Systems and methods for NMR analysis of the physiological state of an analyte - Patents.com

Abstract

The present disclosure relates to systems and methods for analyzing the physiological state of an analyte, and in particular to systems, components of such systems and methods for analyzing the physiological state of an analyte using hyperpolarized (HP) and magnetic resonance (MR) techniques.

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of European Patent Application No. 21382358.6, filed April 26, 2021.

[0002] The present disclosure relates to systems and methods for analyzing the physiological state of an analyte, and in particular to systems, components of such systems and methods for analyzing the physiological state of an analyte using hyperpolarization (HP) and magnetic resonance (MR) techniques. [Background technology]

[0003] Hyperpolarization-enhanced magnetic resonance (MR) techniques, such as dynamic nuclear polarization (DNP), have been shown to enhance the sensitivity of MR scanning by more than 10,000-fold, making it possible to perform in situ metabolic analysis on a wide variety of sample and analyte types.

[0004] Examples of targets or analytes include healthy and diseased cells of human or animal organs and tissues, human or animal cells (or cell fragments), organoids, body fluids from humans or animals and all animals, and cells and fluids from plants. It should also be understood that this refers to animals such as, but not limited to, humans, rats, rabbits, pigs, and mice.

[0005] Traditional cell culture methods have been valuable for many years because they provide a way to study cell-cell interactions and tissue physiology without the use of laboratory animals or human subjects. One of the major obstacles traditional cell culture presents is the morphology and shape in space of the cells. To survive in culture, cells are seeded into culture vessels (flasks and plates) that have been treated to make them adhere to the bottom, which for certain cell types alters their morphology and function.

[0006] The objective of this disclosure is to develop a non-invasive platform using HP and MR techniques that allows real-time acquisition and analysis of metabolic data related to physiological states in a three-dimensional analyte configuration that is particularly suitable for drug toxicity studies and / or clinical diagnostics. The non-invasive nature of the novel platform offers research possibilities that advance the development of functional human-specific drug testing systems that are not possible with some conventional methods.

[0007] In this patent application, by physiological state is meant a healthy or unhealthy (or diseased) state and a transition state from healthy to unhealthy (or diseased) of the target / analyte or a healthy or unhealthy (or diseased) state and any transition state from healthy to unhealthy (or diseased) of any human or animal, or of the human or animal from which the target or analyte is obtained. Summary of the Invention [Means for solving the problem]

[0008] Thus, in a first main example, a system for determining the physiological state of at least one analyte using hyperpolarized (HP) nuclear magnetic resonance (NMR) is used, the system comprising at least the following components: a hyperpolarized (HP) production device configured for hyperpolarizing a substrate, a hyperpolarized (HP) producing apparatus, which is a microfluidic device having at least one measurement chamber structured to receive a sample matrix containing an analyte, the HP producing apparatus being structured to inject at least one analyte contained in the sample matrix, the HP producing apparatus comprising a hyperpolarized substrate; - an NMR apparatus comprising at least a housing defining a target region for containing at least one analyte comprising a hyperpolarized substrate, at least one magnet unit, at least one magnetic gradient unit for applying, during use, one or more magnetic field gradients in the target region, at least one radiofrequency (RF) pulse generating unit for applying one or more sets of RF pulses towards the target region, and an RF receiving unit for acquiring signals; a computer processing unit for determining the physiological state of the analyte by analyzing the acquired signals.

[0009] In this example, a versatile platform is generated that is capable of acquiring and analyzing metabolic data on the physiological state of analytes in real time using HP and MR techniques in a non-invasive manner, which can be used, for example, to achieve the development of more advanced functional human-specific drug testing systems.

[0010] In a preferred embodiment of this platform, the NMR instrument is structured to accommodate in the target region a microfluidic device that includes a sample matrix containing a hyperpolarized substrate and an analyte contained in a measurement chamber. Such an embodiment thus enables a more standardized and reproducible analytical technique capable of processing and manipulating multiple samples of the same or different analytes containing a hyperpolarized substrate in batch and in-line fashion.

[0011] In further examples of the system, the hyperpolarization (HP) producing device may be a dynamic nuclear polarization (DNP) device or a para-hydrogen induced polarization (PHIP) device. In further examples of the system, the hyperpolarization (HP) producing device may include a chemical-induced dynamic nuclear polarization (CIDNP) device, an Overhauser DNP (dynamic nuclear polarization) device, a Signal Amplification by Reversible Exchange (SABRE) device or a Spin Exchange Optical Pumping (SEOP) device.

[0012] In examples, hyperpolarized (HP) production devices configured to hyperpolarize a substrate may include, but are not limited to, devices such as Dynamic Nuclear Polarization (DNP) devices, Parahydrogen Induced Polarization (PHIP) devices, Chemically Induced Dynamic Nuclear Polarization (CIDNP) devices, Overhauser DNP (Dynamic Nuclear Polarization) devices, Signal Amplification by Reversible Exchange (SABRE) devices, or Spin Exchange Optical Pumping (SEOP) devices.

[0013] In example methods described herein, the hyperpolarization step may include hyperpolarization methods such as, but not limited to, d-DNP (dissolution dynamic nuclear polarization), PHIP (parahydrogen induced polarization), CIDNP (chemically induced dynamic nuclear polarization), SABRE (signal amplification by reversible exchange) and SEOP (spin exchange optical pumping).

[0014] To emulate a physiologically relevant cellular microenvironment for the analyte under test, at least one measurement chamber is in fluid communication with a matrix perfusion supply line for supplying a matrix perfusion medium to the measurement chamber, a hyperpolarized substrate supply line for supplying a hyperpolarized substrate to the measurement chamber, and a chamber outlet line. This configuration of the microfluidic device ensures a suitable environment for the analyte (e.g., cells). The measurement chamber serves as a cage incubation enclosure that maintains a stable environment in terms of temperature and gas composition, providing the necessary sterile conditions for carrying out the methods according to the present disclosure.

[0015] In particular, the hyperpolarized agent supply line is located above the matrix perfusion supply line with respect to the bottom wall of the measurement chamber, and the chamber outlet line is located below the hyperpolarized base supply line and above the matrix perfusion supply line with respect to the bottom wall of the measurement chamber. Preferably, the matrix perfusion supply line is located at the bottom wall of the measurement chamber.

[0016] This allows for a continuous flow of matrix perfusion medium (culture medium) through the measurement chamber as well as the injection of hyperpolarized substrate for the analyte being analyzed.

[0017] Preferably, the microfluidic device includes a temperature control unit for maintaining the analyte, including the hyperpolarized substrate, at body temperature when housed in the NMR instrument, thereby achieving and maintaining optimal measurement conditions that mimic the human or animal body. In a particular example, the temperature control unit includes a water circulation circuit.

[0018] In a preferred example, to establish a standardized and reproducible analytical technique capable of processing and manipulating multiple samples of the same or different analytes containing hyperpolarized substrate in a batch and in-line manner, the microfluidic device comprises multiple measurement chambers, each in fluid communication with a hyperpolarized agent supply line, such that all measurement chambers containing (sample matrices containing) the analyte receive the same amount of hyperpolarized substrate.

[0019] According to the present disclosure, the system is intended for real-time acquisition and analysis of metabolic data related to the physiological state of an analyte. In order to process such large amounts of data in real-time and provide a more effective and more accurate analytical technique that is less sensitive to noise, the computer processing unit of the system is structured to analyze at least the acquired signals using one or more machine learning algorithms.

[0020] Preferably, the machine learning algorithm is a computer-implemented artificial neural network, and the computer processing unit further comprises a training unit configured to: train the computer-implemented artificial neural network with a sequence of time-lapse signals, characterized by a training sequence of time-lapse signals for a known physiological state of the analyte comprising the hyperpolarized base; apply a sequence of input signals to the computer-implemented artificial neural network, characterized by at least one test sequence of time-lapse signals for an unknown physiological state of the analyte comprising the hyperpolarized base; and analyze each applied test sequence of time-lapse signals and generate, for each test sequence of time-lapse signals, a predicted physiological state of the analyte comprising the hyperpolarized base.

[0021] Optionally, during training of the computer-implemented artificial neural network, the method may further include an output of the predicted physiological state to verify accuracy and allow adaptation of the training step.

[0022] The use of a trained artificial neural network allows for a suitable and accurate determination of the predicted physiological state. The artificial neural network can be trained based on machine learning techniques, preferably using deep learning techniques. An example of a deep learning technique can be back-propagation learning. The input training data used can be MR acquired images annotated by a medical professional, showing the physiological state of the analyte (e.g. cells).

[0023] Thus, the computer processing unit may include an output unit configured to output the determined physiological state of the analyte.

[0024] The present disclosure also relates to a method for determining a physiological state of an analyte comprising a hyperpolarized substrate, using a system according to any one or more of the preceding claims, the method comprising: a) adding at least one analyte; b) injecting at least one analyte containing a hyperpolarized substrate; c) applying one or more magnetic field gradients to an analyte containing a hyperpolarized substrate positioned in a target region of the NMR instrument using at least one magnet unit and at least one magnetic gradient unit of the NMR instrument; d) applying one or more sets of radio frequency (RF) pulses toward the target region using at least one RF pulse generating unit of the NMR device; e) receiving a sequence of NMR signals over time from an RF receiving unit of the NMR device in response to one or more applied magnetic field gradients; f) generating one or more time-frequency representations of the sequence of NMR signals by a time-frequency transformation; g) analyzing the time-frequency representation using one or more machine learning algorithms to determine a physiological state of the analyte.

[0025] It is noted that according to this disclosure of the invention, the analyte preparation provided in step a) may be contained in a sample matrix that emulates a physiologically relevant cellular microenvironment for the analyte under test.

[0026] In a preferred embodiment of the method according to the invention, the at least one analyte is positioned in the target area of ​​the NMR instrument prior to step b) of injecting the at least one analyte comprising a hyperpolarized substrate, so that the injection step b) is carried out within the NMR instrument.

[0027] In another example of the method according to the invention, the at least one analyte is positioned in the target area of ​​the NMR instrument after step b) of injecting the at least one analyte comprising a hyperpolarized substrate. Thus, the injection step b) is performed outside the NMR instrument and in this example of the method, the at least one analyte comprising a hyperpolarized substrate is positioned in the target area of ​​the NMR instrument.

[0028] The time-frequency transformation of step f) is selected from the group represented but not limited to Fourier transform, while the one or more machine learning algorithms are selected from the group represented but not limited to artificial neural networks, decision trees, regression models, k-nearest neighbor models, partial least squares regression, support vector machines or an ensemble of models combined to define the algorithm.

[0029] In particular, the machine learning algorithm is a computer implemented artificial neural network, and the analyzing step g) comprises: i) training a computer-implemented artificial neural network with sequences of NMR signals over time, the sequences characterizing training sequences of NMR signals over time for known physiological states of an analyte including a hyperpolarized substrate; ii) applying to a computer-implemented artificial neural network a sequence of NMR input signals over time, the sequence characterized by at least one test sequence of NMR signals over time for an unknown physiological state of an analyte including a hyperpolarized substrate; iii) analyzing each applied test sequence of time-lapse NMR signals and generating, for each test sequence of time-lapse NMR signals, a predicted physiological state of the analyte including the hyperpolarized substrate.

[0030] In a further example, while carrying out the method according to the present disclosure, the receiving step b) can occur after the step iv) of hyperpolarizing the substrate with a hyperpolarized (HP) production device. In a further functional embodiment, after step b) but before step c), the method further comprises the step b1) of maintaining the analyte containing the hyperpolarized substrate at body temperature.

[0031] For example, the methods of the present disclosure may be embodied in a computer program or product that includes computer coded instructions that, when executed by a computer, such as a laptop or computer, cause the computer to perform the steps of the computer-implemented methods disclosed herein.

[0032] In a particular embodiment, a computer readable storage medium is proposed that includes computer coded instructions stored therein, which, when executed by a computer, cause the computer to perform the steps of the computer implemented method disclosed herein. Such a computer readable storage medium may be a (solid state) hard drive, a USB drive or a (digital) optical disk.

[0033] In another part of the disclosure, a sample matrix for containing an analyte for in vitro analysis is proposed, the sample matrix being formed as a three-dimensional construct containing open pores, the construct being a gel comprising at least sodium carboxymethylcellulose.

[0034] In particular, the gel comprises at least 0.5-5% by weight of sodium carboxymethyl cellulose (CMC), in particular 1% by weight of sodium carboxymethyl cellulose (CMC).

[0035] Thus, an open 3D construct is formed from 1% CMC cryogel. This construct serves as a sample matrix to accommodate multiple cells or analytes in the open internal space. Such 1% construct has good stability and pore size distribution. In particular, 1% CMC cryogel has been found to have low affinity for cell attachment and good physicochemical properties for NMR applications. Therefore, the open construct of the sample matrix provides optimal conditions to mimic the human or animal body, allowing cell structures to survive and form them into spheroids (3D configurations) to interact with each other instead of the gel material. [Brief description of the drawings]

[0036] The present disclosure will be described with reference to the drawings identified below. [Figure 1] 1 is an example of a system according to the present disclosure implementing an example method according to the present disclosure. [Figure 2a] 1 details several components of a system according to the present disclosure; [Figure 2b] 1 details several components of a system according to the present disclosure; [Figure 2c] 1 details several components of a system according to the present disclosure; [Figure 2d] 1 details several components of a system according to the present disclosure; [Figure 3a]1 details several components of a system according to the present disclosure; [Figure 3b] 1 details several components of a system according to the present disclosure; [Figure 3c] 1 details several components of a system according to the present disclosure; [Figure 3d] 1 details several components of a system according to the present disclosure; [Figure 3e] 1 details several components of a system according to the present disclosure; [Figure 4] 1 details several components of a system according to the present disclosure; [Diagram 5] 1 details several components of a system according to the present disclosure; [Figure 6] 1 details several components of a system according to the present disclosure; DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0037] Detailed Description of the Invention For a proper understanding of the present disclosure, the following detailed description corresponding to corresponding elements or parts of the present disclosure are designated with the same reference numerals in the drawings.

[0038] In Fig. 1 a first example of a system according to the invention is shown. The example in Fig. 1 is schematic and not to scale, the main components of the system are shown with their interrelated functions. Each component will be described in more detail and in a more modified state in relation to the other figures.

[0039] A system according to the present invention for determining the physiological state of at least one analyte using hyperpolarized (HP) nuclear magnetic resonance is indicated generally by the reference numeral 1000 .

[0040] In summary, the system comprises at least the following components: A microfluidic device 100 comprising at least one measurement chamber 110a structured to accommodate a sample matrix 10 containing an analyte 1; a hyperpolarized (HP) producing device 200 configured to hyperpolarize a substrate; an NMR apparatus 300 comprising at least a housing defining a target region 310 for containing at least an analyte 1 comprising a hyperpolarized substrate 6; a computer processing unit 400 for analysing the acquired signals to determine the physiological state of the analyte 1.

[0041] In the next paragraphs of this description, some components of the system 1000 are described in more detail.

[0042] Figure 2a shows an advantageous alternative to the traditional cell culture method that has been used for many years. Also, with this alternative, an effective method for studying cell-cell interactions and tissue physiology can be realized without using laboratory animals or human subjects. One of the major obstacles that traditional cell culture presents is the morphology and shape in space of the cells. To keep the cells (or the analytes of interest in this description) alive in culture, they are seeded in culture vessels (flasks and plates) that have been treated to make them adhere to the bottom, which for certain cell types changes their morphology and function.

[0043] Reference number 10 in Fig. 2a denotes a sample matrix for housing one or more analytes or cells (designated throughout with reference number 1) in a three-dimensional configuration for in vitro analysis. According to the invention, the sample matrix 10 is formed as a three-dimensional construct in this particular alternative. The three-dimensional construct of the sample matrix 10 encompasses a space 10z in which clusters of individual analytes or cells 1 are housed.

[0044] As shown in Figure 2a, the three-dimensional construct of sample matrix 10 is formed from a gel that configures a mesh or crisscross lattice pattern of interconnected gel strands 11a-11b with open pores 11z between them, the open pores 11z being of irregular shape and size distribution.

[0045] The gel of the open three-dimensional construct of the sample matrix 10 comprises at least sodium carboxymethylcellulose, preferably at least 0.5-5% by weight of sodium carboxymethylcellulose, in particular 1% by weight of sodium carboxymethylcellulose.

[0046] The following method steps will explain in detail each step required to properly synthesize 1% carboxymethylcellulose gel. As the basic starting component for the preparation, 1% carboxymethylcellulose gel is used: Sodium carboxymethylcellulose (CMC) [419273, Sigma] Adipic acid dihydrazide (AAD) [MW=174.2g / mol] Morpholino-ethane-sulfonic acid (MES) [MW=195.2g / mol] and N-(3-Dimethylaminopropyl)-N´-ethylcarbodiimide hydrochloride (EDC) [E7750-10g, Sigma] The first step is to prepare 0.5M MES buffer (pH 5.5). For 100 mL, the steps are: 1. Prepare 80 mL of distilled HO in a container. To a solution of 2.80 mL of distilled HO, add 9.76 g of MES free acid. 3. Adjust the pH of the solution to 5.5 using NaOH. For 100 mL, a distilled HO solution may require the addition of 366 mg NaOH (solid) or 9.15 mL of 1 M NaOH. 4. Add additional distilled H2O solution until the volume of the solution reaches 100 mL. For more details: [Table 1] 5. The solution obtained in step 4 is then stored at 4°C. Next, prepare the polymer premix (1.1 mL): Prepare 6.1 wt. % CMC by dissolving 50 mg of it in 5 mL of distilled HO using stirring (e.g., using a magnetic stirrer at 900 rpm). 7. Add 100 μL of AAD (50 mg / mL) to 1 mL of 1% CMC solution and mix this by pipetting. 8. Pre-cool this premix at 4°C. EDC (1 mg / μL) is prepared by dissolving 9.4 mg of EDC in 4 μL of distilled HO. 10. Add EDC to the premix, pipetting up and down to ensure homogeneity of the solution. For more details: [Table 2] 11. The transfer gel is then more quickly transferred to a mold and incubated. If the volume of the mold is large, pre-cool the mold in a -20°C freezer before adding the gel. 12. Incubate at -20° C. The mold containing the gel must be placed in the freezer for 24 hours. Afterwards, the cryogel must be carefully removed from the mold.

[0047] Thus, an open three-dimensional construct is formed from the 1% CMC cryogel. This construct serves as a sample matrix 10 for accommodating a plurality of cells or analytes 1 in its open internal space 10z. In particular, 1% CMC cryogel has been found to have a low affinity for cell attachment. Thus, the open construct of the sample matrix 10 provides optimal conditions to mimic the human or animal body, allowing cell structures to survive and form into spheroids to interact with each other instead of the gel material.

[0048] An example of a sample matrix 10 obtained by the above preparation technique has a diameter of 5 mm and a height of 2 mm. More preferably, the dimensions of the sample matrix 10 may range from 3 to 10 mm in diameter and 1 to 6 mm in height, depending on the intended use.

[0049] The sample matrix 10 increases cell viability and optimizes the metabolic conditions required, for example, to study hepatocyte metabolism. In one example, the AML12 cell line, composed of non-cancerous mouse hepatocytes, was used to study cell senescence under laboratory conditions, as shown in Figure 2b. It was observed that 13 hours after seeding the cells 1 in the cryogel sample matrix 10, the formation of spheroids could already be observed. The open three-dimensional architecture of the sample matrix 10 confirms that the cells 1 are held inside the internal space 10z by the interconnected cryogel strands 11a-11b, with open pores 11z between the cells, and aggregate with themselves.

[0050] In addition, it should be noted that the sample matrix 10 made from CMC materials as disclosed above does not interfere during NMR signal acquisition, thereby allowing for precise perfusion of the hyperpolarized substrate.

[0051] In Figure 2b, 1 million cells were seeded into the cryogel and immediately placed inside the microscope for longitudinal imaging. Spheroid formation was observed starting 3 hours after cell seeding and up to 10 hours later.

[0052] Cryopreservation of cells has traditionally been performed in cell suspension. With this known technique, cells are cultured in a 2D system until they are stress-free and before covering the bottom of the flask in which they are cultured. Typically, the cells are poured into a freezing medium (containing, for example, serum and DMSO as cryopreservatives) at, for example, -80 °C. After about 12-24 hours, the pre-cooled cells are transferred to a nitrogen tank for storage. Once thawed, the cells are cultured back in 2D conditions. However, this process takes 3-4 days before the cells can be used in the experimental setup.

[0053] To analyze the preservation potential of multiple 3D sample matrices 10 containing analytes 1 prior to diffusion through the hyperpolarized substrate 6 and measurement with the NMR device 300, the 3D sample matrices 10 containing analytes 1 are frozen to sub-zero temperatures and structured to maintain said temperatures for several days. Surprisingly, the 3D sample matrices 10 containing analytes 1 are able to maintain the analytes 1 in good condition even after thaw cycles without adversely affecting or damaging the analytes 1. Viability and metabolism assays confirm that the analytes remain functional after the thaw process without the need for a cell culture room or incubator for system use.

[0054] As an example of the functionality of the sample matrix 10 according to the present disclosure to maintain the analyte 1 in good condition even after thaw cycles and without adversely affecting or damaging the analyte, FIG. 2c shows metabolic results obtained from an Alamar Blue test, which shows that cells remain viable inside the cryogel construct after cryopreservation inside the construct for periods of up to 11 days, with a viability similar to cells that have not been subjected to the cryopreservation process. The results are shown in both the fluorescence and absorbance readings of the Alamar Blue test, with the two outputs coming from the same metabolic test.

[0055] Furthermore, FIG. 2d shows the structural integrity of the cryogel before and after undergoing cryopreservation in two confocal microscopy images A and B (scale bar 100 μm) of the interior of the cryogel. FIG. 2dA (left) shows the cryogel of sample matrix 10 imaged before the cryopreservation process, whereas FIG. 2dB (right) shows the cryogel of sample matrix 10 imaged after the cryopreservation process. As can be observed, no damage to the cryogel fibers 11a-11b occurs during cryopreservation, thus preserving the structural integrity of the open 3D architecture of the sample matrix 10.

[0056] The results in Figures 2c and 2d provide evidence that cells / analytes 1 can be seeded into cryogel constructs as shown in Figure 2a and cryopreserved in liquid nitrogen after 24 hours. Tests show high viability up to 11 days after thawing, when compared to results for cryogels that were not subjected to cryopreservation. Furthermore, the cryogel constructs are able to sustain the cryopreservation process after seeding with cells / analytes 1 and maintain their structural integrity, which is necessary for the subsequent injection of hyperpolarized substrate 6 and NMR data acquisition.

[0057] Figures 3a and 3b show further components of the system according to the invention in more detail.

[0058] Reference number 100 denotes a component of the system that is a microfluidic device. In one example, the microfluidic device 100 consists of a stack of structures shown with reference numbers 101, 102 and 104. Reference number 101 describes a substrate structure made up of multiple (e.g. 3 layers) polydimethylsiloxane (PDMS) layers 101-2, 101-3 and 101-4 stacked on a glass slide 101-1. Typical and preferred dimensions of the glass slide are 75 x 50 mm. This can be modified to include the dimensions of the NMR instrument.

[0059] 3b details the microfluidic device 100 in more detail, and in particular a single microfluidic element 110 mounted on a substrate structure 101. Preferably, the microfluidic device 100 comprises a plurality of such microfluidic elements 110, each of which comprises such a well or measurement chamber 110a for individually receiving a sample matrix 10 comprising an analyte 1.

[0060] As shown in Fig. 3a, the substrate structure 101 includes several quadrants or arrays 101v-101w-101x-101y (here four arrays). Each array 101v-101w-101x-101y is composed of several (here four) microfluidic elements 110 of identical configuration. In this example, as shown, the substrate structure 101 configuration is composed of 4 x 4 (16) microfluidic elements 110. In one example, each microfluidic element 110 includes a well / measurement chamber 110a with dimensions of diameter D of 6 mm and height H of 10 mm. See Fig. 3d.

[0061] Each microfluidic device 110 contains channels 111a and 111b, which are in fluid communication with a well or measurement chamber 110a. Channel 111a represents a matrix perfusion supply line for supplying a matrix perfusion medium 5 to the measurement chamber 110a. Channel 111b represents a microfluidic device outlet line. Channels 111a and 111b may be provided in a second PDMS layer 101-2 (e.g., 5 mm thick) on the substrate structure (glass slide) 101.

[0062] Additionally, the microfluidic device 110 includes a channel 111c, which functions as a chamber outlet line. The channel 111c may be provided in a third PDMS layer 101-3 (eg, 1 mm thick).

[0063] Channel 112 represents a hyperpolarized base supply line for supplying hyperpolarized base 6 (e.g., hyperpolarized pyruvic acid) to measurement chamber 110a. Channel 112 can be provided in a fourth PDMS layer 101-4 (e.g., 4 mm thick).

[0064] In an advantageous example, the PDMS layers 101-1, 101-2 and 101-3 are fabricated by replica molding using SU-8 microstructures photolithographically created on a 4-inch silicon wafer. The silicon substrate is dehydrated for 30 minutes using a hot plate at 200° C., and then cleaned / activated using O2 plasma (PDC-002, Harrick Plasma, Ithaca, NY, USA) treatment at 22.5 mL / min and 30 W for 20 minutes. A photoresist material (SU-8 2100, KAYAKU Advanced Materials, Inc., Westborough, MA, USA) is spin-coated to form a 200 μm thick SU-8 layer. A PDMS prepolymer is also prepared with a ratio of 10:1 (base:hardener, w / w) and degassed in a vacuum desiccator. The prepolymer was then cast onto a Petri dish containing the SU-8 mold and allowed to sit at 65°C for 4 hours, followed by overnight at room temperature. O2 plasma was used to activate and bond the three PDMS layers together, including the glass slide, resulting in an 11 mm thick device.

[0065] As shown in detail in Fig. 3d, for each microfluidic element 110 of the microfluidic device 100, the hyperpolarizer supply line 112 is positioned above the matrix perfusion supply line 111a with respect to the bottom wall 110z of the well / measurement chamber 110a. Similarly, the chamber outlet line 111c is positioned below the hyperpolarizer supply line 112 and above the matrix perfusion supply line 111a with respect to the bottom wall 110z of the measurement chamber 110a. In particular, the matrix perfusion supply line 111a is positioned at the bottom wall 110z of the measurement chamber 110a. Thus, during use, the open porous sample matrix 10 containing the analyte 1 is flushed from the bottom side upwards towards the chamber outlet line 111c.

[0066] Thus, a continuous flow of matrix perfusion medium 5 (culture medium) through the measurement chamber 110a is possible, as well as the injection of hyperpolarized substrate 6 for the analyte being analyzed. In particular, by a specific height configuration of the several channels 111a-111c-112, the chamber outlet line 111c withdraws matrix perfusion medium 5 from each well / measurement chamber 110a while maintaining a constant liquid height h above the bottom 110z of the well 111a. In use, this therefore accounts for a total medium volume of about 140 μL per well 110a.

[0067] In the example microfluidic device 100, up to four biological samples can be analyzed by four groups of four microfluidic elements 110, each with a well / measurement chamber 110a. The microfluidic elements 110 are grouped into four separate sets 101v-101w-101x-101y of four wells 110a, such that each set of four wells / measurement chambers 110a shares media 5 via a central matrix perfusion supply line 111a' and a central matrix perfusion outlet line 111b'.

[0068] In Fig. 3b, microfluidic resistors 115 are shown diagrammatically in the microfluidic channels 111a-111b-111c-112 of each microfluidic component 110. Although Fig. 3b shows such microfluidic resistors 115 located in each microfluidic channel 111a-111b-111c-112, it is sufficient to have a microfluidic resistor 115 in at least one of the matrix perfusion supply line 111a and the microfluidic component outlet line 111b. See also Fig. 3e, which shows in more detail the microfluidic resistor 115 structured as a serpentine.

[0069] Microfluidic resistors 115 at the inlets 111a and outlets 111b of the wells / measurement chambers 110a allow for the same flow rate through each well / measurement chamber 110a containing the 3D sample matrix 10 with the analyte 1. Furthermore, all of the wells / measurement chambers 110a of a set 101v-101w-101x-101y are connected to an inset suction reservoir 114 that draws off the matrix perfusion medium 5. In another example, a microfluidic resistor 115 can be placed in each central matrix perfusion supply line 111a' and central matrix perfusion outlet line 111b' that are part of each set 101v-101w-101x-101y of the microfluidic device 110.

[0070] Because the hyperpolarized agent supply line 112 is positioned relative to the bottom wall 110z of the wells / measurement chambers 110a at a position above both the matrix perfusion supply line 111a and the chamber exit line 111c, the channel 112 can distribute the hyperpolarized agent 6 to all of the wells / measurement chambers 110a that are part of the same set 101v-101w-101x-101y. Thus, separation between the sets 101v-101w-101x-101y is maintained, allowing the hyperpolarized agent 6 to be injected into each of the sets 101v-101w-101x-101y, enabling more different and diverse analyses.

[0071] The appropriate environment for the analyte 1 is created by housing the microfluidic device 100 in an incubation enclosure 102. The incubation enclosure 102 maintains stable temperature and gas (O2 and CO2) and sterile conditions while personnel manipulate the microfluidic device 110 into the NMR instrument 300.

[0072] The microfluidic device 100 therefore comprises a temperature control unit 103 in its incubation enclosure 102, which serves to maintain each microfluidic element 110, so that the assembly is made up of a 3D sample matrix containing the analyte 1 with the hyperpolarized substrate 6 at body temperature. This latter requirement is necessary when the microfluidic device 100 with several microfluidic elements 110 is housed in an NMR instrument 300. In this way optimal measurement conditions mimicking the human or animal body are achieved and maintained. In a particular example, the temperature control unit comprises a water circulation circuit 103 with a water inlet 103a and a water outlet 103b.

[0073] The temperature inside the incubation enclosure 102 is controlled by pumping warm water from a water bath into the enclosure base 102a, which acts as a water jacket. Otherwise, the enclosure lid 102b has an inlet 104a and an outlet 104b to provide stable gas incubation.

[0074] For a suitable use of the system and method according to the invention, a sample matrix 10, as described in relation to figures 2a-b, with analytes 1 in an open porous space 10z, is placed inside each well / measurement chamber 110a of each microfluidic element 110 of the microfluidic device 100. A fluidic system is then formed, consisting of several microfluidic channels 111a-111b-111c-112, allowing a continuous flow of culture medium 5 through each well / measurement chamber 110a, thereby perfusing each sample matrix 10 with the analytes 1 contained therein.

[0075] The fluidic system includes a reservoir 114 connected to a vacuum pump and a tube pump (both indicated with reference number 150 in FIG. 5). This setup is replicated for all four sets 101v-101w-101x-101y. The tube pump 150 drives the matrix perfusion medium 5 (culture medium) back and forth from the reservoir 114 to each well / measurement chamber 110a of the microfluidic component 110 (device 100), thereby perfusing the sample matrix 10 and the analytes 1. The vacuum pump 150 applies negative pressure to the reservoir 114 and draws the matrix perfusion medium 5 (culture medium) out of the microfluidic component 110 via the microfluidic component outlet line 111b.

[0076] To perform a hyperpolarized NMR measurement, the tube pump is shut off and a certain amount of hyperpolarized (HP) substrate 6 (e.g., hyperpolarized C13 labeled pyruvic acid) as obtained from the hyperpolarized (HP) production device 200 is injected into a certain central hyperpolarized agent supply line 112'. The central hyperpolarized agent supply line 112' divides the certain amount of hyperpolarized (HP) substrate 6 into smaller samples. The number of smaller samples of hyperpolarized (HP) substrate 6 is equal to the number of microfluidic devices 110 in the set 101v-101w-101x-101y.

[0077] In this example, a particular central hyperpolarized agent supply line 112' is in fluid communication with each of the wells / measurement chambers 110a of the 16 microfluidic devices 110. This configuration allows all wells 110a containing sample matrix 10 with analytes / cells 1 to receive the same hyperpolarized sample volume 6. For example, a sample volume of 800 μL results in 50 μL of hyperpolarized (HP) agent 6 per well 110a / microfluidic device 110.

[0078] Due to its porosity and the sample matrix fabrication methodology, the open three-dimensional architecture of the sample matrix 10 exhibits high permeability, which allows rapid partitioning of the hyperpolarized (HP) substrate 6 through the 3D sample matrix 10 containing the analyte 1. Rapid partitioning of the hyperpolarized (HP) substrate 6 is important due to the rapid decay rate of the hyperpolarized (HP) substrate 6.

[0079] According to one particular example of the system, the hyperpolarization (HP) producing device 200 may be a dynamic nuclear polarization (DNP) device or a parahydrogen induced polarization (PHIP) device.

[0080] As shown in Fig. 4, dissolution DNP is typically performed using two magnet units: one magnet unit 210 is implemented in a hyperpolarized (HP) production device 200 to polarize the substrate 6 in the solid state (preferably at a 3.4-7.0 T magnetic field strength B0), while the other magnet unit is the magnet unit of a dedicated NMR device 300. For adequate data acquisition, current horizontal bore medical magnet units of the NMR device 300 typically range from 1.5-3 T, although higher magnetic fields of up to 7 T are currently available. High resolution vertical bore NMR devices reach higher magnetic fields of up to 28.2 T.

[0081] The dissolution DNP process begins with hyperpolarizing the substrate 6. An example substrate is, for example, C13-labeled pyruvic acid. In addition to the free polarized magnetic field-applied radicals (substances containing unpaired electrons), a vitrification agent may be incorporated into the substrate 6. If the substrate does not form a glass by itself, a vitrification agent such as glycerol or DMSO can be used to vitrify the substrate 6. This ensures homogeneous distribution of radicals in the analyte 1 and good microwave transmission.

[0082] After this preparation step, the substrate 6 was inserted into the DNP magnet unit 210 of the hyperpolarized (HP) production apparatus 200 and irradiated with microwaves to drive the electron-nuclear polarization transfer. The solid-state polarization build-up depends on many factors, including the nature and concentration of the radicals and substrates, the electronic and nuclear relaxation rates, and the composition of the frozen mixture. It may take several hours for the nuclear polarization to reach maximum. Once the 13C is polarized, the usual DNP enhancement is performed by directly irradiating the frozen sample with a MW frequency suitable for 13C. Using a 6.7T DNP magnet at 1.2K, 70% of the 13C polarization can be achieved in 20 minutes before dissolution.

[0083] Once the polarization build-up was completed, the hyperpolarized base 6 was thawed into a cryostat using hot water at 180° C. and 10 bar pressure, finally obtaining the hyperpolarized base in liquid form.

[0084] After being hyperpolarized in the hyperpolarized (HP) production device 200, the hyperpolarized substrate 6 is then injected into a microfluidic device 100 containing at least one 3D sample matrix 10 with the analyte 1 to be analyzed. The microfluidic device 100 containing at least one 3D sample matrix 10 (containing the analyte 1 with an amount of hyperpolarized substrate 6 to be injected) contained in a measurement chamber 110a is then placed into an NMR instrument 300, after which a 13C signal is acquired (either upon injection into an animal or into a test tube).

[0085] In an alternative embodiment, a microfluidic device 100 comprising at least one 3D sample matrix 10 housed in a measurement chamber 110a and containing an analyte 1 is placed into an NMR instrument 300, after which a quantity of hyperpolarized substrate 6 is injected into the microfluidic device 100, thereby injecting the analyte 1 to be analyzed.

[0086] 5, in a preferred embodiment of this platform, an NMR instrument 300 is structured to accommodate, in a target region 310, one or more microfluidic devices 100 each containing an analyte 1 injected with a hyperpolarized substrate 6 contained in an associated measurement chamber 110a. The system 1000 is structured to determine the physiological state of only one sample of analyte 1 (in practice, a large volume of a particular type of cell or analyte 1) containing hyperpolarized substrate 6.

[0087] 3a and 3b, the described microfluidic device 100 has multiple (e.g., 4×4) microfluidic elements 110, each structured to accommodate a 3D sample matrix containing cells or analytes 1 injected with a volume of hyperpolarized substrate 6 to be analyzed in an associated measurement chamber 110a. Such embodiments thus enable a more standardized and reproducible analytical technique capable of processing and manipulating multiple samples of the same or different analytes 1 using the same volume of hyperpolarized substrate 6 simultaneously, as well as in batch and in-line fashion.

[0088] The NMR device 300 also houses at least one magnet unit 310 and at least one magnetic gradient unit (not shown) for applying, during use, one or more magnetic field gradients B0 at the target region 310. The NMR device 300 further includes in its housing at least one radio frequency (RF) pulse generating unit (not shown) for applying a set of one or more RF pulses towards the target region 310 and one or more quantities of hyperpolarized substrate 6 containing cells or analyte 1.

[0089] To simultaneously obtain spectroscopic information 333 from all samples (one or more volumes of cells or analytes 1 injected with hyperpolarized substrate 6) contained in the 3D sample matrix 10 of the measurement chambers 110a of several microfluidic elements 110 of the microfluidic device 100, RF pulse sequences are generated and applied by at least one radio frequency (RF) pulse generating unit. These RF pulse sequences provide as much spatially localized spectral information 333 as possible in a relatively short time window of data acquisition. Since the polarization of the hyperpolarized substrate 6 decays rather rapidly over time, it is desirable to obtain as much spectral information as possible in such a short period of time.

[0090] Traditional pulse sequences such as echo planar imaging (EPI) acquire the information necessary to reconstruct an NMR image in a single data acquisition shot, but do not acquire spectral information. Echo planar spectroscopic imaging (EPSI) allows differentiation of nuclear processes at different chemical shifts, but each data acquisition shot acquires only partial information, so multiple data acquisition shots are required. Chemical shift imaging (CSI) provides 2D spatial information and the third dimension, i.e., spectral information for each voxel, and multiple data acquisition shots are required.

[0091] Recently, imaging techniques such as single-shot NMR and spatiotemporal encoding (SPEN) allow chemical shift discrimination in a single spectral data acquisition. Two versions of these pulse sequences are considered useful for hyperpolarized 13C experiments. One option is to perform localized spectroscopy by exciting a region with the shape of interest. From this selected region, an NMR signal is acquired to obtain a spectrum containing all the chemical shifts of the nuclei in the excitation region of the analyzed sample.

[0092] A time sequence set of NMR signals thus generated in the target region is acquired using an RF coil of an RF receiving unit of the NMR device 400. The RF coil converts changes in the magnetic field in the target region 310 induced by nuclear processing of the analyte 1 into an electromagnetic current signal 333 and transmits this to the computer processing unit 400. The computer processing unit 400 analyses the acquired signals and determines the physiological state of the analyte 1 based on the analysis.

[0093] Once data acquisition has taken place after the NMR measurement, the perfusion system is started again to dilute the remainder of the hyperpolarized (HP) base 6 with matrix perfusion medium 5 .

[0094] As outlined in Fig. 6, the computer processing unit 400 is structured to analyze at least the acquired signals 333 acquired from the NMR device 300 using one or more machine learning algorithms. Thus, the computer processing unit 400 implements a data receiving unit 401 that is structured to acquire NMR signals converted into time-dependent electromagnetic current signals from the NMR device 300. The at least one data receiving unit 401 utilizes suitable signal wiring for receiving a data stream of time sequence sets of NMR signals and may be connected to the NMR device 300 by a wireless data communication interface operating according to a network protocol for exchanging data, such as specified ZigBee™, Bluetooth™, or Wi-Fi based protocols for wireless networks.

[0095] In a step of the computer-implemented method according to the invention, the computer processing unit 400 amplifies and stores the time sequence set of NMR signals (time-domain signal-free induction decay, FID) as intensity versus time. For this purpose, the computer processing unit 400 utilizes at least one data conversion unit 401, which implements at least one time-frequency conversion algorithm for generating one or more time-frequency representations of the time sequence set of NMR signals. In a simple spectroscopy experiment, the NMR signals are zero-filled, Fourier transformed and baseline corrected. The area under the curve of each peak of the spectrum is calculated and correction factors are applied to the flip angle or other acquisition characteristics. More complex pulse sequences may require signal deconvolution (e.g. SPEN).

[0096] The time-frequency transform 444 of the time sequence set of NMR signals may be selected from a group representative of, but not limited to, a short-time Fourier transform, a wavelet transform, a filter bank, or a discrete cosine transform.

[0097] The computer processing unit 400 is structured to analyze the time-frequency representation of the set of time sequences of NMR signals using one or more machine learning algorithms to determine a physiological state of the analyte 1. For illustrative purposes, the one or more machine learning algorithms are indicated by reference numeral 402 in FIG. 6. In one example, the computer processing unit 400 may implement an algorithm processing unit 402 incorporating a computer readable storage unit (not shown) that includes computer coded instructions stored therein, which when executed implement the machine learning algorithms.

[0098] The storage unit of the computer processing unit 400 may be a hard disk unit or a solid state storage unit, or a removable storage unit such as a USB storage unit installed in a laptop or computer that implements computer code performing the computer-implemented methods according to the present disclosure.

[0099] According to the present disclosure, the computer processing unit 400 of the present disclosure may be (part of) a laptop or computer, which may implement computer code performing the computer-implemented method according to the present disclosure. For example, the method of the present disclosure may be embodied in a computer program or product. This computer program or product includes computer coded instructions that, when the computer program or product program is executed by a computer, e.g. a laptop or computer, cause the computer to perform the steps of the computer-implemented method disclosed herein.

[0100] In a particular embodiment, a computer readable storage medium is proposed that includes computer coded instructions stored therein, which, when executed by a computer, cause the computer to perform the steps of the computer implemented method disclosed herein. Such a computer readable storage medium may be a (solid state) hard drive, or a USB drive, or a (digital) optical disk.

[0101] In an alternative example, one or more machine learning algorithms 402 may be embedded within the data transformation unit 401, thus forming a single computational unit performing several distinct steps of the computer-implemented method according to the present disclosure.

[0102] Having determined the physiological state of the analyte 1 , the results may be output or displayed by a data output unit 403 , for example a computer or laptop display, or a separate display unit 403 .

[0103] In order to identify the physiological state of a sample of analyte 1 using one or more machine learning algorithms 402, it is essential that the acquired MRI images are generated from a sample of analyte 1, preferably contained in an open 3D sample matrix 10 located in a measurement chamber 110a of the microfluidic device 100 or injected with a hyperpolarized substrate 6 contained in a vial, as this has the advantage of increased sample sensitivity compared to standard techniques. The present invention requires identifying significant patterns in the acquired MRI images that serve as indicators for the presence or occurrence of a physiological state of analyte 1.

[0104] Thus, in one example, the machine learning algorithm 402 may include a computer-implemented artificial neural network. In such a particular example, the computer processing unit 400 further includes a training unit (not shown) configured to train the computer-implemented artificial neural network 402 with sequences of time-lapse signals characterized by training sequences of time-lapse signals for known physiological states of the analyte injected with the hyperpolarized base, apply sequences of input signals to the computer-implemented artificial neural network characterized by at least one test sequence of time-lapse signals for unknown physiological states of the analyte injected with the hyperpolarized base (which relates to the training data), analyze each applied test sequence of time-lapse signals, and generate, for each test sequence of time-lapse signals, a predicted physiological state of the analyte injected with the hyperpolarized base.

[0105] To aid in training the computer-implemented artificial neural network 402, the NMR / MRI images are first annotated or labeled by an expert. These annotations or labels in the NMR / MRI images indicate significant patterns that correspond to external stimulus responses (such as drugs) to the cell / organ / analyte 1 under question. These labeled significant patterns characterize training sequences of signals for known physiological states of the analyte 1 of interest, including specific hyperpolarized substrates 6.

[0106] In one example, these annotations or labels may be image colors that correspond to signal intensities that in turn indicate certain disease stages and thereby known physiological conditions of the analyte 1 of interest associated with an injection containing a particular hyperpolarized base 6.

[0107] After collecting a statistically significant data cohort of sufficient significant patterns thus labeled, the computer implemented artificial neural network 402 is trained over time on a training set of such signal sequences and then tested on a validation or test set of signal sequences having an unknown physiological state of the analyte injected with hyperpolarized substrate. Once the artificial neural network 402 model achieves acceptable accuracy, it will be able to automatically indicate significant patterns, such as NMR / MRI image signal intensities, in new sets of NMR / MRI images and convert these into a data format that can be incorporated into ongoing data processing, thereby adding further parameters and improving the quality of the final output.

[0108] In further examples, the systems and methods described herein may include the use of other imaging techniques, such as, but not limited to, PET imaging. In examples, the input training images used may include images obtained with other imaging techniques, such as, but not limited to, positron emission tomography (PET) imaging and other molecular imaging techniques. In such examples, the systems and methods include acquiring sample images of analyte 1 with the imaging technique used to acquire images used in training the machine learning algorithm 402.

[0109] Thus, the artificial neural network 402 can detect the disease stage (the physiological state of the analyte being analyzed) if the metabolic changes in the NMR / MRI images correlate with these trained disease stages (e.g., signal intensity).

[0110] One of the key advantages of the system 1000 combining the microfluidic device 100 with the hyperpolarization (HP) production apparatus 200 and the NMR apparatus 300 is the ability to emulate the response of real organs, either of animal or human origin, to external stimuli under in vitro conditions. The more accurate the emulation, the higher the value of the results. The enormous complexity of the inputs involved in creating an accurate emulation and characteristic properties of the test conditions makes it technically demanding and requires the implementation of a computer processing unit 400 utilizing one or more computer implemented artificial neural networks 402.

[0111] In terms of implementation, the artificial neural network 402 is trained with all data derived from the raw materials used in the process stream, including all process parameters. The resulting hardware and analysis output 444 is compared to the actual responses to the same external stimuli in in vitro conditions for both animals and humans. Based on the comparison results on a cohort of statistically significant data, the artificial neural network 402 can determine which parameters contribute to a closer emulation of the actual cell / organ / analyte 1.

[0112] Ultimately, the computer implemented artificial neural network 402 can predict the parameters required to emulate the actual conditions of the cell / organ / analyte 1 under question for in vitro testing. In addition to this, a better understanding of the important parameters can help in building valid models of various other organs that are complex to construct.

[0113] Therefore, by comparing data obtained from cells / organs / analytes using the system according to the present disclosure consisting of a microfluidic device 100 housing one or more sample matrices 10 containing analytes 1, a hyperpolarization (HP) production apparatus 200 and an NMR apparatus 300 for in vitro and in vivo conditions, implementing an artificial neural network 402 helps to bridge the gap that is nearly impossible to establish using current technology.

[0114] The measurement chambers 110a of the multiple microfluidic elements 110 of the microfluidic device 100 allow to test multiple responses of cells / organs / analytes 1 to the same external stimulus, i.e. the same hyperpolarized agent 6 in the target area 310 and the same set of applied magnetic field gradients and RF pulses. The aim is to understand the effect of the external stimulus on other cells / organs / analytes in addition to the main cell / organ / analyte 1 under investigation. Given the complexity of the underlying biological processes, advanced methods such as artificial intelligence are highly beneficial in understanding the correlation of responses in various cells / organs / analytes and ultimately predicting the correlations and hence the responses.

[0115] In terms of implementation, a computer implemented artificial neural network model 402 is trained on the responses of the primary and secondary organs to external stimuli. After training on a statistically significant cohort of data, the computer implemented artificial neural network model 402 predicts the response of the secondary organs to the same external stimuli (specifically the hyperpolarizing agent 6 used) that have been tested on the primary organs. These results help researchers understand and target the underlying biological mechanisms causing these effects.

[0116] System 1000 creates a versatile platform capable of acquiring and analyzing metabolic data on the physiological state of analytes in real time using HP and MR techniques in a non-invasive manner, which can achieve the development of more advanced functional human-specific drug testing systems.

Claims

1. 1. A system for determining a physiological state of at least one analyte using hyperpolarized (HP) nuclear magnetic resonance, the system comprising: a hyperpolarized (HP) production device configured to hyperpolarize the substrate; A sample matrix for containing the analyte; a microfluidic device comprising at least one measurement chamber structured to receive the sample matrix containing the analyte, the microfluidic device being structured to inject the at least one analyte contained in the sample matrix comprising the hyperpolarized substrate; an NMR apparatus comprising at least a housing defining a target region for containing at least the analyte containing the hyperpolarized substrate, at least one magnet unit, at least one magnetic gradient unit for applying one or more magnetic field gradients in the target region during use, at least one radio frequency (RF) pulse generating unit for applying one or more sets of RF pulses towards the target region, and an RF receiving unit for acquiring signals; a computer processing unit for analyzing the acquired signals to determine the physiological state of the analyte.

2. The system of claim 1 , wherein the hyperpolarization (HP) producing device is a hyperpolarization device, such as a dynamic nuclear polarization (DNP) device or a parahydrogen induced polarization (PHIP) device.

3. the at least one measurement chamber is in fluid communication with a matrix perfusion supply line for supplying a matrix perfusion medium to the measurement chamber, a hyperpolarized base supply line for supplying the hyperpolarized base to the measurement chamber, and a chamber outlet line, preferably the hyperpolarized base supply line being disposed at a position above the position of the matrix perfusion supply line with respect to a bottom wall of the measurement chamber; 3. A system according to claim 1 or 2.

4. the chamber outlet line is at a position lower than the position of the hyperpolarized base supply line; and a matrix perfusion supply line disposed at a position higher than the position of the matrix perfusion supply line relative to the bottom wall of the measurement chamber; Preferably, the matrix perfusion supply line comprises: The system of claim 3 , wherein the system is positioned on the bottom wall of the measurement chamber.

5. The microfluidic device comprises: In the NMR device a temperature control unit for maintaining at least the analyte containing the hyperpolarized substrate at body temperature when contained therein. The system of claim 1 .

6. the computer processing unit is configured to analyze at least the acquired signals using one or more machine learning algorithms; The system of claim 1 .

7. The machine learning algorithm, 1. A computer-implemented artificial neural network, the computer processing unit comprising: The known physiological state of the analyte contained in the hyperpolarized analyte agent A time-lapse signal characterized by a training sequence of time-lapse signals. training said computer implemented artificial neural network with a sequence of applying a sequence of input signals to said computer-implemented artificial neural network, said input signals characterized by at least one test sequence of signals over time for an unknown physiological state of an analyte contained in a hyperpolarized analyte agent; 7. The system of claim 6, further comprising a training unit configured to: analyze each applied test sequence of time-lapse signals; and generate, for each test sequence of time-lapse signals, a predicted physiological state of an analyte contained in the hyperpolarized analyte agent.

8. Using the system according to claim 1, 1. A method for determining a physiological state of an analyte containing a hyperpolarized substrate, comprising: The method further comprising: a) adding at least one analyte to a sample matrix; b) injecting said at least one analyte containing a hyperpolarized substrate; c) using at least one magnet unit and at least one magnetic gradient unit of said NMR apparatus; applying one or more magnetic field gradients to the analyte containing the hyperpolarized substrate; d) using at least one radio frequency (RF) pulse generating unit of said NMR apparatus; applying one or more sets of RF pulses to the target region; e) receiving from an RF receiving unit of the NMR apparatus, in response to the one or more sets of RF pulses being applied; receiving a sequence of NMR signals over time; f) By time-frequency transformation, generating one or more time-frequency representations of said sequence of NMR signals; g) using one or more machine learning algorithms to determine the physiological state of the analyte; and analyzing the time-frequency representation.

9. the one or more machine learning algorithms 9. The method of claim 8, wherein the model is selected from the group including, but not limited to, an artificial neural network, a decision tree, a regression model, a k-nearest neighbor model, a partial least squares regression, a support vector machine, or an ensemble of models combined to define the algorithm.

10. The machine learning algorithm, A computer-implemented artificial neural network, The analysis step g) i) using a sequence of NMR signals over time characterized by a training sequence of NMR signals over time for a known physiological state of an analyte comprising a hyperpolarized substrate; training said computer-implemented artificial neural network; ii) for an unknown physiological state of an analyte containing a hyperpolarized substrate; characterized by at least one test sequence of NMR signals over time. applying a sequence of NMR input signals to said computer implemented artificial neural network; 10. The method of claim 8, wherein the step of: iii) analyzing each applied test sequence of time-lapse NMR signals and generating, for each test sequence of time-lapse NMR signals, a predicted physiological state of the analyte including the hyperpolarized substrate.

11. The receiving step b) iv) occurring after the step of hyperpolarizing the base material in a hyperpolarized (HP) production device; The method according to claim 8.

12. Although after step b), Prior to step c), the method further comprises: b1) maintaining the analyte containing the hyperpolarized substrate at body temperature; The method of claim 9 further comprising:

13. 2. The system of claim 1, wherein the sample matrix is ​​formed as a three-dimensional construct containing open pores, the construct being a gel comprising at least sodium carboxymethylcellulose.

14. The gel is 14. The system according to claim 13, comprising at least 0.5-5% by weight of sodium carboxymethylcellulose, in particular 1% by weight of sodium carboxymethylcellulose, and / or said gel is suitable for cryopreservation.

15. 9. The method of claim 8, wherein the sample matrix is ​​formed as a three-dimensional construct containing open pores, the construct being a gel comprising at least sodium carboxymethylcellulose.

Images