Volatile biomarker detector

Abstract

The disclosed device includes a sensor array having sensing materials configured to react to a target analyte / material as well as a background gas from ambient air. The device also includes a controller to read sensor data from the sensor array to detect a presence of the target material in the air sample. The device may be configured to continuously monitor air or be coupled to a sample chamber. The target material may correspond to a volatile organic compound (VOC) such as biomarkers for viruses. Various other methods, systems, and computer-readable media are also disclosed.

Description

VOLATILE BIOMARKER DETECTORGOVERNMENT INTEREST

[0001] This invention was made with government support under Grant # 2344028 awarded by the National Science Foundation (NSF). The government has certain rights in the invention.CROSS REFERENCE TO RELATED APPLICATION

[0002] This application claims the benefit of U.S. Provisional Application No. 63 / 698,245, filed September 24, 2024, the disclosure of which is incorporated, in its entirety, by this reference.BACKGROUND

[0003] The avian influenza virus (AIV), commonly known as bird flu, is among a group of influenza type A viruses. Influenza viruses may infect chickens, turkeys, pheasants, quail, ducks, geese, and guinea fowl, as well as a variety of other birds. Waterfowl and shorebirds are often a natural reservoir of all subtypes of AIV distributed across the globe and are often considered primarily responsible for the spread and maintenance of AIV in nature. However, wild waterfowl and shorebirds often do not exhibit clinical signs of infection in nature. Highly pathogenic avian influenza virus (HPAIV) strains may be extremely infectious, often fatal to domestic poultry, and may spread rapidly from flock-to-flock.

[0004] AIV outbreaks allow zoonotic transmissions to evolve. Farms may lose millions of poultry in a month from such outbreaks. These zoonotic transmissions may spread to other animals, such as goats, cows, and humans. Decreasing the spread and evolution of the virus may require early and rapid disease detection and monitoring of AIV.

[0005] Reverse transcription polymerase chain reaction (RT-PCR) is often considered the gold standard for detecting AIV, due to its ability to test numerous sample types with high specificity and sensitivity. However, this procedure may suffer from certain practical drawbacks. For example, the procedure often requires a one to four day turnaround using lab equipment and a technician. Further, the virus's high genetic variability increases the likelihood of a false negative. Accordingly, there is a need for systems and methods for real-time detection.1ACTIVE 714926490v1BRIEF DESCRIPTION OF THE DRAWINGS

[0006] The accompanying drawings illustrate a number of example implementations and are a part of the specification. Together with the following description, these drawings demonstrate and explain various principles of the present disclosure.

[0007] FIG. 1A is a diagram of an example detector device for detecting volatile compounds.

[0008] FIG. IB is a diagram of an example implementation of the device of FIG. 1A.

[0009] FIG. 1C is an internal diagram of another example implementation of the device of FIG. 1A.

[0010] FIG. 2 is a block diagram of an example computing device used in a detector device.

[0011] FIG. 3 is a table of example sensing materials.

[0012] FIG. 4 depicts devices for testing and training AIV detection systems.

[0013] FIG. 5 depicts an AIV detection device with a 16-sensor array.

[0014] FIG. 6 is a block diagram of an example analyte flow path for a detector device.

[0015] FIG. 7 is a flow diagram of an example method for detecting volatile compounds using a detector device.

[0016] Throughout the drawings, identical reference characters and descriptions indicate similar, but not necessarily identical, elements. While the example implementations described herein are susceptible to various modifications and alternative forms, specific implementations have been shown by way of example in the drawings and will be described in detail herein. However, the example implementations described herein are not intended to be limited to the particular forms disclosed. Rather, the present disclosure covers all modifications, equivalents, and alternatives falling within the scope of the appended claims.DETAILED DESCRIPTION

[0017] The present disclosure is generally directed to systems and methods for detecting volatile biomarkers rather than targeting genes (e.g., detecting AIV gene segments such as RNA and / or DNA). Studies have shown that certain compounds may be present for infected specimens (e.g., birds). More specifically, the ratios between particular compounds2ACTIVE 714926490v1may differ between infected specimens and non-infected specimens such that the compounds or ratios may be used as biomarkers.

[0018] An electronic nose (e-nose) may be a type of artificial detector using optical, electrochemical, and / or chemiresistive gas sensors, and / or other types of gas sensors (e.g., impedimetric sensors, colorimetric sensors, etc.). A chemiresistive e-nose may provide portable, low cost, user-friendly, and real-time multi-detection of volatile organic compounds (VOCs). The chemiresistive sensor may utilize a sensing material that changes electrical resistance when it interacts with target analytes and oxygen, in a process that may involve the adsorption of oxygen and target analytes, surface reactions, and charge transfer between the analytes, ionized oxygen species, and the sensing materials, followed by the desorption of products.

[0019] E-nose systems often detect the presence of VOCs through classificationbased machine learning (ML) algorithms, which may have data size limitations. Common ML algorithms / schemes in e-nose systems include principal component analysis (PCA), support vector machine (SVM), convolutional neural networks (CNN), decision trees, and random forests (RF).

[0020] As referred to herein artificial intelligence (Al) generally refers to machines (e.g., software and / or hardware) developed to think and / or act like humans. Further, although the examples herein relate to Al and / or ML models (e.g., programs and / or systems trained from input data to output predictions or decisions, including unsupervised, supervised, and / or reinforcement learning models), the AI / ML models referred to herein may correspond to any type of Al technology, including various types of machine learning, neural network (NN) (e.g., having a structure of input layers, one or more hidden layers, and an output layer, each layer performing calculations with input from a previous layer using trained / learned weights), deep learning (DL) (e.g., NN with more than three hidden layers), generative Al (GenAI) (e.g., Al trained to generate content such as text, images, video, audio, etc. as learned from existing content), large language model (LLM) (a GenAI for natural language processing to generate human-like text), etc. (e.g., any other Al model).

[0021] The systems and methods provided herein may relate to an e-nose system for identifying and quantifying the VOCs detected to determine the health status of the specimen through ML algorithms, two of which are described herein. Sensor array optimization and the variation of operating temperatures may allow obtaining unique sensing 3ACTIVE 714926490v1performances and features for ML at different adsorption and desorption stages, helping to address the cross-sensitivity and selectivity issues in e-noses.

[0022] The described e-nose system may include various sensors based on sensing materials that may be sensitive or otherwise selective towards particular organic or inorganic volatiles and / or insensitive to organic or inorganic volatiles in various combinations. A machine learning model may be trained to identify temperature and / or illumination parameters for improved sensing performance with the sensing materials, and to also identify AIV biomarkers from sensing data provided by the sensors. The described system may provide a modular form factor allowing flexible, efficient, and accurate AIV biomarker detection in various use cases.

[0023] The systems and methods described herein address challenges in rapidly diagnosing and monitoring animals in the environment. End users, such as farmers, animal husbandry technicians, packing plant inspectors, and others who come into direct contact with animals, may use the systems and methods provided herein to monitor the health of livestock and make critical decisions in real time. The present disclosure provides a low-cost and intelligent e-nose device that may detect AIV volatile biomarkers. The device may produce diverse sensing data with a superior low detection limit achieved through defect- and nano-engineered gas sensing materials. High dimensional sensing data may also be produced by individually modulating the operating temperature and illumination (e.g., wavelength and / or intensity) of the sensors in the e-nose. A machine learning model may be used to optimize the temperature and illumination profiles for the sensors. Another machine learning model may process the sensing data to identify the AIV volatile biomarkers with high sensitivity and specificity.

[0024] As will be described further below, the present disclosure provides various sensing materials, including but not limited to: a group of sensing materials that are highly sensitive and selective toward organic volatiles such as acetoin and l-octen-3-ol; a group of sensing materials that are sensitive and selective toward to inorganic volatiles such as ammonia and hydrogen sulfide; a group of sensing materials that are insensitive to organic volatiles (e.g., acetoin and l-octen-3-ol), but which are sensitive to inorganic volatiles (e.g., ammonia and hydrogen sulfide); and a group of sensing materials that that are insensitive to organic and inorganic volatiles, but sensitive to other gases.4ACTIVE 714926490v1

[0025] The present disclosure also provides various methods, including but not limited to a method for detecting Avian Influenza (or other disease) by detecting chemicals emitted by infected animals; and a method for detecting Avian Influenza (or other disease) by detecting gases emitted by the feces of infected animals.

[0026] In addition, the present disclosure provides various systems, including but not limited to: an e-nose system that implements these methods described herein for detecting Avian Influenza; an e-nose system using the sensing materials described herein to detect Avian Influenza; an e-nose system optionally with individually controllable light and temperature modulation of the sensing materials; machine learning and artificial intelligence methods that process sensor signals from the e-nose system to identify Avian Influenza; an e- nose system with a modular architecture, enabling it to be used in various configurations including a wall-mounted system, a hand-held system, a table-top system, etc.; and a modular e-nose system that may be used with multiple sample sources including vials, single-point air monitoring, multi-point air monitoring, etc.

[0027] Some diseases may be diagnosed using odor changes. Alterations of body odor (expressed in feces or urine) may occur as a result of viral infection, inflammation, immunization, brain injury, and genetic mutation. Of note for detection of avian influenza, infection with a low pathogenic avian influenza (H5N2) may cause an increase of acetoin (3- hydroxybutanone) in feces of infected ducks as compared to fecal concentrations of 1-octen- 3-ol.

[0028] To mimic the hundreds to thousands of unique odorant receptors found in biological olfactory systems, a high-density sensor array architecture may include combinations of nano- and defect- engineered sensing materials. To provide diverse sensing response with part per billion detection limit and fast response time, two different types of gas sensing materials (i.e., phyllosilicates, and noble metal doped / decorated metal oxide semiconducting (MOS)) may be used, having temperature tunable electrical properties (e.g., charge carrier concentrations and mobility) and enhanced chemical reactivity.

[0029] Phyllosilicates, or sheet silicates, are a class of silicates comprising tetrahedral (T) and octahedral (O) sheets. In 1:1 phyllosilicates, layers comprising a cornersharing pseudohexagonal network of Si tetrahedra alternate with layers comprising edgesharing metal octahedra. These layers may be delaminated and leveraged to behave as pseudo-two-dimensional layers, with a semiconducting metal oxide layer connected to an 5ACTIVE 714926490v1insulating silicon oxide layer. These materials may accommodate the entire periodic table on either the octahedral or tetrahedral site or interstitial sites.

[0030] Metal oxide semiconductor (MOS)-based nanofibers with varying compositions and structures may be used as sensing materials. Such sensing materials may be fabricated as metal doped / decorated MOS nanofibers by combining electrospinning followed by post heat treatment where the diameter of nanofiber was confined by tuning electrospinning parameters and crystallinity were altered by controlling post annealing conditions. Introducing metal dopants may alter chemical interaction between MOS and target analytes.

[0031] Zinc nitrate hexahydrate (Zn (NOsh ’ 6H2O) and potassium tetrachloroplatinate (KzPtC ) as the Pt precursor may be used to synthesize pristine ZnO nanosheets (NS). For example, zinc nitrate hexahydrate can be dissolved in a mixture of ethanol and water (1:2) under continuous stirring until homogeneity. The solution can be transferred to a Teflon-lined autoclave and heated at 180 °C for 24 hours, where it can later be washed and dried at 80 °C for 6 hours, and can then be annealed at 200 °C and 550 °C for 4 hours under argon gas.

[0032] The Pt decorated ZnO NS may be synthesized with the addition of 0.01 at% h PtCk, instead of gold precursor, in 40mL of deionized water and treated with the microwave-assisted method of 1000 W for 1 minute or less. The solution can then be washed and dried at 80 °C for 6 hours.

[0033] Preparing phyllosilicate nanosheets may include nickel (II) nitrate hexahydrate (Ni (NOsh ’ 6H2O), aluminum chloride hexahydrate (AICI3 ’ 6H2O), silicic acid hydrate (l-hSiCh), sodium hydroxide (NaOH), zinc nitrate hexahydrate (Zn (NOsh ’ 6H2O), and gallium (III) nitrate hexahydrate (Ga (NOsh ’ XH2O). The weight of each component is based on the chemical formula for each material. First the silicic acid hydrate can be dissolved in 9mL of deionized (DI) water, then the nickel (II) nitrate hexahydrate and the other precursor (aluminum, zinc, or gallium) are added and stirred for 30 minutes until dissolved. Next, three pellets of sodium hydroxide can be dissolved in 6mL of DI water and added to the mixture solution dropwise with a syringe pump (0.15 mL / min) while stirring continuously. After all the sodium hydroxide has been added the solution is gelled for three days and transferred to a Teflon-lined autoclaved and heated at 200 °C for 50 hours. Finally, the solution can be washed with DI water and absolute ethanol and dried at 80 °C for 24 hours.6ACTIVE 714926490v1

[0034] Preparing SnCh nanofibers (NFs) may include polyvinylpyrrolidone (PVP, molecular weight of 1,300,000) and Tin (II) chloride dihydrate (SnCIz’ 2H2O), N, N- dimethylformamide (DMF). The SnCh NFs can be synthesized by electrospinning, where the precursor solution is composed of 7wt.% of PVP dissolved into DMF, and 7wt. % tin (II) chloride dihydrate dissolved into ethanol and stirred until homogeneity is achieved and mixed afterwards. The NFs may follow electrospinning parameters such as 15 kV, 0.4 mL / h, 10 cm, 40°C, and 12% RH. Then they can be annealed at 550 °C for one or more hours under air.

[0035] Preparing WO3 NFs may include ammonium metatungstate hydrate (AMH, H26N6O40W12, 99.99 % trace metal basis), PVP (same as above), and gold chloride trihydrate (HAuC ’ 3H2O). The WO3 NFs may be synthesized by electrospinning where 6wt.% PVP and 12wt.% of AMH can be dissolved into DMF and stirred until homogeneity is achieved. The NFs may follow electrospinning parameters such as 13 kV, 0.25 mL / h, 10 cm, 40°C, and 12% RH, then annealed at 700 °C for 24 hours at a rate of 10 °C / hour under air.

[0036] FIGS. 1A-1C illustrate a system 100 corresponding to an example e-nose system as provided herein. FIG. 1A illustrates a diagram of system 100, FIG. IB illustrates an example device arrangement of an example implementation of system 100, and FIG. 1C illustrates an example internal component layout of an example implementation of system 100.

[0037] System 100 may include a sensor array 160, a pump 150, a sample chamber 156, electronics 102, a valve 152, and a valve 154. As also illustrated in FIG. 1A, system 100 may include tubing as needed to form an airway connecting sensor array 160 to sample chamber 156 and / or the ambient environment (e.g., for air intake and / or exhaust of ambient air from the environment of system 100), as well as electrical connections (e.g., coupling electronics 102 to pump 150, sensor array 160, valve 152, and valve 154).

[0038] Pump 150 may correspond to any pump device, such as a diaphragm pump, configured to move fluid (e.g., liquid or gas such as an air sample for analyte detection) via mechanical action. Valve 152 and valve 154 may each correspond to any valve device configured to direct, regulate, or otherwise control mechanical flow of fluids (as may be pumped by pump 150) via opening or partially / fully obstructing passageways. In some examples, valve 152 and / or valve 154 may correspond to 3-way valves having three ports (e.g., inlet or outlet) that may be connected in any desired combination. Pump 150, valve 152,7ACTIVE 714926490v1and valve 154 may be electronically controlled by electronics 102 (e.g., having electronically controllable motors).

[0039] Electronics 102 may include a microcontroller unit (MCU) 130, a storage device 140, a display 132, a readout 136, and an h-bridge 134. H-bridge 134 may correspond to an h-bridge circuit for switching polarity of a voltage applied to a load (e.g., for reversing motor operations such as pump 150, valve 152, and valve 154) although H-bridge 134 may generally represent any support circuits. Storage device 140 may represent any type of memory / storage device. Display 132 may represent any type of device capable of optical and / or auditory output. Readout 136 may correspond to any circuit for interfacing with sensors / detectors (e.g., a readout board). MCU 130 may represent any type of processor, and as will be described further below, may control one or more of pump 150, valve 152, and / or valve 154 via h-bridge 134, read sensor data from sensor array 160 via readout 136, perform analysis based on data / software via storage 140, store results in storage 140, and provide result output via display 132.

[0040] Sensor array 160 may include one or more sensor device 162 which may comprise a sensing material as described herein. In some examples, sensor array 160 may include multiple sensor devices 162 arranged in a grid (e.g., 4x4 for 16 different sensing materials). Readout 136 may be configured with a channel for each sensor device 162.

[0041] FIG. 1C further illustrates system 100 including additional components such as on / off buttons 166, and other sensors 164. A battery 168 (e.g., an internal battery) may correspond to a power supply for system 100, although in other implementations system 100 may use other power supplies as needed. A vial holder 158 may be configured to hold a container such as sample chamber 156 in airtight connection with the airway of system 100.

[0042] The AIV detection system provided herein may support a variety of configurations suitable for diverse monitoring applications. These include a wall-mounted configuration for continuous environmental surveillance in settings such as poultry barns, as well as a portable, handheld configuration designed for point-of-care and field-based testing. FIGS. 1B-1C illustrate a handheld system using a diaphragm pump and pair of three-way valves to enable sampling either from ambient air or from the head space of a vial containing a fecal sample. For ambient air sampling, valves 152 and 154 are configured so that pump 150 draws in ambient air, passes it over sensor array 160, and then exhausts it. For head space sampling, valves 152 and 154 are configured so that pump 150 recirculates air from an attached vial 8ACTIVE 714926490v1(e.g., sample chamber 156) over sensor array 160. Buttons 166 allow the user to operate the system and an OLED display (e.g., display 132) to indicate device status and detection results.

[0043] The MCU (e.g., MCU 130) from the sensor readout board (e.g., readout 136) can be used for all user interface functions and operates valves 152 and 154 and pump 150 via one or more H-bridge circuits (e.g., H-bridge 134). The AIV detection system is designed with a modular architecture that allows internal components to be placed into appropriate cases to support a range of monitoring applications, such as a wall-mountable case for continuous surveillance, and a portable handheld case capable of single- and multipoint sampling, as well as headspace analysis of vials containing fecal samples as described herein.

[0044] The sensor system comprises sensor array 160, readout 136, and MCU 130. The MCU contains firmware for operating the system. Sensor array 160 comprises 16 chemiresistive sensors fabricated on micro-electromechanical systems (MEMS)-based micro hotplates where different nano- and defect- engineered sensing materials were drop-casted on top of the micro hotplate (although in other implementations other arrangements / combinations of sensing materials can be used). The readout board uses digital-to-analog converters (DACs) and op amps to control the temperatures of the micro hot plates. Analog-to-digital converters (ADCs) and appropriate voltage divider circuits are used to measure the resistance of the sensing elements. For example, each sensing element is connected in series with a 100-kQ reference resistor such that the combination is placed between 0V and 5V voltage references.

[0045] The ADCs measure the voltage across the sensing elements, enabling their resistances to be computed using an appropriate voltage divider formula. The sensor board is connected to the readout board via a flexible flat cable (FFC) enabling the use of interchangeable sensor boards. The readout board includes an IC having 16 12-bit DACs, 16 12-bit ADCs, and eight op amps, and two ICs each containing four op amps, for a total of 16 op amps. The op amps buffer the DAC outputs to provide sufficient current to drive the micro hot plates. This three-chip design provides high-performance in a low power, compact form.

[0046] Turning to FIG. 2, FIG. 2 is a block diagram of an example computing device 202 that may correspond to electronics 102. As illustrated in this figure, example computing device 202 may include one or more modules 204 for performing one or more tasks. As will be explained in greater detail below, modules 204 may include a pump module 206, a sensor 9ACTIVE 714926490v1module 208, an machine learning (ML) module 210, and an output module 212. Although illustrated as separate elements, one or more of modules 204 in FIG. 2 may represent portions of a single module or application.

[0047] In certain embodiments, one or more of modules 204 in FIG. 2 may represent one or more software applications or programs that, when executed by a computing device, may cause the computing device to perform one or more tasks. For example, and as will be described in greater detail below, one or more of modules 204 may represent modules stored and configured to run on one or more computing devices, such as the devices illustrated in FIGS. 1A-1C. One or more of modules 204 in FIG. 2 may also represent all or portions of one or more special-purpose computers configured to perform one or more tasks. In some examples, pump module 206 may correspond to instructions / hardware for operating / reconfiguring valves and / or pumps (e.g., pump 150, valve 152, and / or valve 154 via H-bridge 134). Sensor module 208 may correspond to instructions / hardware for interfacing with sensors (e.g., sensor array 160 and / or sensor device 162 via readout 136). ML module 210 may correspond to instructions / hardware for training, using, and / or using the results of an ML model. Output module 212 may correspond to instructions / hardware for interfacing with an output device (e.g., display 132) for providing output.

[0048] As illustrated in FIG. 2, computing device 202 may also include one or more memory devices, such as memory 240 (corresponding to storage 140). Memory 240 generally represents any type or form of volatile or non-volatile storage device or medium capable of storing data and / or computer-readable instructions. In one example, memory 240 may store, load, and / or maintain one or more of modules 204. Examples of memory 240 include, without limitation, Random Access Memory (RAM), Read Only Memory (ROM), flash memory, Hard Disk Drives (HDDs), Solid-State Drives (SSDs), optical disk drives, caches, variations or combinations of one or more of the same, and / or any other suitable storage memory.

[0049] As illustrated in FIG. 2, computing device 202 may also include one or more physical processors, such as physical processor 230 (corresponding to MCU 130). Physical processor 230 generally represents any type or form of hardware-implemented processing unit capable of interpreting and / or executing computer-readable instructions. In one example, physical processor 230 may access and / or modify one or more of modules 204 stored in memory 240. Additionally or alternatively, physical processor 230 may execute one ioACTIVE 714926490v1or more of modules 204 to facilitate biomarker detection. Examples of physical processor 230 include, without limitation, microprocessors, microcontrollers, Central Processing Units (CPUs), Field-Programmable Gate Arrays (FPGAs) that implement softcore processors, Application-Specific Integrated Circuits (ASICs), portions of one or more of the same, variations or combinations of one or more of the same, and / or any other suitable physical processor.

[0050] As illustrated in FIG. 2, computing device 202 may also include one or more data elements 220, such as sensor data 222, detection data 224, and configuration data 226. Sensor data 222 may generally represent any type or form of data relating to sensor signals, including signals received from sensor devices (e.g., sensor array 160 and / or sensor device 162). Detection data 224 may generally represent any type or form of data relating to analyte analysis (e.g., relevant thresholds, historical data, etc.). In some examples, sensor data 222 and / or detection data 224 may also represent quantities computed from the received signals (e.g., for computing features from the signals). Configuration data 226 may generally represent any type or form of data relating to device operation (e.g., parameters, options, thresholds, etc.).

[0051] In addition, although not illustrated in FIG. 2, in some implementations one or more of modules 204 and / or data elements 220 may be remote (e.g., residing on a different computing device communicatively coupled to computing device 202 and accessible to computing device 202). For example, ML module 210 may be at least partially implemented on a remote device such that certain aspects of machine learning (e.g., training) may be performed on a different computing device, the results of which may then be incorporated into ML module 210 locally on computing device 202.

[0052] Machine learning may be used for biomarker detection, by allowing an e- nose system to process sensor signals based on one or more trained ML models (e.g., by exposing an e-nose to various analytes, and analyzing the resulting sensor signals). Nonlimiting examples of training models for classifying and quantifying analytes will be further discussed. For example, classifiers can identify the species of the analyte and regression models can quantify the concentration. Training the classifiers may include characterizing sensor responses by a variety of features including time derivatives of the sensor response during both exposure and recovery, the peak response during exposure, and the time (T90) to reach 90% of the peak exposure and recovery responses. These features may be computed 11ACTIVE 714926490v1for each of multiple (e.g., three) sensor operating temperatures. The classification performance of models trained with features from a single operating temperature may be compared to that of models trained with features from all three temperatures. A feature subset selection method may identify the most important features for classification. Likewise, the classification accuracy (the percentage of examples correctly classified) of multiple (e.g., four) different classification algorithms may be compared. The analyte concentration may be quantified with multiple (e.g., two) types of regression models.

[0053] In one example training, each training event was characterized by nine features: ES60, ES120, ES300, RS60, RS120, RS300, PR, ET90, and RT90. Prior to computing these features, the resistance, R(t), of a sensor is normalized by the baseline resistance, Ro. This baseline may be computed, for example, as the average of R(t) during the last 60 minutes of a warmup period prior to exposure of the e-nose to analytes. ES60, ES120, ES300 are the average time derivatives (slopes) of R(t) / Ro during the first 60-sec, 120-sec, and 300-sec, respectively, of the exposure of the e-nose to the analytes. Likewise, RS60, RS120, RS300, are the average time derivatives of R(t) / Ro during the first 60-sec, 120-sec, and 300-sec, respectively, of the recovery of the e-nose after the exposure to the analytes has been completed. The average time derivatives may be computed as the slope of a least squares line fit of the relevant data, which may mitigate the effects of sensor noise. For example, ES60 ("exposure slope 60 seconds") may be the slope of a least square line fit of the first 60 seconds of the exposure.

[0054] PR is the peak response during exposure. This may be computed as the average of R(t) / Ro during the last 60 seconds of the exposure. Averaging may mitigate the effects of noise. ET90 is the time required for a sensor to reach 90% of this peak response. To compute ET90, the signal R(t) / Ro may be smoothed to remove noise using a moving average in which each sample point is averaged with the one or more neighbors (e.g., five neighbors) on either side. T90 may be computed as the time, t, of the first sample point for which the condition | R(t) / Ro - SR | > 0.91 PR - SR | , where SR is the average response during a time interval (e.g., 60 seconds) just prior to the exposure. To provide further robustness to noise, it may be required that this condition be satisfied for several (e.g., at least five) consecutive data points.

[0055] If this is not satisfied, RT90 may be taken to be the full duration of the exposure. Analogously, RT90 is the time required for the sensor to complete 90% of the 12ACTIVE 714926490v1recovery. More specifically, it is the time it takes for the response to transition 90% of the way from PR to the average response during the last time interval (e.g., 60 seconds) of the recovery. While the six slope features characterize the initial speed of the response during exposure and recovery, the two T90 features characterize the overall speed of the response.

[0056] The sensing performance of nano sensing materials often depends on their temperature. To examine this dependence, sensing data may be collected at multiple different sensor operating temperatures, for example: 150° C, 300° C, and 450° C. For a given temperature, each of several sensor systems (e.g., as five) may be exposed to multiple concentrations (e.g., seven) of each of multiple analytes (e.g., nine), although any appropriate quantities / combinations of operating temperatures, sensor systems, concentrations, and / or analytes may be used. In this example, the corresponding experiments produced a total of 315 positive examples. The experiments also produced 45 negative examples of the sensor response to dry air in the absence of an analyte. These negative examples were taken from the 3 hour warmup period.

[0057] The last 30 minutes of the warmup may be a 15 minute exposure (with no analyte) followed by a 15 minute recovery. In total, the data comprised 360 examples for each of three operating temperatures.

[0058] To characterize the temperature dependence of the sensing performance, two kinds of classification models may be considered: single temperature models based on data from a single operating temperature and multi-temperature models based on data from all three temperatures (e.g., 150 °C, 300 °C, and 450 °C). Each single temperature model included 360 examples, each characterized by 144 features (nine features for each of 16 sensors). For multi-temperature models, the single-temperature examples were concatenated to produce 360 examples with 432 features. In this way, each multitemperature example included nine features for each of 16 sensors operating at each of three temperatures.

[0059] Various algorithms - k-nearest neighbor, decision tree, random forest, and neural network - may be used for classification to compare their performance. Prior to training neural networks, the features may also be z-normalized. For a rigorous evaluation of model performance, in one example, various models were trained and tested with 10-fold cross validation.13ACTIVE 714926490v1

[0060] To avoid over-fitting in a model with unnecessary features, forward sequential feature selection was used to identify the most important features. This process uses greedy search to iteratively identify the most important features. Consider, for example, selecting features for a single-temperature model. The process begins with an empty list of selected features. Then 144 single-feature models are trained, one for each of the 144 features. The feature from the model with the highest accuracy is added to the set of selected features. Then 143 two-feature models are trained, one for each of the remaining 143 features. The model with the highest accuracy determines the second most important feature. The process continues in this way until the desired number of features has been selected. Feature selection was conducted using a random forest classifier with 10-fold cross validation.

[0061] Once a classifier has identified the species of the analyte, a regression model is used to quantify the concentration as a function of the peak response during exposure. One or more types of regression models may be considered, such as least squares logarithmic models and multi-layer perceptron models. The former are a least squares fit to the logarithmic model log C = aPR + b, where C is the concentration, PR is the peak response, and a and b are the regression coefficients. This model uses the response of a single sensor to quantify the concentration of a particular analyte. The particular sensor and operating temperature that provides the most accurate quantification may vary with the analyte. Least squares models were created by aggregating the data from all five sensor systems, resulting in nine models (one for each analyte). Another set of models was created by disaggregating the data from the individual sensor boards, resulting in 45 models (nine models for each of the five systems). Model accuracy was characterized by the coefficient of determination (R2).

[0062] Multi-layer perceptron regression models were also used to quantify analyte concentration. These models included 48 features: the peak response (PR) values of the 16 sensors at each of the three operating temperatures. In addition to aggregating temperatures, these models also aggregate data from all five sensor systems.

[0063] The ML training described above may be applied to any appropriately selected sensing materials for the sensor systems. A non-limiting screening process for selecting appropriate sensing materials to identify and quantify desired target materials (e.g., AIV biomarkers) will now be described. In an example, an initial screening of 206 sensors was conducted on the two biomarkers (l-octen-3-ol and acetoin) and two barn / fecal VOCs 14ACTIVE 714926490v1(ammonia and hydrogen sulfide) at different operating temperatures (250 °C, 300 °C, 350 °C, 400 °C, 450 °C). Each sensor consists of a unique material that can be categorized into three synthesis methods: electrospinning (ES), hydrothermal (HT), or commercially available materials. 120 NFs were synthesized through ES (WO3, SnCh, ZnO, and TiCh); 16 phyllosilicate plates and 35 ZnO sheets through HT synthesis; and 37 commercial materials (WO3, SnO2, ZnO, TiOz, Fe2O3, CuO). Analyzing results allow selecting based on criteria including: 1) Selective / highly sensitive to acetoin; 2) Selective / highly sensitive to l-octen-3-ol; 3) Highly sensitive to acetoin and l-octen-3-ol; 4) Selective to ammonia; 5) Selective to hydrogen sulfide; 6) No response towards acetoin and l-octen-3-ol; and 7) No response to all analytes. FIG. 3 illustrates a table 300 summarizing selected sensors (e.g., nanomaterials and synthesis methods as discussed herein), including MOS (metal oxide semiconductor) materials, NP (nanoparticles), NF (nanofibers), NS (nanosheets), C (commercially available nanomaterials), HT (hydrothermal synthesis), and ES (electrospinning).

[0064] The selected sensors show the sensing variability required to detect biomarkers in a complex mixture. The materials were composed of different dimensionalities and compositions (e.g., MOS and phyllosilicates). The lowerthe dimensionality, the higherthe surface area to volume ratio; however, they tend to agglomerate, reducing the available surface area for gas analytes to interact. The catalytic activity and electronic properties of chemiresistive materials significantly affect the sensing response towards gas analytes. The types of defect engineering executed on the ES and HT nanomaterials were doping and vacancies. Metal doping and oxygen vacancies enhance catalytic activity by increasing the active sites. These active sites promote oxygen dissociation and incite a more effective reaction and electron transfer, increasing the sensing response. Oxygen vacancies may be designed either through reducing atmospheres (argon) with the ZnO nanosheets or by decreasing the diameter and grain size in the NFs. Electrospinning and environmental conditions may significantly affect the NF diameter, while post-treatment may affect the grain size. A lower PVP content, a higher applied voltage, a slower feed rate, and a higher temperature and lower absolute humidity environment may decrease NFs' diameters. In addition, increasing the dopant concentration increases the electrical conductivity of the NF solution, which in turn may increase the jet elongation, resulting in a smaller NF diameter. A smaller diameter may have a larger surface area-to-volume ratio, favoring the adsorption and desorption of target analytes due to its wider depletion region compared to its diameter.15ACTIVE 714926490v1Meanwhile, a lower annealing temperature and time may decrease the grain size, creating more grain boundaries where a depletion layer can form from oxygen adsorption and generate oxygen ionic species. Oxygen vacancies in these small grain boundaries may play a crucial role in the structure's stability, since the surface structure can be destroyed by reducing the oxygen vacancy concentration through oxygen adsorption

[0065] The electronic properties of the materials may be defined through the band gap and the temperature coefficient of resistance (TCR). The UV-VIS diffuse reflection spectra can be used to calculate the band gap based on the following equation:Equation 1

[0066] where a is the absorption coefficient, h is Planck's constant, hv is the photon energy, A is a constant, n is direct (2) or indirect (0.5), and Egis the optical band gap. An increase in band gap was seen in higher annealing temperature and a longer duration due to an increase in grain size, which increased the gap between the conduction and valence band. While there was a decrease in grain size, reducing the band gap, the major contributor for these materials is the oxygen vacancies due to the vacancies forming defective states below the conduction band.

[0067] The TCR experiment was performed at operating temperatures ranging from room temperature to 450 °C at varying oxygen concentrations. Overall, resistance increases with the increase of oxygen concentration compared to a nitrogen atmosphere, due to the depletion layer formation from oxygen adsorption. This energy barrier (Eb) can be calculated through a linear fitting of the calibration curve of the logarithm of resistance versus the inverse of temperature. Equation 2

[0068] Where R is the resistance of the sensing materials, Ro is a pre-exponential factor, Eb is the energy barrier, KB is the Boltzmann constant, and T is the temperature in Kelvin. Since many of the sensors have various linear segments, the Eb for each operating temperature was calculated based on its linear segment.

[0069] With the above selected sensing materials and applying the ML training described above, the corresponding experiments have shown that classification accuracy may be greater with higher sensor operating temperatures. However, using sensing data from all three operating temperatures gives the highest accuracy. For all four types of classification16ACTIVE 714926490v1models described above, the top 10 features contribute much of the classification accuracy while the peak accuracy is achieved with between about 23 and 30 features. The top 23 features for the multi-temperature models include features from 12 of the 16 sensors and all six feature types (exposure slopes, peak response, recovery slopes, and T90s). Furthermore, the top 23 include three features at an operating temperature of 150 °C, 10 at 300 °C, and 10 at 450 °C. It is apparent that the diversity of sensing materials, operating temperatures, and feature types contributes to the accuracy of the models at distinguishing between the nine analytes.

[0070] Based on the experiments described herein, for a given set of features, random forest models nearly always provided the best accuracy for classifying analytes, although other models may be used for other sensing materials. While the k-nearest neighbor and decision tree models had lower accuracy than the random forest models, they also required considerably less computation for training and testing. Neural networks models had lower accuracy than random forest models and required more computation.

[0071] To better characterize the accuracy of random forest models using the top 23 features, 10 models were trained and evaluated for each of the four temperature conditions. The average accuracies for 150 °C, 300 °C, 450 °C, and all temperatures combined were 85.6%, 91.2%, 95.1%, and 96.1%, respectively. The multi-temperature models were, on average, one percentage point more accurate than models trained with only 450 °C features (which is a 19.4% reduction in misclassifications).

[0072] The experiments have shown that for a random forest model trained with the top 23 multitemperature features, all 45 examples without an analyte were correctly classified, 34 examples of l-octen-3-ol were correctly classified, while one example was misclassified as formaldehyde. For this model, only 14 examples were misclassified, for an overall accuracy of 91.1%.

[0073] The experiments have shown that for regression models for quantifying the concentrations of analytes, the R2values for least squares models that aggregate the data from all five sensor systems ranged from 85.6% to 96.5%, with an average of 91.4%. To handle possible variability between sensors, another set of least squares models was created by disaggregating the data from the individual sensor systems, resulting in five models for each analyte. R2for these models ranged from 93.0% to 99.0%, with an average of 96.9%. The higher performance of the disaggregated models indicates that there is some sensor-to- 17ACTIVE 714926490v1sensor variation. Finally, the R2values for models that aggregate data from all five sensor systems and employ features from all three operating temperatures, ranged from 98.5% to 100.0%, with an average of 99.8%.

[0074] FIG. 4 illustrates an example high-throughput testing system 401 to rapidly test and train an AIV detector device 400 (e.g., system 100), for detecting the avian influenza (AIV) volatile biomarkers, for instance using the material screening and ML training described above. The AIV detector device 400 may include, for example, a 16-sensor array, although in other examples may include a fewer or greater number of sensors in the array. The testing system 401 may test multiple AIV detector devices with modulation of temperature and light illumination.

[0075] The testing system 401 may simultaneously test, for example, 20 e-nose systems, each comprising an array of 16 sensors. The testing system 401 may include a vapor generation system that may produce known concentrations of target analytes. The testing system 401 may also control the temperature and optical excitation of the sensors of the e- nose systems to investigate effects of this excitation on sensing performance. The testing system 401 may measure the sensor responses by measuring the electrical properties of the sensors through various techniques, including real-time measurement of the electrical resistance of the sensing materials. The sensing performance of the sensors at detecting AIV may be characterized by exposing the sensors to biomarkers and background fecal analytes (e.g., acetoin, l-octen-3-ol, hydrogen sulfide, and / or ammonia) at various concentrations and / or combinations.

[0076] Successful detection of AIV volatile biomarkers may require detecting organic biomarkers (e.g., acetoin and / or l-octen-3-ol) and distinguishing these organic biomarkers from background inorganic volatiles (e.g., ammonia and / or hydrogen sulfide) contained in feces. The present disclosure provides nano- and defect-engineered sensing materials to detect AIV volatile biomarkers from background fecal analytes. These sensing material include, but are not limited to: sensing materials that are highly sensitive and selective toward ketones such as acetoin; sensing materials that are highly sensitive and selective toward alcohols such as l-octen-3-ol; sensing materials that are highly sensitive toward both ketones and alcohols; sensing materials that are sensitive and selective toward to ammonia; sensing materials that are sensitive and selective toward hydrogen sulfide; sensing materials that are insensitive to ketones and alcohols, but which are sensitive to 18ACTIVE 714926490v1hydrogen sulfide and ammonia; sensing materials that are insensitive to ketones, alcohols, hydrogen sulfide, and ammonia, but sensitive to other gases, etc.

[0077] In another material screening example, various testing conditions were examined using the high-throughput testing system 401. For example, 7,210 different testing condition combinations were examined, based on 206 different sensing materials, five sensor operating temperatures (in the range 250°C to 450°C), and seven concentrations of the four potential fecal analytes (acetoin, l-octen-3-ol, hydrogen sulfide, and ammonia). Each of the 206 sensing materials was implemented into a sensor, resulting in 206 unique sensors. These sensors were assembled into 16-sensor arrays (e.g., the AIV detector device 400) and tested with the high-throughput testing system 401 in FIG. 4.

[0078] To further improve the performance of the e-nose system at detecting AIV infection, the e-nose system may optionally include hardware and software to individually and dynamically control the light and temperature excitation of each sensor in an e-nose system 500 (corresponding to system 100) in FIG. 5. FIG. 5 illustrates a sixteen sensor e-nose platform with individually controllable temperature and light modulation. The modulation may produce high-dimensional sensing data, enabling AIV biomarkers to be accurately identified from complex fecal samples. In other implementations, the e-nose system 500 may include a fewer or greater number of sensors with individually controllable temperature / light.

[0079] Dynamically adjusting the excitation during sensing may produce additional data to characterize analytes, thus enabling more accurate detection of the target analytes. The individual sensors in the e-nose system 500 may be fabricated by drop-casting sensing material on MEMS-based micro hotplates. The hotplates may enable the temperatures of the sensing materials to be controlled. Additionally, an array of multiwavelength LEDs may be used for optical excitation. In some implementations, each sensing material (e.g., each sensor) may be coupled to its own heat element (e.g., hotplate) and light element (e.g., LED), although in other implementations the heat elements and / or light elements may be shared by sensing materials of different sensors.

[0080] The e-nose system 500 may include a machine learning model or other machine learning / artificial intelligence algorithm as described herein that may process the response of the 16 sensors in the e-nose system 500 to identify the target analytes, and thus AIV infection, with high-accuracy.19ACTIVE 714926490v1

[0081] As described herein, a modular architecture enables use with multiple sample sources including vials, single-point air monitoring, multi-point air monitoring, etc., as also illustrated in FIG. 6. FIG. 6 illustrates a flow chart 600 of an example analyte flow path through an AIV detector device as described herein (e.g., system 100). FIG. 6 illustrates an example configuration of the e-nose system for use with sample vials. Three-way valves enable continuous sampling from a vial or recirculation (using a pump) of a sample.

[0082] FIG. 6 illustrates a sample 656 (corresponding to a gas sample held in sample chamber 156 and / or from ambient air), a valve 654 (corresponding to valve 152 and / or valve 154), a sensor 660 (corresponding to sensor array 160 and / or sensor device 162), a pump 650 (corresponding to pump 150), a valve 652 (corresponding to valve 152 and / or valve 154), electronics 602 (corresponding to electronics 102 and / or computing device 202), and air 606 representing ambient air (e.g., from the device's environment).

[0083] In one example, embedded electronics (e.g., electronics 602) may control a sampling process. Sample 656 from a sample chamber (e.g., corresponding to a vial, one or more points in air, etc.) may circulate through a three-way valve (e.g., valve 654 as controlled by electronics 602 such as via H-bridge 134) to sensor 660 (e.g., including one or more of the sensing materials described herein). As sample 656 is exposed to sensor 660, electronics 602 may control temperature and / or light of sensor 660, which in some examples may also be based on input from one or more of the machine learning models described herein. Electronics 602 may further measure electrical resistance of the sensing materials in sensor 660 over time (e.g., using readout 136). Electronics 602 may further control pump 650 to further pump sample 656 through a three-way valve (e.g., the same and / or a different three- way valve such as valve 652) to an exhaust or recirculate back to sensor 660 as desired. In addition, one or more of the machine learning models described herein may analyze the read sensor data and determine, for example, the presence of biomarkers (e.g., acetoin and / or 1- octen-3-ol) as distinguished from background analytes.

[0084] Additionally, clean air (air 606) may be passed through the system (e.g., through the three-way valve) to purge the e-nose. In other implementations, other types of valves and pumps may be used.

[0085] This flexibility in the device form factor allows an end user to use the same sensor system in a variety of monitoring modalities (e.g., portable and / or stationary) without incurring extra costs.20ACTIVE 714926490v1

[0086] FIG. 7 is a flow diagram of an example method 700 for detecting volatile biomarkers using an e-nose system as described herein. The steps shown in FIG. 7 may be performed by any suitable computer-executable code and / or computing system, including system 100 in FIGS. 1A-1C, computing device 202 in FIG. 2, and / or variations or combinations of one or more of the same. In one example, each of the steps shown in FIG. 7 may represent an algorithm whose structure includes and / or is represented by multiple sub-steps, examples of which will be provided in greater detail below. A target material, as described herein, may refer to an analyte, such as a biomarker or volatile compound, and / or may represent any other material that is targeted for detection.

[0087] As illustrated in FIG. 7, at step 702 one or more of the systems described herein may pump an air sample over a sensor array that includes a first sensor device comprising a first sensing material configured to react to a target volatile compound in ambient air and a second sensor device comprising a second sensing material configured to react to a background gas in ambient air (e.g., one or more background gases that may be distinct from the target volatile compound). For example, pump module 206 may, as part of computing device 202 in FIG. 2, control a pump device such as pump 150.

[0088] The systems described herein may perform step 702 in a variety of ways. In one example, the pump device further comprises or is coupled to a valve device (e.g., valve 152 and / or valve 154) and tubing between at least the pump device and the sensor devices (e.g., sensor array 160).

[0089] In some examples, the valve device is coupled to an intake from a device environment (e.g., in a wall-mounted mode) and an exhaust to the device environment, and the ambient air is from the device environment. Pump module 206 may be further configured to control the pump device to continuously pump the ambient air from the device environment and further configured to detect the target material from the device environment by detecting the target material from the ambient air pumped from the device environment.

[0090] In some examples, the valve device is coupled to a sample vessel (e.g., sample chamber 156) configured to hold a sample and wherein the ambient air is from the sample vessel (e.g., in a hand-held mode). Pump module 206 may be further configured to control the pump device to pump the ambient air from the sample vessel and further21ACTIVE 714926490v1configured to detect the target material from the sample by detecting the target material from the ambient air pumped from the sample vessel.

[0091] In some examples, the target material corresponds to a VOC, which may also correspond to a biomarker indicative of a virus. In other examples, the target material may correspond to any other volatile compound, which ML module 210 has been trained to detect and sensor array 160 having sensor devices 162 in an appropriate arrangement (e.g., with respect to number / placement as well as sensing material). In some examples, configuration data 226 may establish parameters for the different modes (e.g., operation for wall-mounted mode or continuous ambient air monitoring, operation for hand-held or point sample mode, selecting biomarkers and thresholds for reporting, etc.).

[0092] In some examples, the first sensing material may be configured to react upon interaction with the target material, which can be detected to produce an output signal (e.g., sensor data) that corresponds to a measurable change (e.g., change electrical resistance, change electrical capacitance, change electrical impedance, change electrochemical potential, change color, change mass, and / or change another measurable property) in a property of the first sensing material. In some examples, the first sensing material may be configured to react upon interaction with the background gas, which can be detected to produce an output signal (e.g., sensor data) that corresponds to a measurable change (e.g., change electrical resistance, change electrical capacitance, change electrical impedance, change electrochemical potential, change color, change mass, and / or change another measurable property) in a property of the second sensing material.

[0093] At step 704 one or more of the systems described herein may receive sensor data from the sensor array in response to pumping the air sample. For example, sensor module 204 may, as part of computing device 202 in FIG. 2, receive sensor data 222 from sensor array 160 (e.g., via readout 136). In some implementations, sensor data 222 may include the output signals from the sensor materials / sensor devices. In some implementations, sensor data 222 may include data based on processing the output signals (e.g., processing the output signals to determine the corresponding measured changes of properties of the sensing materials).

[0094] At step 706 one or more of the systems described herein may determine, based on the sensor data, a presence of the target volatile compound in the air sample. For example, ML module 210 may, as part of computing device 202 in FIG. 2, use sensor data 222 22ACTIVE 714926490v1and detection data 224 to detect biomarkers (e.g., concentrations thereof, and / or ratios with other biomarkers) as described herein.

[0095] The systems described herein may perform step 706 in a variety of ways. In one example, ML module 210 may be configured to identify the target material / analyte using a machine learning model that was trained to identify a presence of the target material / analyte in the ambient air based on sensor data such as electrical resistance data or other property change data (e.g., change in electrical resistance, electrical capacitance, electrical impedance, electrochemical potential, color, mass, etc.) from the first and second sensor devices. The electrical resistance data (or other property change data) from the first sensor device may correspond to the first sensing material reacting to a presence of the target material / analyte and the electrical resistance data (or other property change data) from the second sensor device may correspond to the second sensing material reacting to a presence of the ambient air. In some examples, the machine learning model may be configured to classify the presence of the target material / analyte in the ambient air based on the electrical resistance data (or other property change data) from the first and second sensor devices. In other examples, the sensor data may correspond to other measurable changes described above (e.g., chemiresistive, chemicapacitive, impedimetric, electrochemical potential, color, mass, etc.) and the ML model may be trained to classify the presence of the target material / analyte in the ambient air based on this sensor data. In addition, detecting / classifying the presence of the target material / analyte may include measuring an amount of the target material / analyte detected (e.g., a concentration relative to the ambient air, such as parts per million, parts per billion, parts per trillion, etc.). In other words, the presence of the target material / analyte may represent detecting the target material / analyte above a concentration threshold or other proportional metric.

[0096] In some examples, the ML model may be trained to directly receive the output signals from the sensor devices (e.g., raw sensor data), although in other examples, the ML model may be trained to receive processed sensor data (e.g., converting the raw sensor data or signals into values corresponding to the measured changes in property of the sensing materials, which may be processed by the sensor array, the ML model, and / or another processor).

[0097] At step 708 one or more of the systems described herein may provide a notification in response to detecting the presence of the target analyte (e.g., volatile 23ACTIVE 714926490v1compound). For example, output module 204 may, as part of computing device 202 in FIG. 2, output an appropriate notification using display 132.

[0098] The systems described herein may perform step 708 in a variety of ways. In one example, the notification may include a report of findings. In other examples, display 132 may be a simple display, such as a warning light or other visual indicator that may distinguish between positive (e.g., presence detected) and negative (e.g., not detected) results. In addition, in some examples, the notification may include or alternatively be an audible notification (e.g., a chime, warning tone, etc.). In further examples, system 100 may be communicatively coupled (e.g., via any appropriate wired and / or wireless medium including a computer network) to a remote computing device, such as a central server. The notification may include or alternatively be an electronic message to the remote computing device, such as a log of results, a warning message (e.g., critical notification), etc.

[0099] In some examples, the e-nose system may be configured to continuously monitor the device environment for a presence of the target material / analyte. The notification may be appropriate for continuous monitoring (e.g., an alert).

[0100] In some aspects, the techniques described herein relate to a device including: a plurality of sensor devices including: a first sensor device including a first sensing material configured to produce a first signal in response to a target material / analyte in ambient air; and a second sensor device including a second sensing material configured to produce a second signal in response to a background gas in the ambient air; and a computing device including: a memory; and a processor coupled to the memory and configured to detect the target material based on sensor data from the plurality of sensor devices.

[0101] In some aspects, the techniques described herein relate to a device, wherein the processor is configured to identify the target material using a machine learning model trained to identify a presence of the target material in the ambient air based on the sensor data from the plurality of sensor devices.

[0102] In some aspects, the techniques described herein relate to a device, wherein the first signal from the first sensor device corresponds to a measurable change in a property of the first sensing material reacting to a presence of the target material and the second signal from the second sensor device corresponds to a measurable change in a property of the second sensing material reacting to a presence of the background gas.24ACTIVE 714926490v1

[0103] In some aspects, the techniques described herein relate to a device, wherein the machine learning model is configured to classify the presence of the target material in the ambient air based on the measured changes of the first and second sensing materials.

[0104] In some aspects, the techniques described herein relate to a device, further comprising a pump device configured to pump the ambient air to the plurality of sensor devices and controlled by the processor, wherein the pump device further comprises a valve device and tubing between at least the pump device and the plurality of sensor devices.

[0105] In some aspects, the techniques described herein relate to a device, wherein the valve device is coupled to an intake from a device environment and an exhaust to the device environment, and wherein the ambient air is from the device environment.

[0106] In some aspects, the techniques described herein relate to a device, wherein the processor is further configured to control the pump device to continuously pump the ambient air from the device environment and further configured to detect the target material from the device environment by detecting the target material from the ambient air pumped from the device environment.

[0107] In some aspects, the techniques described herein relate to a device, wherein the processor is further configured to continuously monitor the device environment for a presence of the target material.

[0108] In some aspects, the techniques described herein relate to a device, wherein the valve device is coupled to a sample vessel configured to hold a sample and wherein the ambient air is from the sample vessel.

[0109] In some aspects, the techniques described herein relate to a device, wherein the processor is further configured to control the pump device to pump the ambient air from the sample vessel and further configured to detect the target material from the sample by detecting the target material from the ambient air pumped from the sample vessel.

[0110] In some aspects, the techniques described herein relate to a device, wherein the processor is further configured to provide a notification in response to detecting the target material.

[0111] In some aspects, the techniques described herein relate to a device, wherein the target material corresponds to a volatile organic compound (VOC).25ACTIVE 714926490v1

[0112] In some aspects, the techniques described herein relate to a device, wherein the target material corresponds to a biomarker indicative of a virus. In other examples, the biomarker may be indicative of any other kind of infection, such as bacterial and / or fungal infections.

[0113] In some aspects, the techniques described herein relate to a biomarker detector device including: a sensor array including: a first sensing material configured to react to a biomarker to produce sensor data; and a second sensing material configured to react to a background gas in the ambient air to produce sensor data; a pump device configured to pump the ambient air to the sensor array; and a computing device including: a storage; and a controller coupled to the storage and configured to control the pump device and detect the biomarker based on the sensor data from the sensor array.

[0114] In some aspects, the techniques described herein relate to a biomarker detector device, wherein: the controller is configured to identify the biomarker using a machine learning model trained to identify a presence of the biomarker in the ambient air based on electrical change data from the sensor array; the electrical change data from the sensor array corresponds to a measurable electrical change of the first sensing material reacting to a presence of the biomarker and a measurable electrical change of the second sensing material reacting to a presence of the background gas in the ambient air; and the machine learning model is configured to classify the presence of the biomarker in the ambient air based on the electrical change data from the first and second sensor devices.

[0115] In some aspects, the techniques described herein relate to a biomarker detector device, wherein the pump device further includes a valve device and tubing between at least the pump device and the sensor array.

[0116] In some aspects, the techniques described herein relate to a biomarker detector device, wherein: the valve device is coupled to an intake from a device environment and an exhaust to the device environment, and wherein the ambient air is from the device environment; the controller is further configured to control the pump device to continuously pump the ambient air from the device environment and further configured to detect the biomarker from the device environment by detecting the biomarker from the ambient air pumped from the device environment; and the controller is further configured to continuously monitor the device environment for a presence of the biomarker.26ACTIVE 714926490v1

[0117] In some aspects, the techniques described herein relate to a biomarker detector device, wherein: the valve device is coupled to a sample vessel configured to hold a sample and wherein the ambient air is from the sample vessel; and the controller is further configured to control the pump device to pump the ambient air from the sample vessel and further configured to detect the biomarker from the sample by detecting the biomarker from the ambient air pumped from the sample vessel.

[0118] In some aspects, the techniques described herein relate to a biomarker detector device, wherein the controller is further configured to provide a notification in response to detecting the biomarker.

[0119] In some aspects, the techniques described herein relate to a method including: pumping an air sample over a sensor array that includes a first sensor device including a first sensing material configured to change in response to a target volatile compound in the air sample and a second sensor device including a second sensing material configured to change in response to a background gas in the air sample; receiving sensor data from the sensor array in response to pumping the air sample; determining, based on the sensor data, a presence of the target volatile compound in the air sample; and providing a notification in response to detecting the presence of the target volatile compound.

[0120] The examples herein generally refer to a background gas, which may correspond to a gas (e.g., O2, N2) commonly in atmosphere (e.g., occupying a significant portion by volume), although may refer to any gas in ambient air which is not the target material / analyte or biomarker. In addition, although the examples herein generally describe measuring air (e.g., ambient air, air samples, etc.,) in other examples, the measurements may be made with respect to any fluid. For example, the air sample and / or ambient air may represent any fluid (e.g., liquid or gas) in which the target material is to be distinguished from and / or measured against the background fluids (e.g., any fluid in the ambient / environmental fluid that is not the target material).

[0121] Features from any of the implementations described herein may be used in combination with one another in accordance with the general principles described herein. These and other implementations, features, and advantages will be more fully understood upon reading the following detailed description in conjunction with the accompanying drawings and example claims / clauses.27ACTIVE 714926490v1

[0122] Clause 1. A device for detecting avian influenza virus (AIV) comprising: a plurality of sensor devices, each sensor device comprising: a sensing material configured for a biomarker profile; optionally a heat element; and optionally a light element; a memory; and a processor coupled to the memory and configured to control operating temperature and operating illumination of the plurality of sensor devices and identify AIV biomarkers based on sensing data from the plurality of sensor devices.

[0123] Clause 2. The device of clause 1, wherein the biomarker profiles correspond to at least one of: sensitivity to ketones; sensitivity to alcohols; sensitivity to ketones and alcohols; sensitivity to aldehydes; sensitivity to ammonia; sensitivity to hydrogen sulfide; sensitivity to hydrogen sulfide and ammonia and insensitivity to ketones and alcohols; insensitivity to ketones, alcohols, hydrogen sulfide, and ammonia and sensitivity to other gases; and sensitivity to any combination of ketones, alcohols, aldehydes, ammonia, hydrogen sulfide and insensitivity to any combination of ketones, alcohols, aldehydes, ammonia, hydrogen sulfide.

[0124] Clause 3. The device of clause 1 or 2, wherein the identifying the AIV biomarkers is further based on one or more machine learning models trained on operating temperatures, operating illumination, and sensing data.

[0125] As detailed above, the computing devices and systems described and / or illustrated herein broadly represent any type or form of computing device or system capable of executing computer-readable instructions, such as those contained within the modules described herein. In their most basic configuration, these computing device(s) each include at least one memory device and at least one physical processor.

[0126] In some examples, the term "memory device" generally refers to any type or form of volatile or non-volatile storage device or medium capable of storing data and / or computer-readable instructions. In one example, a memory device stores, loads, and / or maintains one or more of the modules and / or circuits described herein. Examples of memory devices include, without limitation, Random Access Memory (RAM), Read Only Memory (ROM), flash memory, Hard Disk Drives (HDDs), Solid-State Drives (SSDs), optical disk drives, caches, variations, or combinations of one or more of the same, or any other suitable storage memory.

[0127] In some examples, the term "physical processor" generally refers to any type or form of hardware-implemented processing unit capable of interpreting and / or 28ACTIVE 714926490v1executing computer-readable instructions. In one example, a physical processor accesses and / or modifies one or more modules stored in the above-described memory device. Examples of physical processors include, without limitation, microprocessors, microcontrollers, Central Processing Units (CPUs), Field-Programmable Gate Arrays (FPGAs) that implement softcore processors, Application-Specific Integrated Circuits (ASICs), systems on a chip (SoCs), digital signal processors (DSPs), Neural Network Engines (NNEs), accelerators, graphics processing units (GPUs), portions of one or more of the same, variations or combinations of one or more of the same, or any other suitable physical processor.

[0128] Although illustrated as separate elements, the modules described and / or illustrated herein can represent portions of a single module or application. In addition, in certain implementations one or more of these modules can represent one or more software applications or programs that, when executed by a computing device, cause the computing device to perform one or more tasks. For example, one or more of the modules described and / or illustrated herein represent modules stored and configured to run on one or more of the computing devices or systems described and / or illustrated herein. In some implementations, a module can be implemented as a circuit or circuitry. One or more of these modules can also represent all or portions of one or more special-purpose computers configured to perform one or more tasks.

[0129] In addition, one or more of the modules described herein transforms data, physical devices, and / or representations of physical devices from one form to another. For example, one or more of the modules recited herein receives sensor data to be transformed, transforms the sensor data, outputs a result of the transformation to detect one or more biomarkers, uses the result of the transformation to provide a notification, and stores the result of the transformation to continuously monitor for the biomarker. Additionally, or alternatively, one or more of the modules recited herein can transform a processor, volatile memory, non-volatile memory, and / or any other portion of a physical computing device from one form to another by executing on the computing device, storing data on the computing device, and / or otherwise interacting with the computing device.

[0130] In some implementations, the term "computer-readable medium" generally refers to any form of device, carrier, or medium capable of storing or carrying computer-readable instructions. Examples of computer-readable media include, without 29ACTIVE 714926490v1limitation, transmission-type media, such as carrier waves, and non-transitory-type media, such as magnetic-storage media (e.g., hard disk drives, tape drives, and floppy disks), optical- storage media (e.g., Compact Disks (CDs), Digital Video Disks (DVDs), and BLU-RAY disks), electronic-storage media (e.g., solid-state drives and flash media), and other distribution systems.

[0131] The process parameters and sequence of the steps described and / or illustrated herein are given by way of example only and can be varied as desired. For example, while the steps illustrated and / or described herein are shown or discussed in a particular order, these steps do not necessarily need to be performed in the order illustrated or discussed. The various example methods described and / or illustrated herein can also omit one or more of the steps described or illustrated herein or include additional steps in addition to those disclosed.

[0132] The preceding description has been provided to enable others skilled in the art to best utilize various aspects of the example implementations disclosed herein. This example description is not intended to be exhaustive or to be limited to any precise form disclosed. Many modifications and variations are possible without departing from the spirit and scope of the present disclosure. The implementations disclosed herein should be considered in all respects illustrative and not restrictive. Reference should be made to the appended claims and their equivalents in determining the scope of the present disclosure.

[0133] Unless otherwise noted, the terms "connected to" and "coupled to" (and their derivatives), as used in the specification and claims, are to be construed as permitting both direct and indirect (i.e., via other elements or components) connection. In addition, the terms "a" or "an," as used in the specification and claims, are to be construed as meaning "at least one of." Finally, for ease of use, the terms "including" and "having" (and their derivatives), as used in the specification and claims, are interchangeable with and have the same meaning as the word "comprising."30ACTIVE 714926490v1

Claims

WHAT IS CLAIMED IS:

1. A device comprising: a plurality of sensor devices comprising: a first sensor device comprising a first sensing material configured to produce a first signal in response to a target material in ambient air; and a second sensor device comprising a second sensing material configured to produce a second signal in response to a background gas in the ambient air; and a computing device comprising: a memory; and a processor coupled to the memory and configured to detect the target material from the ambient air based on sensor data from the plurality of sensor devices.

2. The device of claim 1, wherein the processor is configured to identify the target material using a machine learning model trained to identify a presence of the target material in the ambient air based on the sensor data from the plurality of sensor devices.

3. The device of claim 2, wherein the first signal from the first sensor device corresponds to a measurable change in a property of the first sensing material reacting to a presence of the target material and the second signal from the second sensor device corresponds to a measurable change in a property of the second sensing material reacting to a presence of the background gas.

4. The device of claim 3, wherein the machine learning model is configured to classify the presence of the target material in the ambient air based on the measured changes of the first and second sensing materials.

5. The device of claim 1, further comprising a pump device configured to pump the ambient air to the plurality of sensor devices and controlled by the processor, wherein the pump device further comprises a valve device and tubing between at least the pump device and the plurality of sensor devices.31ACTIVE 714926490v16. The device of claim 5, wherein the valve device is coupled to an intake from a device environment and an exhaust to the device environment, and wherein the ambient air is from the device environment.

7. The device of claim 6, wherein the processor is further configured to control the pump device to continuously pump the ambient air from the device environment and further configured to detect the target material from the device environment by detecting the target material from the ambient air pumped from the device environment.

8. The device of claim 7, wherein the processor is further configured to continuously monitor the device environment for a presence of the target material.

9. The device of claim 5, wherein the valve device is coupled to a sample vessel configured to hold a sample and wherein the ambient air is from the sample vessel.

10. The device of claim 9, wherein the processor is further configured to control the pump device to pump the ambient air from the sample vessel and further configured to detect the target material from the sample by detecting the target material from the ambient air pumped from the sample vessel.

11. The device of claim 1, wherein the processor is further configured to provide a notification in response to detecting the target material.

12. The device of claim 1, wherein the target material corresponds to a volatile organic compound (VOC).

13. The device of claim 1, wherein the target material corresponds to a biomarker indicative of at least one of: a viral infection, a bacterial infection, or a fungal infection.

14. A biomarker detector device comprising:32ACTIVE 714926490v1a sensor array comprising: a first sensing material configured to react to a biomarker in ambient air to produce sensor data; and a second sensing material configured to react to a background gas in the ambient air to produce sensor data; a pump device configured to pump the ambient air to the sensor array; and a computing device comprising: a storage; and a controller coupled to the storage and configured to control the pump device and detect the biomarker based on the sensor data from the sensor array.

15. The biomarker detector device of claim 14, wherein: the controller is configured to identify the biomarker using a machine learning model trained to identify a presence of the biomarker in the ambient air based on electrical change data from the sensor array; the electrical change data from the sensor array corresponds to a measurable electrical change of the first sensing material reacting to a presence of the biomarker in the ambient air and a measurable electrical change of the second sensing material reacting to a presence of the background gas in the ambient air; and the machine learning model is configured to classify the presence of the biomarker in the ambient air based on the electrical change data from the first and second sensor devices.

16. The biomarker detector device of claim 14, wherein the pump device further comprises a valve device and tubing between at least the pump device and the sensor array.

17. The biomarker detector device of claim 16, wherein: the valve device is coupled to an intake from a device environment and an exhaust to the device environment, and wherein the ambient air is from the device environment; the controller is further configured to control the pump device to continuously pump the ambient air from the device environment and further configured to detect the33ACTIVE 714926490v1biomarker from the device environment by detecting the biomarker from the ambient air pumped from the device environment; and the controller is further configured to continuously monitor the device environment for a presence of the biomarker.

18. The biomarker detector device of claim 16, wherein: the valve device is coupled to a sample vessel configured to hold a sample and wherein the ambient air is from the sample vessel; and the controller is further configured to control the pump device to pump the ambient air from the sample vessel and further configured to detect the biomarker from the sample by detecting the biomarker from the ambient air pumped from the sample vessel.

19. The biomarker detector device of claim 14, wherein the controller is further configured to provide a notification in response to detecting the biomarker.

20. A method comprising: pumping an air sample over a sensor array that includes a first sensor device comprising a first sensing material configured to change in response to a target volatile compound in the air sample and a second sensor device comprising a second sensing material configured to change in response to a background gas in the air sample; receiving sensor data from the sensor array in response to pumping the air sample; determining, based on the sensor data, a presence of the target volatile compound in the air sample; and providing a notification in response to detecting the presence of the target volatile compound.34ACTIVE 714926490v1

Images