds on the layers of said layered separating medium. In a certain embodiment, the detector is configured to generate a string or an array of measurements from each said sensor on each said layer of the layered separating medium. The detector may further comprise one or more microfabricated components and/or embedded electronics. Non-limiting examples of said microfabricated components are capillary or chip-based microcapillary separation columns, a source of carrier gas, pre-concentrator-injector, MEMS components, micro- and/or nano-optical components, microfluidics components, pumps, filters, and valves. The detector may further comprise hardware and software for instrument control, data acquisition, and analysis. In a specific embodiment, the detector is selected from a micro-gas chromatograph, miniaturised dispersive optical spectrometer, fibre-coupled optical spectrometer, MEMS-based spectrometer, plasmon-enhanced Raman spectrometer, on-chip plasmonic spectrometer, piezoelectric crystal detector, and spin-induced mass spectrometer. In a particular embodiment, the detector is embedded in the layered separating medium. In some embodiments, the detector is in a form of a paper card with embedded electronics.
In yet further aspect of the present invention, a system for detection and analysis of chemical compounds in a sample subjected to a chromatographic separation on a layered separating medium, said system comprises:
In some embodiment, communication between the sensors and the external memory of the system of the present invention is either passive or active, or combination thereof. When the communication between the sensors and the external memory is passive, the system is configured to perform a spectral encoding of information using a single radiative structure with multiple resonators each of which is dedicated either to a bit encoding or to a sensor readout. When the communication between the sensors and the external memory is active, the system is configured to carry out a parallel route for powering and communicating between the sensors and external memory using a semiconductor device.
The aforementioned embedded electronics may comprise at least one of the following components:
The connection module may be wireless for wireless connection of said system with the external memory. The external memory may comprise another wireless connection module connecting said system to a user interface via a digital-to-analogue converter (DAC). In some embodiments, both wireless connection modules are either Bluetooth or NFC, thereby providing wireless communication between the sensors and the readout module for up to 20 m. If these two modules are Wi-Fi, the connection between them can be established for up to 200 m, while the GSM may provide the worldwide communication.
In another embodiment, the system of the present invention further comprises at least one of: (i) a feedback control microcontroller unit (MCU) for energy level adjustment and de-trapping via an external or integrated gate electrode; (ii) a harvester for harvesting energy of the system; (iii) a power management unit (PMU) for transforming said harvested energy and powering an analogue read-out of the sensors; (iv) an analogue front-end; and (v) a gate electrode for discharging parasitic electric current. In a particular embodiment, the system further comprises a remote powering with miniaturised receiver antenna.
The system may further comprise at least one radio-frequency identification (RFID) out-input tag for remote readout and zero-power operation, each RFID tag connected to said embedded electronics via an electric circuit for receiving or transmitting a signal. In a particular embodiment, this RFID system further comprises:
In still another embodiment, the external memory is a mobile device, wearable gadget, smartphone, smartwatch, desktop computer, server, remote storage, internet storage or internet cloud. The external memory may comprise a processor, a microcontroller or a memory-storing controller suitable for storing executable instructions, which when executed by the processor cause the processor to perform the machine-learning method on the measurement results.
Various embodiments may allow various benefits and may be used in conjunction with various applications. The details of one or more embodiments are set forth in the accompanying figures and the description below. Other features, objects and advantages of the described techniques will be apparent from the description and drawings and from the claims
Disclosed embodiments will be understood and appreciated more fully from the following detailed description taken in conjunction with the appended figures. The drawings included and described herein are schematic and are not limiting the scope of the disclosure. It is also noted that in the drawings, the size of some elements may be exaggerated and, therefore, not drawn to scale for illustrative purposes. The dimensions and the relative dimensions do not necessarily correspond to actual reductions to practice of the disclosure.
In the following description, various aspects of the present application will be described. For purposes of explanation, specific configurations and details are set forth in order to provide a thorough understanding of the present application. However, it will also be apparent to one skilled in the art that the present application may be practiced without the specific details presented herein. Furthermore, well-known features may be omitted or simplified in order not to obscure the present application.
The term “comprising”, used in the claims, is “open ended” and means the elements recited, or their equivalent in structure or function, plus any other element or elements which are not recited. It should not be interpreted as being restricted to the means listed thereafter; it does not exclude other elements or steps. It needs to be interpreted as specifying the presence of the stated features, integers, steps or components as referred to, but does not preclude the presence or addition of one or more other features, integers, steps or components, or groups thereof. Thus, the scope of the expression “a device comprising x and z” should not be limited to devices consisting only of components x and z. Also, the scope of the expression “a method comprising the steps x and z” should not be limited to methods consisting only of these steps.
Unless specifically stated, as used herein, the terms “about” and “approximately” are understood as within a range of normal tolerance in the art, for example within two standard deviations of the mean. In one embodiment, the term “about” means within 10% of the reported numerical value of the number with which it is being used, preferably within 5% of the reported numerical value. For example, the term “about” can be immediately understood as within 10%, 9%, 8%, 7%, 6%, 5%, 4%, 3%, 2%, 1%, 0.5%, 0.1%, 0.05%, or 0.01% of the stated value. In other embodiments, the term “about” can mean a higher tolerance of variation depending on for instance the experimental technique used. Said variations of a specified value are understood by the skilled person and are within the context of the present invention. As an illustration, a numerical range of “about 1 to about 5” should be interpreted to include not only the explicitly recited values of about 1 to about 5, but also include individual values and sub-ranges within the indicated range. Thus, included in this numerical range are individual values such as 2, 3, and 4 and sub-ranges, for example from 1-3, from 2-4, and from 3-5, as well as 1, 2, 3, 4, 5, or 6, individually. This same principle applies to ranges reciting only one numerical value as a minimum or a maximum. Unless otherwise clear from context, all numerical values provided herein are modified by the term “about”. Other similar terms, such as “substantially”, “generally”, “up to” and the like are to be construed as modifying a term or value such that it is not an absolute. Such terms will be defined by the circumstances and the terms that they modify as those terms are understood by those of skilled in the art. This includes, at very least, the degree of expected experimental error, technical error and instrumental error for a given experiment, technique or an instrument used to measure a value.
As used herein, the term “and/or” includes any and all combinations of one or more of the associated listed items. Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the specification and relevant art and should not be interpreted in an idealized or overly formal sense unless expressly so defined herein. Well-known functions or constructions may not be described in detail for brevity and/or clarity.
It will be understood that when an element is referred to as being “on”, “attached to”, “connected to”, “coupled with”, “contacting”, etc., another element, it can be directly on, attached to, connected to, coupled with or contacting the other element or intervening elements may also be present. In contrast, when an element is referred to as being, for example, “directly on”, “directly attached to”, “directly connected to”, “directly coupled” with or “directly contacting” another element, there are no intervening elements present. It will also be appreciated by those of skill in the art that references to a structure or feature that is disposed “adjacent” another feature may have portions that overlap or underlie the adjacent feature.
The term “compound/s”, “tested compound/s” or “chemical compound/s” used in the present application are entirely equivalent and encompass virtually any chemical and biomolecular compound, including but not limited to organic compounds, volatile organic compounds (VOCs), biochemicals, etc. By common definition, “chemical compound is any substance composed of identical molecules (that can be biomolecules as well) consisting of atoms of two or more chemical elements”. Non-limiting examples of chemical compounds being tested in the present invention and then included in a library of micro-GC patterns of the present invention are:
The present invention relates to design and examination of spatiotemporal nano/micro-structural arrangement that is inspired by origami, kirigami and also bio-inspired by a butterfly wing for endowing real-time and miniaturised sensors with capabilities of miniaturised spectrometers towards complex structural and chiral molecules.
In one aspect, the present invention directs at a method and system for detection and analysis of chemical compounds in a sample, wherein said method comprises:
Throughout the description, the terms “spintronics spectroscopy” or “spintronics spectrometry” are used to refer to any spectroscopic technique or instrument, relatively, based on measuring the intrinsic spin of electron and its associated magnetic moment, in addition to its fundamental electronic charge, in solid-state devices. The field of spintronics spectroscopy concerns spin-charge coupling in physical systems. Spintronics fundamentally differs from traditional electronics in that, in addition to charge state, electron spins are exploited as a further degree of freedom, with implications in the efficiency of data storage and transfer. Spintronic systems are of particular interest in the field of quantum computing and miniaturised spectroscopic devices, for example, nano-nuclear magnetic resonance spectrometer or spin-induced mass spectrometer.
Throughout the description, the terms “MEMS-based spectroscopy” or “MEMS-based spectrometry” are used to refer to any spectroscopic technique or instrument, relatively, that uses a special discrete-time measuring principle which makes it possible to scan a spectrum with a single highly sensitive detector only by the rotational movement of the integrated micro-electro-mechanical system (MEMS) grating.
Throughout the description, the term “piezoelectric crystal detector” refers to the detector that detects the mass of chemical vapours absorbed into chemically selective coatings. It comprises an array of surface acoustic wave (SAW) sensors designed to detect changes in propagation characteristics of acoustic waves near the surface of a piezoelectric material, such as quartz.
In one embodiment, the layered separating medium is a layered cellulose paper or layered nitrocellulose film. The layered separating medium may be optionally patterned, and may be in a form of origami or kirigami, as will be explained below. In some embodiments, the physico-chemical separation is a chromatographic separation as will also be explained below.
In a particular embodiment, a method for processing micro-gas chromatography data of a sample containing chemical compounds subjected to separation on an origami or kirigami paper, said data generated with a micro-gas chromatograph comprising an array of sensors inkjet-printed on each layer of said origami or kirigami paper, and providing information on the presence and properties of said chemical compounds in the sample, comprises:
Throughout the description, the term “micro-gas chromatograph” (μGC) is used to refer to any field portable versions of a GC comprising one or more microfabricated components, selected from capillary or microcapillary separation columns, a source of carrier gas, pre-concentrator-injector, detector, pump, valves, and filters. The μGC may further comprise software for instrument control, data acquisition, and analysis.
The external memory according to the present invention can be, for example, a mobile device, wearable gadget, smartphone, smart watch, desktop computer, server, remote storage, internet storage, or internet cloud.
In the present invention, the inventors have developed all hybrid electronic sensing system printed on a layered separating medium suitable for separating complex mixtures, for example paper, cellulose or nitrocellulose film, in a flexible origami or kirigami format to conduct spectroscopic measurements in time space resolved hierarchical manner. The developed strategy shows its huge potential for simultaneous separation and discrimination of various complex mix organic compounds (for example, alcohols, ketones, aldehydes, organic acids, hydrocarbons) using layer by layer approach by modulating the mass transport phenomena of wide range VOCs.
In conventional sensor/sensor array measurements, each sensor in the array exhibits an exponential transient reaction kinetics which is the superposition of the response profile of multiple VOCs present in the mix condition, thus, it is challenging for analysing the VOC contents when the number of VOCs become higher. In this context, time-space resolved approach like a gas chromatography, using printed origami or kirigami architecture, is highly useful for analysing complex VOC mixtures in real applications for performing direct micro-gas chromatography like measurements in skin or other physiological samples in wearable format.
In the design of the present invention shown in
In the present invention, the inventors have optimised inkjet-print grade reduced graphene oxide (rGO) heterostructures without using any aggressive chemicals or reducing agents (such as, hydrazine, dimethyl hydrazine, hydroquinone, and NaBH4), stabiliser or binder (for example, Triton x-100, pyrene sulphonic acid, Xanthan gum, ethylene/propylene glycol). In this context, a mussel-inspired bio-adhesive dopamine, a surfactant and reducing agent, has been used to simultaneously reduce the GO and to use as a capping agent by forming the poly-dopamine through self-polymerisation during the reduction reaction of GO for wide range of substrate.
Dopamine is biochemically signified as hormone and neurotransmitter inside the human body and it transforms to poly-dopamine through self-polymerisation in weak alkaline medium. The catechol groups of poly-dopamine coated rGO could be oxidised to quinone form in weak alkaline condition and further could be used to functionalize with thiol or amine containing compounds either by Michel addition or Schiff based mechanism. Then the powder (DrGO) is dried in vacuum for functionalisation. This versatile reaction pathway has been used to nano-engineer the grafting process for tuning the surface chemistry of graphene plane with variety of multiple ligands for constructing inkjet-printed grade sensor array which is connected through printed silver nanowires prepared with a rapid solvothermal chemical route.
The present rGO or rGO-ligand ink formulations do not require any stabilising agent or binder (thus no post printing annealing is required for surfactant/binder removal) and could be readily dispersed in various organic solvents to print the different geometry on rigid, flexible and porous substrates (for example, Si/SiO2, Kapton, glass, PET, silicone rubber and various types of paper) by simply modulating the printer parameters for film deposition in layer-by-layer approach.
To get a stable suspension of graphene ink for uniform printing geometry without nozzle blockages, various attempts have been reported. These include use of various stabiliser/surfactant (such as, Triton or pyrene sulfonic acid) for very good dispersion of the ink, adjustment of desired viscosity of the ink without any coffee ring effect, by suitably chosen dispersion medium (such as, terpineol, NMP or alcohol), surface tension matching between substrate and ink droplet by tuning hydrophobicity (such as, spin coating of bis(trimethylsilyl)amine (HMDS) or hydrophilicity (plasma) of the surface, use of binders and additives (such as, PVP, PEDOT: PSS or ethyl cellulose) for improvement of the post printing adhesion between the ink and the substrate. Although the use of stabiliser promotes better graphene stabilisation, removal of the excess of the stabiliser from the ink for getting the better conductivity is tedious. Similarly, use of binder/additives may promote better adhesion for wide range substrate, but again post-printing thermal annealing is needed to remove extra adhesives, and this high temperature treatment is not suitable for various flexible substrate like paper/plastic. In this scenario, dopamine could be a better alternative as this specific material could be used as direct solvent mediumted reducing agent to remove oxygen containing functional group from graphene oxide (GO) to convert into reduced graphene oxide (rGO), as well as it serves as surfactant through the formation of polydopamine by in-situ polymerization on rGO surface.
The obtained hybrid system (rGO-polydopamine (DrGO)) is stable in water and various organic solvents (such as, dimethyl formamide (DMF), chloroform) and readily available for inkjet printing without using any additional surfactant, thereby, no need for post-printing high temperature annealing. In addition, due to the presence of catechol group inside the DrGO nano hybrid, the adhesion property is very similar to natural adhesive mussel protein (enriched with amines and catechol group) that has higher affinity to bind with wide range organic/inorganic substrates, such as metal/metal oxides, polymers. This additional inherent adhesion property of the DrGO hybrid could be an excellent alternative for an additive/binder-free printing technology without using any tedious viscosity adjustments, such as, using high boiling point solvent, thus, suitable for printing with wide range substrate application.
As prepared cleaned and dried, DrGO powders are then used for grafting reaction to bridge various thiols, amines, and chiral compounds to make FDrGO powders. The functionalised powder (FDrGO) is further redispersed in DMF and sonicated at low power for dispersion of the ink for about 1-2 minutes and subsequently used for inkjet printing on paper.
Table 1 lists thiol and amine terminated various achiral and chiral ligands used for synthesis of the functionalised hybrid multifunctional ink (FDrGO):
Table 2 provides a comparative printing approach for the binder- and surfactant-free printing of the present invention with the state-of-art. It shows the benchmarking of the chemicals and components used for traditional inkjet-printed graphene-based materials and for the binder- and surfactant-free printing of the present invention:
In the printing process of the present invention shown in
An exemplary microprocessor-based computational unit is Mega 2560 IC which is suitable for interfacing sensor pixels in a credit card sized printed paper as described above. Schematic of the microprocessor embedded μGC design with I2C inter connection is shown
The sensors used in the present invention are any suitable sensors, which can be used with the detector of the present invention, such as micro-gas chromatograph, miniaturised dispersive optical spectrometer, fibre-coupled optical spectrometer, MEMS-based spectrometer, plasmon-enhanced Raman spectrometer, on-chip plasmonic spectrometer, piezoelectric crystal detector, or spin-induced mass spectrometer, capable of sensing chemical compounds. Non-limiting examples of these sensors are:
The origami-based sensors layout has advantages of long-range printing capability in mass scale (printing) and synthesis of inkjet grade ligand modified graphene on paper, PET, PDMS etc. Origami/kirigami based paper platforms are getting attention in various applications, such as, sensor, robotics, microfluidics, battery, tissue engineering, microscopy, and artificial muscles. These self-folding structures are emerging at the recent interest due to their capacity to perform certain programmed folding/unfolding motions, 2D planar to 3D transition by manipulating layers, easy integration of printed features in its layer by layer to show multi-functionality in one system. In particular, for wearable electronics and sensors, this integrated platform is highly appealing.
Microscopic (SEM), spectroscopic (Raman, FTIR, NMR), electronic (FET and I-V) data on GO, DrGO, thiol-FDrGO, amine-FDrGO and chiral-DrGO were collected to find morphological, structural, functionalisation status and basic electronic quality of the materials of the present invention.
N-type semiconducting properties of DrGO/FDrGOs suggest integration of N2 adatom at graphitic plane during GO-reduction with polydopamine. This has been proved by the DFT analysis of the FDrGO (with 4-mercaptobenzoic acid as a ligand) interaction with S (+)/R (−)-butanol. The DFT analysis was performed using Gaussian 16 software to optimise the molecular structure, as well as the HOMO and LUMO energy of FDrGO with 4-mercaptobenzoic acid) as well as their complexes R- and S-2-butanol molecules. The DFT method was combined with Austin-Frisch-Peterson functional with dispersion (APFD) and the basis set 6-31G (d), including treatments of dispersion effects, which represents the best trade-off between accuracy and computational cost for a relatively large system. The corresponding modelling images are available upon request.
To explain further, the spin polarisation effect of the modified band structure of graphene due to presence of doping or defect was considered. In general, graphene is diamagnetic. However, it shows magnetic behaviour when some foreign atom or defect is present in C—C hexagonal lattice. Due to doping of other element (e.g., H2, N2, F), time reversal symmetry breaks and due to orbital overlap of p electron, the local crystal structure exhibit finite magnetic moment. This make graphene magnetic, which is highly desirable for spintronics and chiral recognition.
In general, during the chiral based charge transfer, the interaction is also accompanied by spin injection which is completely opposite to each enantiomer type. This phenomenon is commonly termed as “chiral induced spin selectivity”. Using this unique spin injection, the surface become spin polarised (i.e., majority spin direction is either up or down). Due to the specific spin population, energy band of graphene structure modulates (increased or decreased). This band splitting phenomena is also predicted by quantum mechanical DFT when graphitic structure is doped with various nitrogen-based ligand.
In the present invention, during the reduction reaction of GO and polymerisation reaction of polydopamine, the oxygen containing group from C—C lattice is removed and N-based compound from polydopamine chain is impregnated in C—C lattice. The presence of N2 adatom breaks time and space reversal symmetry, possesses finite magnetic moment, opens the Dirac band gap and becomes n-type.
The DFT analysis for two typical extreme case (considering all the spin configuration in the N-site is either all up or all down) and calculated band gap difference between these two cases show that energy difference is indeed different in magnitude. This distinct band modulation (increase or decrease of band gap) due to specific spin population and polarisation on the surface exhibits opposite directional resistance change (increase/decrease of resistance) for different enantiomer type using CISS mechanism. This result shows a high potential for the next generation spintronics, spin-based logic and spin-based memory/magnet free storage device applications.
Reference is now made to
For evaluating performance of the detectors of the present invention, a single layer including functionalised sensors array was examined upon direct exposure to various chemical structural/chiral mixtures, without spatiotemporal part (see
Reference is made to
As seen in figure, the direct exposure provides superposed exponential response/recovery kinetics of overall mixture, but no further information could be obtained on each alcohol component in mixture. This superposed response could originate from infinitely possible ratios of mix components that would generate every time same exponential superposed response kinetics that is contributed from each of the component in mixture. This issue would be more challenging when number of components (analytes) in the mixture become larger. On top of that, it will be very hard mission to train intelligent system for infinitely possible combination of ratios to predict from real sample analyses. The method of the present invention changes the result dramatically.
Thus, the major issue with a conventional sensor array and direct exposure is a difficulty to separate compounds having a similar structure and molecular weight (see
The principal notion for utilising layered medium together with the detectors of the present invention in the VOC sensing is to utilise the unique time-space resolved geometry for effective semi-separation and instantaneous detection of various VOCs from a complex mixture in different instant of time. Since each VOC exhibit different mass transport phenomena (due to their distinct molar weight) while passing through porous origami paper layer, it is reasonable to expect a non-linear resistance profile (with multiple peaks at different moment of time) from sensors in each origami layer (see
In conventional measurements, each sensor in the array exhibits an exponential transient reaction kinetics which is the superposition of the response profile of multiple VOCs present in the mix condition, thus, it is challenging for analysing the VOC content when the number of VOCs become higher. In particular, for real human physiological samples from breath, blood, urine and skin, the confounding VOCs and background VOCs are very large in number. In this context, time-space resolved approach like a gas chromatography, using printed origami architecture, would be highly useful for analysing complex VOC mixtures in real applications for performing direct micro-gas chromatography like measurements in skin or other physiological samples in wearable format.
Reference is now made to
As expected in the first layer, sensing results showed no separation. However, already for the second and third layer, response kinetics shows three distinct peaks with a clear realisation of the μGC time-space-resolved-architecture. This simultaneous separation and sensing capability are originated from porous layered structure of the cellulose film used in this experiment within the μGC architecture which modulates unique mass transfer rate of each molecule in the mixture and creates distinct time to reach each layer, thereby creating three major peaks and area: for methanol, ethanol and isopropanol.
To confirm identification of each peak, sensing of additional vapour combinations was examined for different ratios (1:1:2, 2:1:1, 1:1:2), as shown in
As mentioned above, the Gaussian distribution was used for fitting the peaks from mix gas analyses data to extrapolate the VOC component position. However, the Gaussian distribution is symmetric in nature. This means the response and recovery kinetics should be the same. Although it is easier to find fast discrimination of the peaks, in practice this is not applicable. The distribution of each VOC profile in mix condition is asymmetric in nature due to different adsorption/desorption energy.
To represent entire VOC kinetics, the inventors have developed a new equation that can represent the real-time picture. In continuous flow, it is assumed that during the adsorption and desorption, sensor resistance should reach a maximum point tmax that is defined as the “maximum availability” of a specific VOC on a certain origami paper layer.
Based on the Langmuir absorption kinetics, the response-recovery kinetics is modelled, as shown in
where ka and kd are the adsorption and desorption constants, Pvoc is the partial pressure of a volatile organic compound to be tested, and φ(t) is a fractional occupancy of the adsorption sites or the surface coverage which is linearly proportional to the sensor's conductance G (t).
Reference is now made to
where G (t) is continuous measurement of sensor resistance at a certain time (t), tmax is time of electrical resistance maxima representing for a specific analyte and different for different analytes, and a, b, c, d, e are respective kinetical constants, adsorption, and desorption parameters.
In the present case, some specific gas molecule enters a specific layer, react with sensor (response), leaves (recovery) and provides a unique time moment (tmax) where the electrical resistance reaches maximum. To express mathematically the entire phenomena, the entire electrical resistance transients should be the superposition.
Thus, in the method of the present invention, due to continuous flow, the adsorption/desorption is simultaneously active and therefore, the entire conductance transients should be the superposition of both contributions. This is markedly different from constant exposure where response and recovery never overlap. In that case, only after the sensor reached an equilibrium in constant VOC exposure during the response, the recovery begins. But in the present invention, due to continuous flow, both adsorption and desorption is active and therefore, the resultant equation should be written as:
Due to difference in adsorption and desorption energy (c and e in the above equation), fitted sensing profile is asymmetric and broad (see
In this case, G (t) represent a maximum where the probability for finding the VOC is also maximum (see
Solving the above equation returns the following value of tmax:
Adding the tmax value into the above G (t) total equation gives the following expression:
From the above equation, it becomes clear that tmax corresponds to each VOC peak in the continuous micro-GC chromatogram and depends on the physico-chemical properties of each tested compound (VOC), where each VOC is eluted at different time and creates separate tmax for each VOC. Thus, the resultant equation for the multi-VOC mixture represents the superposition of each component (VOC1, VOC2, . . . . VOCn) (see
The method of the present invention was further challenged with more complex mixtures of 24 analyte types (alcohols, aldehydes, ketones, hydrocarbons and organic acids) Reference is now made to
The mixing ratio in the above experiment was 10:1. For example:
All complex data set was used as a sequence input to machine learning for generating the predicted continuous chromatogram or spectrogram of typical mix samples MVOC1-12. Details of input-generation, hidden-layer optimisation and data-compressing using wavelet-based signal processing are provided below.
The Layer 2 of the layered medium-derived sensing profile of the VOCs mixes of 24 types is thus analysed through machine learning and wavelet-based signal processing to generate the chromatogram or spectrogram.
Machine-learning neural network architecture is constructed by MATLAB program with a typical sequence input layer which is fed from the time space resolved data (read from Layer 2). This specific time-resistance sequence is further processed to LSTM layer and fully connected layers. Then finally a regression layer is added in the end of the process to generate the chromatogram or spectrogram for continuous prediction. Due to large data set, the wavelet-based signal processing (with Haar Level 3 decomposition) is used in order to compress the time sequence data (N) from the approximated coefficients values (N/2), and then used for deep neural network's sequence input layer.
For enantiomers and isomers mix state, the image processing is used by synthetically constructed input image from Layer 2 μGC-derived data and fed to the self-learning architecture of the deep net layers, which has the subsequent consecutive layers, such as, image-input layer, a 2-D convolutional layer (“convolution2dLayer” operation in MathWorks®), rectified linear unit (ReLU) layer (“reluLayer” operation in MathWorks®), a 2-D max pooling layer (“maxPooling2dLayer” operation in MathWorks®), fully connected layer, classification layers. Convolutional and batch normalisation layers are usually followed by a nonlinear activation function such as a rectified linear unit (ReLU), which is specified by a ReLU layer. The ReLU layer performs a threshold operation to each element of the input, where any value less than zero is set to zero. Before the training, each image input sequence has been resized with same dimension. Deep net system automatically samples the features from an image by itself and uses them for learning to classify the various mixed enantiomers and isomers.
The predicted result does not match properly with the target ratio and could be due to large time stamp and data information and need to optimise the hidden layer information with longer training time. To solve this, the μGC-derived input large data set is further compressed with wavelet transformation without losing important information and then used for deep net input as described above.
The abbreviation “LSTM layers” stand for the long short-term memory layers of a neural network. Recurrent neural networks (RNNs) process inputs in a sequential manner, where the context from the previous input is considered when computing the output of the current step. This allows the neural network to carry information over different time steps rather than keeping all the inputs independent of each other. However, a significant shortcoming that plagues the RNN is the problem of vanishing/exploding gradients. This problem arises when back-propagating through the typical RNN during training, especially for networks with deeper layers. The gradients have to go through continuous matrix multiplications during the back-propagation process due to the chain rule, causing the gradient to either shrink exponentially (vanish) or blow up exponentially (explode). Having a gradient that is too small prevents the weights from updating and learning, whereas extremely large gradients cause the model to be unstable. Due to these issues, RNNs are unable to work with longer sequences and hold on to long-term dependencies, making them suffer from “short-term memory”. The LSTMs come to solve this problem
While LSTMs are a kind of RNN and function similarly to traditional RNNs, its gating mechanism is what sets it apart. This feature addresses the “short-term memory” problem of RNNs.
On the other hand, the LSTM layers can retain the earlier information of all the previous time stamps, and this will aid the model in accurate determination of compounds in complex mixtures, even when the compounds are structurally very similar, such as isomers (see examples below). The LSTM is capable of generating the input at any time stamp due to the contextual information from a much earlier time stamp. This advantage of the LSTM actually lies in its gating mechanism within each LSTM cell. In the normal RNN cell, the input at a certain time stamp and the hidden state from the previous time step is passed through a hyperbolic tangent (tanh) activation function to obtain a new hidden state and output. On the other hand, the LSTM cell at each time stamp takes in three different pieces of information, which are the current input data, the short-term memory from the previous cell (similar to hidden states in RNNs) and lastly the long-term memory. The short-term memory is commonly referred to as the hidden state, and the long-term memory is usually known as the cell state. The cell then uses gates to regulate the information to be kept or discarded at each time step before passing on the long-term and short-term information to the next cell. These gates can be seen as filters. Ideally, the role of these gates is supposed to selectively remove any irrelevant information. At the same time, only desired data can pass through these filters, just like how the gates only hold on to the useful information. Of course, these gates need to be trained to accurately filter what is useful and what is not. These gates are called the Input Gate, the Forget Gate, and the Output Gate. There are many variants to the names of these gates; nevertheless, the calculations and workings of these gates are mostly the same.
In the present invention, a chemical or biological compound generates a spectrogram or chromatogram which is read by the detector, for example μGC, miniaturised dispersive optical spectrometer, fibre-coupled optical spectrometer, MEMS-based spectrometer, plasmon-enhanced Raman spectrometer, on-chip plasmonic spectrometer, piezoelectric crystal detector, or spin-induced mass spectrometer. As mentioned above, each VOC, as any chemical or biological compound, has its own unique fingerprint (pattern) in a generated library and can be served as a layer input data from a layered separating medium, resulting in a ‘0’ if there is no compound in the mixture, or a ‘1’ if there is a compound in the mixture. Alternatively, these could be probed, for example by a regular GC chromatography or mass spectrometry readout resulting in a binary result as well. The readout could be done via a series of measurements resulting in a sum of the ones and zeros from all the detections. Therefore, in the method of the present invention, an input is either an array of binary values or an array of integers generated by the series.
An “output” of a neural network used in the method of the present invention is defined as a single bit whose value is ‘0’ or ‘1’, or an array of bits, or an array of integers, or an array of complex numbers, wherein said single bit, or said array of bits, or said array of integers, or said array of complex numbers corresponds to an estimated frequency, voltage, sensor conductance, electrical resistance, or time of chromatographic elution, mass-to-charge ratio, nuclear magnetic resonance shift, and to maximum availability or amplitude of the input.
As mentioned above, a neural network of the present invention is trained using a training dataset of inputs with known fingerprints or patterns. During the training, the parameters of the neural network are optimised to output the correct fingerprints or patterns for the inputs in the training dataset. The goal of the training is thus to make the neural network learn the general relation between the inputs and outputs such that it would be able to output the correct result of a known input with the highest possible probability.
The machine-learning method of the present invention is able to overcome the lack of knowledge of the physical model under supervised learning. As noted above, the objective is to use a train data set that contains inputs together with their known true patterns (true outputs) in order to train a deep neural network such that the trained neural network is an optimised function, which outputs the correct results for new inputs with the optimal (or near optimal) probability.
The problem of discrimination between at least two signals corresponding to at least two different time stamps by the detector of the present invention, for example μGC, miniaturised dispersive optical spectrometer, fibre-coupled optical spectrometer, MEMS-based spectrometer, plasmon-enhanced Raman spectrometer, on-chip plasmonic spectrometer, piezoelectric crystal detector, or spin-induced mass spectrometer, is solved in the present invention. In accordance with the present invention, to overcome the model's lack of knowledge, a supervised machine-learning model is used. In this model, a train dataset of measurement results of known patterns from a pre-generated library is used to train the neural network. The trained network is then applied to a test dataset and results in estimations of the time stamps of the test measurement results.
In a further aspect of the present invention, a system for performing a physico-chemical analysis of a sample containing compounds subjected to a physico-chemical separation, said analysis is performed by processing spectrometry data of the sample, said system comprises:
The aforementioned embedded electronics may comprise:
In a particular embodiment, the system of the present invention further comprises a remote powering with miniaturised receiver antenna. Inclusion of a battery for powering the system would affect the overall size and would cause several economic and environmental problems due to the toxic nature of battery chemicals. Thus, the system of the present invention is optionally based on sensing nodes that use the energy provided by incoming electromagnetic waves to ensure the read-out of the sensors and the transmission of the resulting data to the reader node. The major problem solved by the present inventors in this regard relates to the size of the antenna, which affects both the coupling and the selection of carrier frequencies. The coupling decreases with the reduction of antenna size, while the optimal carrier frequency increases. Only high carrier frequencies can permit the miniaturisation of the antenna, but the propagation losses also increase with the frequency which requires both careful selection and EM modelling.
When considering an implementation, the present inventors proposed that the harvested energy would require transformation to power the analogue read-out of sensors. This is typically handled by a PMU (Power Management Unit) but the extremely low received energy would challenge the use of existing solutions. A tightly combined implementation of these components in the system: harvester, PMU and analogue front-end is another embodiment of the present invention that allows to reach performances beyond state of the art.
In still another embodiment, the system of the present invention further comprises at least one of: a feedback control microcontroller unit (MCU) for energy level adjustment and de-trapping via an external or integrated gate electrode, a harvester for harvesting energy of the system, a power management unit (PMU) for transforming the harvested energy and powering an analogue read-out of the sensors, an analogue front-end and a gate electrode for discharging parasitic electric current. The system may further comprise at least one radio-frequency identification (RFID) out-input tag for remote readout and zero-power operation, each RFID tag connected to said embedded electronics via an electric circuit for receiving or transmitting a signal. In a particular embodiment, this RFID system further comprises:
In some embodiments, the external memory is a mobile device, wearable gadget, smartphone, smartwatch, desktop computer, server, remote storage, internet storage or internet cloud. The external memory may comprise a processor, a microcontroller or a memory-storing controller suitable for storing executable instructions, which when executed by the processor cause the processor to perform the machine-learning method on the measurement results.
In other embodiments, communication between the sensors and the external memory can be either passive or active, or combination thereof. In the passive read-out of the sensors, the system uses a spectral encoding of the information that is the route of choice, since a passive structure cannot handle any kind of communication protocol. The necessary information (RFID and sensors readings) is encoded in the passive mode using a single radiative structure with multiple resonators each of which is dedicated either to a bit encoding or to a sensor readout.
In case of the active communication, a parallel route is carried out to power and communicate with the system using a semiconductor device. In order to do that, a miniaturised die (25 μm thick, inert and non-toxic) on the substrate and the management of the energy are integrated in the system. Importantly, this approach also enables operation of the sensor elements at optimum operating frequencies (DC to kHz). The antenna and RF propagation channel are viewed jointly and closely co-designed with the materials.
In the above approach, the number of sensing nodes accessed (millions) lies orders of magnitudes beyond the state of the art (thousands) found nowadays in the conventional UHF RFID solutions, in which many tags can be seen almost simultaneously by a user. Spatial, time and frequency multiplexing are considered together with data encoding and modulation schemes to push beyond the limits of the state-of-the-art. Depending on the architectural implementation (passive or active) of the sensing node of the system, the communication scheme is different. When the complexity (intelligence) of the sensing node is limited as in a passive mode, specific distributions of sensing node configurations (by construction) are used to reduce overlaps in time or frequencies. This requires the design of complementary separation algorithms pushing back the complexity at the reader level.
Due to the extreme miniaturisation of the system, the returned signal may considerably diminish. To solve this problem, the present inventors proposed to combine the contribution of multiple nodes, and to use a communication scheme allowing an intrinsic averaging of the sensed data together with an increase of the back scattered energy.
In yet further embodiment, the connection module may be wireless for wireless connection of said system with the external memory. The external memory may comprise another wireless connection module connecting said system to a user interface via a digital-to-analogue converter (DAC). In a certain embodiment, both wireless connection modules are either Bluetooth or NFC, thereby providing wireless communication between the sensor and the readout module for up to 20 m. If these two modules are Wi-Fi, the connection between them can be established for up to 200 m, while the GSM may provide the worldwide communication.
In a particular embodiment, the external memory supports its own communication network and the reporting to the cloud. To achieve this, a secondary communication link providing long-range and variable data rates is integrated into the node. A specific multi-interface control protocol is designed and prototyped to handle both the locally dense data exchanges (localisation, consolidation, and pre-processing) and the long-range but low data-rate reporting to the cloud. It comes on top of the communication interfaces installed on the external memory devices (such as cellular, Wi-Fi or V2V). It is inspired from heterogeneous V2V algorithms.
In a certain embodiment, the connection to the cloud is direct and simple. Ideally, there is only one physical device that interconnects the swarm of the sensors with the cloud, i.e., the external memory device may have a cellular interface. For areas with low cellular coverage, a private cellular network (4G/5G) may be deployed, for instance on a drone. However, the association of spatial and temporal data to measurements requires a form of collaboration between the external memory devices, and in the hierarchical architecture, the collaboration is enforced by the gateway-cloud node. However, this node does not need to be a physical device, it can be virtual. In 5G, most network functions are virtualised, and specific edge functions can be deployed near the actual working location to minimise latency. The virtual gate-way cloud generates the time signal to synchronise all the external memory devices and runs more advanced protocols: data cleaning, localisation and spectrum optimization. The latter is understood to minimise spectrum usage by the external memory device, for instance by leveraging the sensors' redundancy.
As a first example, where the method and system of the present invention are used is the screening for human breast cancer and skin-based VOC sampling. Reference is now made to
In another example,
In another example, the present inventors were able to discriminate between two enantiomers of butanol in their racemic (1:1) and enantiomeric mixtures using the method of the present invention, which is normally impossible to do with a regular GC-MS chromatography:
The method of the invention mimics an alternative rapid approach to chromatographic enantiomer separation and detection, in the sense that it enables the enantiomers to bind differently to chirality/helicity of the cellulose/lignin fibres of the layered separating medium of the present invention. As a result, the enantiomers are released to each next layer at different time, thereby enabling spatiotemporal separation.
Complementary to the spatiotemporal part, the hybrid FDrGO sensors were loaded with various chiral/achiral ligands that exhibit remarkable sensing difference between each enantiomer. In this regard, the reference is made to
Reference is now made to
For all the hidden layers, the ReLU activation function is used. The output of the ReLU layer is then fed as an input to the next (fully-connected layer) layer, followed by soft-max and classification layers so on until the last layer (output) layer is reached. According to the present invention, the number of the hidden layers can be any, but at least one. In this model of the present invention, the output layer has at least one neuron having low and high activations levels associated with the two possible patterns and time stamps. The output probability is calculated by the loss function (the mean-squared error) between the output patterns of compounds (for example, enantiomers) and their true patterns.
As mentioned above, the specific (exemplary) parameters of the neural network of the present invention are in no way limiting. They may further be fine-tuned and adjusted, including the number of hidden layers, the number of neurons in each layer, the activation function of each layer, the activation function of the output layer, the optimisation method and learning rate, and the addition of regularisation methods.
The machine-learning method of the present invention can also be used for time stamp estimation. This can be achieved by generalising the discrimination problem between two patterns to a discrimination problem between a few to many patterns, where each pattern corresponds to a small time-stamp interval. The structure of the neural network, including for instance the number of layers, number of nodes in each layer, type of activation functions and loss functions, might have to be adjusted in that case.
Reference is now made to
In another aspect of the present invention, the inventors developed all the smart paper e-card when connected in a layered medium format, specifically in a folded origami format. In this case each layer communicates with 2 wires I2C protocol. One “master” card communicates in each layer with each “slave” card in the layer. In this way, it is possible to communicate with all the sensors with minimum number of wire lines. The inventors developed and designed a program externally for communicating all 128-sensor device using I2C communication in 1 second interval. It is also possible to increase the layer numbers (or more sensors) if needed. All software programs and algorithms have been developed in ATMEL MEGA 2560 micro-processor in C language.
In some embodiments, the method and system of the present invention are suitable for use in discriminating various isomers and enantiomers in various mixing conditions by utilising the chiral chemical ligands, such as chiral organic acids, functionalized on graphene network. That allows excellent separation and detection of structurally similar and identical compounds, such as enantiomers.
The obtained results can be seen as a strong indication that the machine-learning methods are actually the methods of choice when analysing the data, for example μGC data in a variety of scenarios and in a variety of applications, including real-time molecular analysis. The present invention clearly shows that machine-learning methods can effectively learn the physical spectroscopic and chromatographic models taking into account various kinetic and thermodynamic parameters and by that constitute an advantageous alternative to modern spectroscopic and chromatographic methods, such as GC-MS, MALDI-TOF and NMR, which are time-consuming, cumbersome, require sophisticated and bulky equipment and trained personnel to carry out the tests.
In a typical synthesis, 100 mg of GO and 50 mg of dopamine hydrochloride (Sigma) were added into 200 mL of 10 mM Tris-solution (PH ˜ 8.6) and sonicated for 15 min in an ice cool water bath for dispersion. The reaction mixture was stirred at 400 rpm at 65° C. for 24 h for reduction (rGO) and subsequent encapsulation by polymerization (DrGO) through Michel addition or Schiff mechanism (see the current specification for details). The product (DrGO) was cooled to room temperature and washed several times with DI water (˜10 times) and ethanol (˜5 times) and collected by centrifuge. In each step of washing with water, pH of the suspension medium was checked until it reaches neutral (pH˜7) to remove the alkaline by products. As prepared, cleaned and dried DrGO powder (5 mg) is dispersed in 20 mL DI water (pH (˜8.6) has been then adjusted with trometamol (Trizma®, Sigma) and mixed with 10 mg of amino-terminated or 20 mg of thiol-terminated chiral/achiral ligand (please see the list in Table 1).
In case of water insoluble ligand, the mixture of water: ethanol (1:1 volume ratio) was used as dispersion reaction medium. The reaction mixtures are gently stirred for ˜15 h at room temperature. After that, mixtures are washed several times with water (10 times) and alcohol (5 times) and centrifugally collected and dried in reduced pressure. In each step of washing pH of the solution was checked until it washes out all alkalinities from the suspension. The dried functionalized powder (FDrGO) is redispersed in DMF and sonicated at low power for dispersion of the ink for ˜1-2 minutes and subsequently used for inkjet printing on paper.
The details of binder free print approach, printing on wide range substrate, reaction mechanism, list of bio-chemical ligands (thiols, amines, chiral) and its advantage over conventional printing from literature is discussed in the present specification).
One-dimensional silver nanowire using solvothermal seed mediumted method. Due to its photosensitivity, the AgCl seed solution was prepared in the dark condition. Typically, aqueous solution of silver nitrate (5 mL, 0.5 M) is mixed with aqueous solution of sodium chloride (5 mL, 1 M) and stirred at ˜700 rpm for a minute. Within few moments white silver chloride immediately precipitated which is separated and washed with deionized water and kept under vacuum for 1 hr in dark. Next, 1.36 g of PVP (40000 MW) are separately dissolved into 160 mL of ethylene glycol in a round flask (with reflux) at 700 rpm at ˜170° C. for 5 minutes. Then an excess (˜50 mg) of dried fine AgCl powder is added in this hot solution and within 1-2 minute the transparent solution will change to light yellow which is an indication of the very fine silver NP. After 3-4 minutes, 0.22 g of silver nitrate AgNO3 are added in this hot solution for grow at 300 rpm. The solution colour was changing in this sequence: light yellow, dark yellow, brown, green and finally grey. This indicates the formation of silver long nanowire.
After that, the solution is kept for cooling at room temperature and cleaned with centrifuge periodically by water and ethanol 10 times to remove extra PVP and EG. The fresh and cleaned silver nanowire is then dispersed in isopropyl alcohol with mild ultra-sonification for 30 s and is used for inkjet printing on paper.
Morphological study using SEM (Sigma 500, Zeiss Ultra-Plus High-Resolution SEM, Germany) was taken to see the microscopic view of synthesized materials dispersed in alcohol and spin coat on silicon wafer (Si/SiO2) to distribute flakes uniformly.
Raman spectra (recorded with Horiba Jobin Yvon Lab RAM HR Evolution Micro-Raman, Japan) was obtained with 532 nm laser. All samples dispersed in ethanol are spin coated on silicon wafer.
FTIR (recorded with Bruker Vertex 70V KBr BS vertical ATR-FTIR) measurements (in the range of 300-4000 cm−1) are done by mixing cleaned and dried (overnight) sample powder with KBr to form pellets. Typical spectrum (see
1H NMR analysis was done with Bruker ARX 300 MHz spectrometer using d-DMSO (Sigma) as the solvent, in 1000 scans at a relaxation time of 2 s.
Typical gate dependent FET properties for carrier type determination were done using spin coated samples on Si/SiO2 wafer with electrometer (Keithley 2636A and 3706) interfaced with the Lab VIEW software.
Cytotoxic assessments are performed for human epithelial lung cells (see
Human epithelial lung cells (BEAS-2B, ATCC, CRL-9609, produced in Israel) were seeded at 96×103 cells per well into 24-wells plates and culture with 0.5 mL full RPMI-1640 medium (Sigma-Aldrich, Israel) at 370 C with 5% CO2. After 24 hours of recovery, cells were washed with the PBS Ca2+/Mg2+ (Phosphate Buffered Saline, Sigma-Aldrich, Israel), and treated for 24 hours. All treatments were performed in the cell culture full medium. Graphene was vortexed shortly before making final dilutions for the treatment in the cell culture medium. Cells were then exposed to 10 and 100 μg/ml graphene.
After 24 hours of the treatment at the indicated concentrations, cells were gently washed with PBS. Annexin-V and propidium iodide (PI) staining was performed according to the manufacturer's instructions (Bio Legend, California, US) in 0.2 mL full RPMI-1640 medium per well. Cells were gently washed again with PBS, and Hoechst 33342 (Invitrogen by Thermo-Fisher Scientific, Israel) 0.2 μg/mL solution was added to the cells. Cells were observed and analysed by an In-Cell Analyzer 2000 System (Technion Life Sciences and Engineering Infrastructure Centre, Technion, Israel).
Frozen tumorous and healthy breast tissues, of no specific type, from different patients were purchased from the Israeli Biorepository Network for Research MIDGAM (Haifa, Israel). Both tumorous and non-cancerous tissues from patients that did not undergo through chemotherapy, were collected after surgery. The frozen tissues were transported inside their plastic cryovials in dried ice and kept in a −80° C. freezer until experiment performance. All tissue samples were thawed in their plastic cryovials to room temperature for 10-20 minutes and 50-100 mg tissue were prepared to be inserted in (500 [μL]-mtissue [mg]) μL 0.9% NaCl solution in double-distilled water, and EPA 524 internal standard mix with 2000 μg/mL 1,4-dichlorobenzene-d4 in methanol, purchased from Sigma-Aldrich, was used as internal standard with concentration of 40 ppb. Tissues from samples were analysed with GC-MS and with the μGC detector on an origami paper-based system of the present invention.
Chromatographic analyses involved a GC-7890B/MS-5977A instrument (Agilent) equipped with an SLB-5 ms capillary column (30 m length; 0.25 mm internal diameter; 1 μm thickness; Sigma-Aldrich) combined with the ITEX PAL RSI 120 headspace autosampler system. Trap preclean conditions were 180 seconds and 260° C. Incubation of samples was set at 55° C. for 15 minutes. Agitator speed was defined as 500 RPM. Desorption flow was 10 μL/sec at 250° C., while extraction strokes were set at 100, extraction volume at 1000 μL and aspirate and dispense flow rates at 100 μL/sec. Injection aspirate flow rate was defined as 10 μL/sec. The carrier gas was helium, the flow through the column was set at 1.5 mL/min and the temperature in the column ranged between 35° C. and 300° C. at a rate of 15° C./min. The GC-MS obtained chromatograms were analysed using Mass Hunter qualitative analysis version B.07.00, Agilent Technologies, and spectral library match NIST was used for compounds identification. The measurements performed with the system and method of the present invention was done with all clinical samples kept at 55° C. water bath and ultra-pure N2 as baseline with ˜10 Hz sampling frequency using Arduino Mega board and the I2C interconnected protocol.
Results obtained with the system and method of the present invention in time domain (see
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/IL2022/050460 | 5/3/2022 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 63228107 | Aug 2021 | US |
