Multi-modal smart adhesive bandage for ai-based quantification of pain

Abstract

A method for determining an objective pain score for pain experienced by a patient includes placing adhesive bandages on a patient, each of which has multiplexed sensors and providing information from the sensors to a scoring agent that has been trained through contrastive learning to provide the objective pain score based on the information from the sensors.

Description

[0001] MULTI-MODAL SMART ADHESIVE BANDAGE FOR AI-BASED QUANTIFICATION OF PAIN

[0002] RELATED APPLICATIONS

[0003] This application claims the benefit of the July 9, 2024 priority date of U.S. Provisional Application 63 / 668,949, the contents of which are incorporated herein by reference.

[0004] BACKGROUND

[0005] Pain arises as a result of various mechanisms, among which are nociceptive pain, neuropathic pain, and pain due to central nervous system sensitization.

[0006] Unfortunately, it is difficult to provide an objective way of measuring pain. In general, one asks the patient. But the patient’s response is often subjective. Different patients express pain in different ways based on various factures such as differences in cultural background, past experiences, psychological state, and social context.

[0007] Each person possesses a unique pain tolerance level. This tolerance level defies quantification. Despite significant advancements in understanding and managing pain, current practices for monitoring pain rely a great deal on questionnaires and surveys. These lack objectivity due to their heavy reliance on self-reported pain. They are also susceptible to influences arising from societal expectations and cultural norms.

[0008] Current practices often overlook the distinct physiological pain conditions specific to pain such as endometriosis / fibromyalgia, as well as hormonal factors like menstrual cycle, pregnancy, and menopause, which affect pain sensitivity and perception.

[0009] Socioeconomic status significantly affects pain management. Individuals of lower socioeconomic standing face barriers to accessing healthcare, including pain management services. This results in delayed or inadequate pain treatment, which in turn increases suffering and compromises health outcomes.

[0010] SUMMARY

[0011] The invention provides a basis for overcoming the foregoing limitations by providing a simple quantifiable metric for chronic pain that takes into account the diverse conditions and experiences of patients and that allows for a more personalized and effective pain management. Moreover, the invention provides a way for clinicians to track pain more objectively and longitudinally in response to different treatments. An advantage of the methods and devices described herein arises from their portability. This, in turn, improves their accessibility, thus improving their impact and significance.

[0012] An apparatus for measuring and quantifying pain relies on monitoring multiple cardiorespiratory and neuropathic responses using wearable sensors to monitor a combination of biophysical and biochemical markers. A suitable platform comprises first and second adhesive bandages that incorporate therein various sensors. The first adhesive bandage is one for continuous and daily use for monitoring cardiorespiratory and sympathetic nervous system responses indicative of pain. The second adhesive bandage is a single-use self-powered biochemical sensing patch that monitors a suite of chemical biomarkers in interstitial fluid. The use of interstitial fluid is useful as a non-invasivc surrogate for blood.

[0013] In one aspect, the invention features a method that includes placing a first adhesive bandage on the patient, the first adhesive bandage having first multiplexed sensors integrated therein, placing a second adhesive bandage on the patient, the second adhesive bandage having second multiplexed sensors integrated therein, receiving a first signal and receiving a second signal. The first signal, which is from the first multiplexed sensors, contains information indicative of a first set of biomarkers for pain. The second signal, which is from the second multiplexed sensors, contains information indicative of a second set of biomarkers for pain. The biomarkers include biochemical markers for pain and biophysical markers for pain. The method continues with the step of providing the first signal and the second signal to a scoring agent that has been trained to determine an objective pain score that represents pain being experienced by the patient, the scoring agent having been trained by contrastive learning to determine the objective pain score based on the biomarkers for pain. The scoring agent then uses information from the first and second signals to transmit a signal that includes the objective pain score. In some practices, the first and second signals are wireless signals. Among these practices are those in which receiving the first and second signals includes receiving the first and second signals at a local processing device that is in wireless data communication with the first and second multiplexed sensors and that is data communication with the scoring agent.

[0014] In other practices, the scoring agent is disposed at a cloud server. In such practices, the method further includes transmitting a third signal to the cloud server, the third signal including information from the first and second signals.

[0015] Practices further include those in which placing the first adhesive bandage on the patient includes placing the first adhesive bandage on the patient’s chest.

[0016] Also among the practices of the method arc those in which placing the second adhesive bandage on the patient includes placing the second adhesive bandage on an arm of the patient, those in which placing the second adhesive bandage on the patient includes causing plural hollow needles to penetrate the patient such that the plural hollow needles extend into the patient’s dermis to draw interstitial fluid from the patient’s arm, and those in which placing the second adhesive bandage on the patient includes causing the second multiplexed sensors to be in fluid communication with a dermis of the patient and with interstitial fluid drawn from the dermis of the patient.

[0017] In some practices, receiving the second signal includes receiving information about the biochemical markers for pain.

[0018] Other practices include those in which receiving the first signal includes receiving information about the patient’s heart-rate variability, the patient’s skin conductance, and the patient’s respiration, those in which receiving the first signal includes receiving information about the biophysical markers for pain, and those in which receiving the first signal consists of receiving information about biophysical markers for pain and wherein receiving the second signal consists of receiving information about biochemical markers for pain. Still other practices include using contrastive loss to train an encoder to map information concerning the pain markers in the suite of pain markers into embeddings in an embedding space. In such practices, the scoring agent uses the embeddings as a basis for determining the objective pain score.

[0019] In still other practices, the scoring agent includes an encoder. In such practices, after having trained the encoder using a self-supervised feature extraction, the method includes an additional training step of carrying out a fine-tuning process using labeled that that has been curated to avoid individual biases in pain reporting.

[0020] In those cases in which the scoring agent includes an encoder and a classifier, practices include those that include, after having trained the encoder using a selfsupervised feature extraction, using pain questionnaires as a basis for labeling a corpus of training data and using the labelled corpus of training data to fine tunc the classifier to identify boundaries generated by the encoder.

[0021] Other practices include training the classifier by carrying out a domain-adjustment process to reduce variability of the classifier caused by a subjective impression of pain experienced by the patient or by carrying out a domain-generalization process to reduce variability of the classifier caused by a subjective impression of pain experienced by the patient.

[0022] The various practices claimed herein are limited strictly to non-abstract implementations, where “non-abstract” is hereby defined to be the converse of “abstract” as that term has been construed by the courts of the United States as of the filing of this application. Any party who construes the claims as covering abstract implementations will simply be demonstrating that it is possible to construe the claims in a manner contrary to the specification.

[0023] It has further been discovered experimentally that it is not possible to carry out all the steps of the foregoing methods mentally. Accordingly, the claims cannot be construed as covering mental steps. Any person who construes the claims as reading on mental steps will simply be demonstrating the possibility of construing claims contrary to the specification.

[0024] In another aspect, the invention features a non-abstract apparatus including first and second adhesive bandages having corresponding first and second pluralities of multiplexed sensors integrated therein, respectively. The apparatus also includes a receiver for receiving a first signal and a second signal. The first signal, which is from the first multiplexed sensors, contains information indicative of a first set of biomarkers for pain. The second signal, which is from the second multiplexed sensors, contains information indicative of a second set of biomarkers for pain. The biomarkers for pain include both biochemical markers for pain and biophysical markers for pain. The apparatus also includes a scoring agent that has been trained been trained to determine an objective pain score that represents pain being experienced by the patient. This scoring agent has been been trained by contrastive learning to determine the objective pain score based on the biomarkers for pain.

[0025] In some embodiments, the scoring agent includes an encoder that is configured to embed features obtained pain markers into an embedding space and a classifier that provides a basis for determining the objective pain score using features from the first and second signals.

[0026] In other embodiments, the scoring agent includes nodes arranged in first and second layers, each of said nodes including a summation circuit that outputs a weighted sum of voltages present at inputs thereof. The outputs of nodes in said first layer are used as inputs of nodes in said second layer.

[0027] In still other embodiments, the scoring agent is implemented as a neural network.

[0028] It has been found that a generic computer is unable to include the foregoing features. Accordingly, the apparatus is implemented either on a non-generic computer or using application-specific digital circuitry. The apparatus as claimed is limited to a nonabstract apparatus, where “non-abstract” is hereby defined as the converse of “abstract” as construed by the courts of the United States as of the filing date of this application. DESCRIPTION OF THE FIGURES

[0029] FIG. 1 shows a patient wearing a patient-mounted platform that is in communication with a scoring agent;

[0030] FIG. 2 shows a process for training the scoring agent of FIG. 1;

[0031] FIG. 3 shows a first sensor from the platform of FIG. 1;

[0032] FIG. 4 shows an alternative view of the first sensor from the platform in FIG. 1;

[0033] FIG. 5 shows pain-related biomarkers available to the second sensor from the platform in FIG. 1;

[0034] FIG. 6 shows an example of a second sensor from the platform of FIG. 1;

[0035] FIG. 7 shows details of a molecularly-imprinted polymer used in the second sensor shown in FIG. 6;

[0036] FIG. 8 shows another embodiment of the second sensor shown in FIG. 6;

[0037] FIG. 9 shows yet another embodiment of the second sensor shown in FIG. 6;

[0038] FIG. 10 shows an alternative embodiment of the second sensor shown in FIG. 6;

[0039] FIG. 11 shows steps in making a needle patch as shown in FIG. 10; and

[0040] FIG. 12 shows a variety of sensors for use in the platform of FIG. 1.

[0041] DETAILED DESCRIPTION

[0042] FIG. 1 shows a patient 10 wearing a patient-mounted platform 12 for obtaining information about markers that are known to be indicative of the presence of pain. These markers include biophysical markers and biochemical markers.

[0043] The platform 12 has a first sensor 14 and a second sensor 16, both of which are integrated into corresponding adhesive bandages. The first sensor 14, which adheres to the patient’s chest, monitors a first band of markers. The second sensor 16, which adheres to the patient’s arm, monitors a second band of markers. In the illustrated embodiment the first band comprises biophysical markers and the second band comprises biochemical markers that are present in the patient’s interstitial fluid. As used herein, a “band” is a set whose elements comprise one or more “markers.”

[0044] The first sensor 14 provides a biophysical-marker signal to a portable device 18. Similarly, the second sensor 16 provides a biochemical-marker signal to the portable device 18. A suitable portable device 18 is a smartphone.

[0045] The biophysical-marker signal includes information about biophysical markers measured by the first sensor 14. The biochemical-marker signal includes information about biochemical markers measured by the second sensor 16.

[0046] The portable device 18 provides the information about the suites of markers to a scoring agent 20 that has been trained using machine-learning methods to objectively quantify pain experienced by the patient 10 and to do so based on the suite of signals provided by the first and second sensors 14, 16. In a typical embodiment, the scoring agent 20 is implemented on a cloud-based server 22. In response to receiving the scoring agent 20 outputs an objective pain score.

[0047] In the illustrated embodiment, the first sensor 14 measures biophysical markers and the second sensor 16 measures biochemical markers. This is a convenient arrangement because of the differences in the construction of the two sensors 14, 16. However, the scoring agent 20 does not care which sensor measured which markers or how many sensors exist.

[0048] The platform 12 is inherently modular because the first and second sensors 12, 14 are easily replaced so that different markers can be measured by simply replacing the relevant sensors 12, 14. As a result, the platform 12 is simple to fabricate and easy to use. Because the platform 12 transmits data to a portable device 18 and ultimately to a cloud server 22, there is no need for expensive storage on the platform 12 itself. Additionally, by avoiding reliance on a single pain marker and instead using a suite of pain markers, the scoring agent 20 provides a more robust pain score. Referring now to FIG. 2, the process of training the scoring agent 20 includes a feature-extraction process 24, a fine-tuning process 26, and a domain-adjustment process 28.

[0049] The feature-extraction process 24 relies on the observation that markers that are correlated with pain are similar across individuals. For example, when a person is stricken with sudden pain, it is common for the heart rate to change or for the person to suddenly gasp. It is also common for certain biochemicals to increase in concentration. What tends to differ is what a person says about the experience of pain, i.e., the person’s subjective tolerance for pain.

[0050] The feature-extraction process 24 addresses this difficulty by using contrastive representation in combination with machine learning to train the scoring agent 20 to use features that arc relatively invariant to subjective tolerances and to then fine tunc the scoring agent’s training by incorporating individual perception in the fine-tuning process 26 that follows.

[0051] The feature-extraction process 24 trains an encoder 30. It does so by using biomarker data and contrastive information concerning whether pain levels are similar or not. The feature-extraction process 24 thus provides an alternative to the more subjective pain levels reported by human subjects.

[0052] Since contrastive information relies only on contrast, the feature-extraction process 24 carries out a form of training with weak supervision or self-supervision. The resulting self- supervised feature-extraction process 24 uses self-supervised feature extraction on an unlabeled corpus of training data to generate a suite of markers that are to be used by the scoring agent 20 for providing an objective pain score based on measurements obtained by the first and second sensors 14, 16 that are mounted on the patient 10.

[0053] The feature-extraction process 24 relies on contrastive loss as a way to update the encoder 30. The use of contrastive loss as a loss function permits the encoder 30 to learn useful embeddings by contrasting similar items against dissimilar items, thereby urging the encoder 30 to produce embeddings in which similar items are close together in the dissimilar items are far apart in an embedding space. By using contrastive loss as an updating mechanism, the feature-extraction process 24 trains the encoder 30 to identify the distinct classes that will ultimately serve as a basis for an objective pain score.

[0054] The fine-tuning process 26 that follows fine tunes the output of the featureextraction process 24 based on labeled data that has been curated to avoid individual biases and socio-economic factors in pain reporting. This fine-tuning process 26 includes the use of pain questionnaires as a basis for labelling a small corpus of training data to fine tune a classifier 32 that identifies boundaries within the classes generated by the encoder 30. This fine-tuning process 26 thus causes the classifier 32 to take into account, to a limited extent, the subjective impressions of individual subjects.

[0055] It has been discovered that the subjective impression of pain as experienced by a patient 10 tends to vary as a function of the various demographic groups to which that subject belongs. As a result, a classifier 32 may perform more poorly on a domain that differs from that used to train it. The domain-adjustment process 28 executes a domain generalization process and / or a domain adaptation process to reduce variability of the classifier 32 when applied to patients who are in demographic groups other than those used to train the classifier 32.

[0056] Referring back to FIG. 1, the first sensor 14 is conveniently located to gather information about the manner in which the patient’s cardiovascular system, respiratory system, and sympathetic nervous system respond to pain. In particular, the first sensor 14 obtains information about heart-rate variability, changes in skin conductance, including both the tonic component and the phasic component thereof, and mechanical movement indicative of activity by the respiratory system.

[0057] As shown in FIGS. 3 and 4, the first sensor 14 includes first, second, and third electrodes 34, 36, 38. The electrodes 34, 36, 38 cooperate for monitoring heart rate, heartrate variability, and electrodermal response. Each electrode 34, 36, 38 comprises a eutectic gel coupled to silver or silver chloride. The use of eutectogel for long-term monitoring is advantageous because eutectogel tends to be non-toxic, to be unlikely to cause irritation, and to not dissipate as a result of volatility thereof. A eutectogel fiber 40 connects the first and second electrodes 34. 36. As the patient 10 breathes, this fiber 40 stretches and relaxes. This, in turn, provides a basis for measuring respiration.

[0058] A flexible printed-circuit board 42 couples to the various electrodes 34, 36, 38 using one or more cross-linked polydimethylsiloxane vias 44. The printed-circuit board, in turn, is attached to a fabric 46.

[0059] The second sensor 16 provides measurements of any one of a number of biochemical markers that have been observed to be present when pain is present. These arc summarized in the table shown in FIG. 5.

[0060] Referring now to FIG. 6, the second sensor 16 comprises hollow needles 48 enclosed in a casing 50. In a preferred embodiment, the casing 50 is a flexible polycarbonate casing having a one-way valve 52 passing therethrough.

[0061] The needles 48 penetrate into the patient’s dermis. These needles 48 sample interstitial fluid. This interstitial fluid serves as a close surrogate to blood serum. The interstitial fluid is drawn up the needles 48 via capillary action. In those embodiments having a flexible casing 50, the lower pressure that builds up in the casing 50 promotes passive suction and further assists in drawing interstitial fluid through the needles 48. The interstitial fluid is drawn towards a set of one or more biochemical-marker detectors 54, each of which is configured to detect a particular biochemical marker. Referring now to FIG. 7 one implementation of a biochemical-marker detector 54 comprises a bilayer electrode 56 having a polymer layer 58 disposed on a metal layer 60. The polymer layer 58 comprises a molecularly-imprinted polymer having cavities that are selective to bind to a particular biochemical marker. FIG. 7 shows the biochemical-marker detector before and after having been exposed to the particular biochemical marker for which it has been configured to bind. The binding of the biochemical marker to the polymer layer 58 modulates an electrical property of the bilayer electrode 56. The modulation of this electrical property is detectable using a variety of modes, among which are amperometry and electrochemical impedance spectroscopy. In some embodiments, the electrical property is the bilayer electrode’s charge-transfer resistance.

[0062] The use of a bilayer electrode 56 provides femtomolar sensitivity to a biochemical marker in a complex biological fluid, such as serum, saliva, or urine. In addition, a bilayer electrode 56 is easily synthesized to accommodate newly discovered biochemical markers for pain.

[0063] FIG. 8 illustrates another example of a biochemical-marker detector 54. The illustrated biochemical-marker detector 54 includes a sample zone 62, a first detection zone 64, and a second detection zone 66, both of which are in fluid communication with the sample zone 62. Each detection zone has a proximal end that begins at the sample zone 62. Each detection zone 64, 66 extends distally away from this proximal end.

[0064] The first and second detection zones 64, 66 are functionalized with first and second reagents to detect first and second biochemical markers. The sample zone 62 and the first and second detection zones 62, 66 are formed on a paper substrate. As a result, fluid in the sample zone 62 wicks its way into the detection zones 64, 66.

[0065] A needle 48 includes a proximal end in fluid communication with the sample zone 62 and a distal end that penetrates the epidermis 68 so as to be placed in contact with interstitial fluid 70 in the dermis 72. In some embodiments, the needles 48 are hydrogel microneedles.

[0066] In operation, the needles 48 penetrate the skin. This causes transport of interstitial fluid from beneath the skin towards the sample zone 62. This transport also brings at least first and second biochemical markers along with it. From the sample zone 62, the interstitial fluid proceeds to the proximal ends of the first and second detection zones 64, 66 and travels distally along the first and second detection zones 64, 66.

[0067] As interstitial fluid traverses the first detection zone 64, the first biochemical marker reacts with the first reagent in the first detection zone 64. This causes a segment of the first detection zone 64 to change color. This segment begins at the proximal end and extends distally along the first detection zone 64 by some length that can be measured. The length of this colored segment provides a basis for inferring the concentration of the first biochemical marker.

[0068] Similarly, as interstitial fluid traverses the second detection zone 66, the second biochemical markers reacts with the second reagent in the second detection zone 66. This causes a segment of the second detection zone 66 to change color. This segment begins at the proximal end and extends distally along the second detection zone 66 by some length that can be measured. The length of this colored segment provides a basis for inferring the concentration of the second biochemical marker.

[0069] In some embodiments, the first and second biochemical markers are cortisol and dopamine, respectively. Among these embodiments are those in which the first and second reagents are tetramethylammonium hydroxide (TMAOH)Zblue tetrazolium and tetravalent cerium cation (Ce4+) solutions to cause a colorimetric response to the presence of cortisol and dopamine, respectively.

[0070] FIG. 9 shows an alternative embodiment of the biochemical-marker detector 54 in which a bivalve clamshell housing 74 comprises a sampling side 76 and a detection side 78 on either side of a fold line 80.

[0071] The sampling side 76 includes a sampling zone 82 made of a paper substrate. This sampling zone 82 contacts proximal ends of plural hydrogel needles 48. Each needle 48 extends away from the sampling side 76 and terminates at a distal end thereof. The distal ends of the needles 48 penetrate a patient’s skin 48 and thus draw interstitial fluid 70 therefrom. The proximal ends of the needles 48 are in fluid communication with the sampling zone 82. As a result, interstitial fluid 70 drawn by the needles 48 reaches the sampling zone 82.

[0072] The detection side 78 features a paper substrate on which is formed a detecting zone 84 that comprises molecularly imprinted polymers and carbon nanodots integrated therein. These are selected such that binding of the biochemical marker to the polymer tends to quench fluorescence of the carbon nanodots to an extent that depends on the concentration of the biochemical marker.

[0073] An assay for the analyte begins with folding the housing 74 along its fold line 80 so that the sampling zone 82 and detecting zone 84 come into contact. The housing 74 is then placed in fluid communication with the hydrogel needles 48 to draw interstitial fluid 70 into the sampling zone 82. After a brief sampling interval to allow interstitial fluid 70 to flow from the distal ends of the needles 48 to the sampling zone 82 and onward to the detecting zone 84, the process continues with unfolding the housing 74 thereby detaching the sampling zone 82 and the detecting zone 84 from each other and permitting visual inspection of the detecting zone 84.

[0074] Binding of biochemical marker to the molecularly imprinted polymer quenches the fluorescence of the carbon nanodots as a result of surface adsorption and phot- induced electron transfer. The extent to which such quenching occurs depends on the concentration of the biochemical marker. As a result, by observing the fluorescence, it is possible to infer the concentration of the biochemical marker by observing the extent of fluorescence in the sampling zone 82.

[0075] In another embodiment, shown in FIG. 10, the second sensor 16 relies on solid needles 48 instead of hollow needles as in FIG 6. In this embodiment, the needles 48 comprise a material that draws interstitial fluid 70 spontaneously without the need for passive suction as described in connection with FIG. 6. These needles extent from a nonwoven substrate 86. A suitable needle 48 is a sodium poly(acrylate) needle. Such needles 48 are useful for the embodiments shown in FIG. 7 and FIG. 8. FIG. 11 shows a process for making needles 48 of the type shown in FIG. 10. The process begins with casting PDMS on an acrylic mold. The molded PDMS is then detached, treated with oxygen plasma, and silanized. The resulting structure is used to construct a PDMS mold which can then be used to receive a casting of methacrylate hyaluronic acid maltose solution, which is then exposed to light to promote cross linking. A non-woven substrate is then disposed on the base ends of the resulting needles, photocrosslinked, and detached to form a patch having needles 48 extending therefrom.

[0076] FIG. 12 shows examples of sensors for use in one or more of the first and second sensors 14, 16. The illustrated examples are: (a) a needle array platform for interstitial fluid sampling, (b) a transdermal sensing-patch for monitoring electrolytes and metabolites using thread electrodes, (c) a cortisol-sensing disposable sensor, (d) a flexible paper-based ECG electrode, (e) a disposable paper-based sensor for cytokines detection, (f) a eutectogel-based flexible sensor, (g) a sensor correlation with salimetric and enzyme-linked immunosorbent assay techniques, and (h) an eMIP-based oxytocin sensor.

[0077] The combination of the platform 10 and the scoring agent 20 in communication therewith thus provides an objective basis for quantifying pain experienced by a patient 10 in part based on the mining of diverse pain-related biophysical and biochemical markers. The resulting multi-modal monitoring of a diverse panel of biophysical and biochemical markers related to pain provides a more accurate and reproducible basis for evaluating chronic pain. Since pain can be measured without having to go to a clinic, the platform 10 reduces the cost and burden of travel to centralized facilities. Moreover, the sensors that measure biochemical markers are able to use synthetic recognition elements, thus providing greater shelf-stability, accuracy, selectivity, and tunability for detecting a wide range of analytes.

[0078] Among the platform’s advantages is that of using a flexible renewable textile substrate and the use of common and inexpensive textile processing methods to fabricate the adhesive bandage that incorporates the sensors 12, 14. This considerably lowers cost and increases access. Wireless cloud connectivity will facilitate active caregiver engagement using tele-health allowing them to provide timely care to the patient’s pain issues. Beyond the immediate impact of helping patients manage their pain, this platform holds immense value for discovering new treatments. For example, it will create a universal benchmark to compare the effectiveness of different treatments. As more and more patients utilize the platforms for managing pain, the data will help researchers gain a deeper understanding of different pathways that mediate pain and how it relates to the unique physiology and circumstances. This will help identify novel drug candidates for pain treatment.

[0079] Having described the invention and a preferred embodiment thereof, what is claimed as new and secured by letters patent is:

Claims

CLAIMS1. A method comprising: placing a first adhesive bandage on said patient, said first adhesive bandage having first multiplexed sensors integrated therein, placing a second adhesive bandage on said patient, said second adhesive bandage having second multiplexed sensors integrated therein, receiving a first signal, wherein said first signal is from said first multiplexed sensors, wherein said first signal contains information indicative of a first set of biomarkers for pain, and wherein said first set of biomarkers comprises biomarkers for pain selected from the group consisting of biochemical markers for pain and biophysical markers for pain, receiving a second signal, wherein said second signal is from said second multiplexed sensors, wherein said second signal contains information indicative of a second set of biomarkers for pain, and wherein said second biomarkers comprise biomarkers for pain selected from the group consisting of biochemical markers for pain and biophysical markers for pain, providing said first signal and said second signal to a scoring agent that has been trained to determine an objective pain score that represents pain being experienced by said patient, said scoring agent having been trained by contrastive learning to determine said objective pain score based on said biomarkers for pain, and causing said scoring agent to use information from said first and second signals to transmit a signal that includes said objective pain score.

2. The method of claim 1, wherein said first and second signals are wireless signals and wherein receiving said first and second signals comprises receiving said first and second signals at a local processing device that is in wireless data communication with said first and second multiplexed sensors and that is data communication with said scoring agent.

3. The method of claim 1, wherein said scoring agent is disposed at a cloud server and wherein said method further comprises transmitting a third signal to said cloud server, said third signal comprising information from said first and second signals.

4. The method of claim 1, where placing said first adhesive bandage on said patient comprises placing said first adhesive bandage on said patient’s chest.

5. The method of claims 1 or 4, wherein placing said second adhesive bandage on said patient comprises placing said second adhesive bandage on an arm of said patient.

6. The method of claim 1, wherein placing said second adhesive bandage on said patient comprises causing plural hollow needles to penetrate said patient such that said plural hollow needles extend into said patient’s dermis to draw interstitial fluid from said patient’s arm.

7. The method of claim 1, wherein receiving said second signal comprises receiving information about said biochemical markers for pain.

8. The method of claim 1, wherein receiving said first signal comprises receiving information about said patient’s heart-rate variability, said patient’s skin conductance, and said patient’s respiration.

9. The method of claim 1, wherein receiving said first signal comprises receiving information about said biophysical markers for pain.

10. The method of claim 1, wherein placing said second adhesive bandage on said patient comprises causing said second multiplexed sensors to be in fluid communication with a dermis of said patient and with interstitial fluid drawn from said dermis of said patient.

11. The method of claim 1, wherein receiving said second signal comprises receiving information about said biochemical markers for pain.

12. The method of claim 1, wherein receiving said first signal consists of receiving information about biophysical markers for pain and wherein receiving said second signal consists of receiving information about biochemical markers for pain.

13. The method of claim 1, further comprising using contrastive loss to train an encoder to map information concerning said pain markers in said suite of pain markers into embeddings in an embedding space, wherein said scoring agent uses said embeddings as a basis for determining said objective pain score.

14. The method of claim 1, wherein said scoring agent comprises an encoder, said method further comprising, after having trained said encoder using a selfsupervised feature extraction, carrying out a fine-tuning process using labeled that that has been curated to avoid individual biases in pain reporting.

15. The method of claim 1, wherein said scoring agent comprises an encoder and a classifier, said method further comprising, after having trained said encoder using a self- supervised feature extraction, using pain questionnaires as a basis for labeling a corpus of training data and using said labelled corpus of training data to fine tune said classifier to identify boundaries generated by said encoder.

16. The method of claim 1, wherein said pain score is that of said patient, said method further comprising training a classifier, wherein training said classifier comprises carrying out a domain-adjustment process to reduce variability of said classifier caused by a subjective impression of pain experienced by said patient.

17. The method of claim 1, wherein said pain score is that of said patient, said method further comprising training a classifier of said scoring agent, wherein training said classifier comprises carrying out a domain-generalization process to reduce variability of said classifier caused by a subjective impression of pain experienced by said patient.

18. An apparatus comprising: a first adhesive bandage, first multiplexed sensors integrated into said first adhesive bandage, a second adhesive bandage, second multiplexed sensors integrated into said second adhesive bandage, a receiver for receiving a first signal and a second signal, wherein said first signal is from said first multiplexed sensors, wherein said first signal contains information indicative of a first set of biomarkers for pain, wherein said first set of biomarkers comprisesbiomarkers for pain selected from the group consisting of biochemical markers for pain and biophysical markers for pain, wherein said second signal is from said second multiplexed sensors, wherein said second signal contains information indicative of a second set of biomarkers for pain, and wherein said second biomarkers comprise biomarkers for pain selected from the group consisting of biochemical markers for pain and biophysical markers for pain, and a scoring agent, said scoring agent having been trained been trained to determine an objective pain score that represents pain being experienced by said patient, said scoring agent having been trained by contrastive learning to determine said objective pain score based on said biomarkers for pain.

19. The apparatus of claim 18, wherein said scoring agent comprises an encoder that is configured to embed features obtained pain markers into an embedding space and a classifier that provides a basis for determining said objective pain score using features from said first and second signals.

20. The apparatus of claim 18, wherein said scoring agent comprises nodes arranged in first and second layers, each of said nodes comprising a summation circuit that outputs a weighted sum of voltages present at inputs thereof, wherein outputs of nodes in said first layer are used as inputs of nodes in said second layer.

Images