Pest detection system and method

Abstract

The present invention provides a device for sensing the presence of pests including, but not limited to, bed bugs, and the use of such devices in the methods for the detection of pest insect infestations. In particular, the device is a smart device that comprises a sensing arrangement and an associated controller that together operate to autonomously detect the presence of said pests and, optionally, distinguish between pests of different species. Also provided is a method for training said smart device to detect the current or past presence of a live population of pests.

Description

[0001] Pest Detection System and Method The present invention provides a device for sensing the presence of pests including, but not limited to, bed bugs, and the use of such devices in the methods for the detection of pest insect infestations. In particular, the device is a smart device that comprises a sensing arrangement and an associated controller that together operate to autonomously detect the presence of said pests and, optionally, distinguish between pests of different species. Also provided is a method for training said smart device to detect the current or past presence of a live population of pests. Background of the Invention A wide variety of pests have been biting, or otherwise causing nuisance to, people since the beginning of recorded time. A prime example of such a pest is the common bed bug, Cimex lectularius, which feeds upon human blood, and has a worldwide distribution. The biting nuisance caused by an infestation of such a pest can have physical and mental effects on the human host. Blood feeding can result in allergic reactions to vasodilatory substances in the bed bug saliva and insect bite wounds can become infected by opportunistic pathogens, if inappropriately treated or in immuno- compromised individuals. Furthermore, realisation of a domestic bed bug infestation can affect mental health, causing a range of problems, including emotional distress, anxiety, insomnia and paranoia. For most of the past 40 years, C. lectularius has been an unimportant pest in developed countries. However, in recent years, bed bugs have re-emerged as a significant global pest, with increased rates of infestations occurring in North America, Europe, Australia and elsewhere. The banning of certain insecticides, resistance to other insecticides and the increase in international travel are cited as the main reasons for the re- emergence of this pest problem. Bed bug infestations have a large impact on the hotel and hospitality industry which can be extremely costly. In the US, it is reported to represent around 20% of the US$4 billion extermination insect and pest control services sector. It has grown to be the third largest segment of this market. Monitoring of bed bug and other pest infestations, both spatially and temporally, is essential for the development of targeted treatments. Early detection of bed bug activity is acknowledged as a key factor in reducing the disruption, cost and effort of remedial treatment. The monitoring of bed bug populations can also be used to assess the efficacy of remedial and preventative treatment. However, many currently available pest trap devices, such as simple sticky traps, are based on the principle of random interception and are not efficient at monitoring or controlling pest populations. Many pests, including bed bugs, will preferentially crawl under them, and they are not sufficiently sensitive to detect low-level infestations. Refuge-type monitors also detect the presence of an infestation if the pest insects use the device as a temporary or permanent refuge. However, as with interceptive devices, these will be unlikely to detect low-level infestations. Traps that claim to attract bed bugs currently on the market use either heat or chemical lures or both. However, there are few scientific studies that have tested the efficacy of these devices. Some devices use CO2 to attract bed bugs. However, traps baited with CO2 are expensive and their maintenance is time consuming. Therefore, it is unlikely that such a device would be suitable for long term, wide scale bed bug surveillance. Furthermore, there are safety issues associated with the use of pressurised gas and / or chemical reactions for the production of CO2. In addition, surveillance of pest infestations using these conventional traps requires the user to regularly inspect deployed traps for evidence of infestation which, depending on the size of the surveillance area and number of traps deployed, may be a hugely time consuming and almost unsurmountable task. Therefore, with the high and increasing global prevalence of the common bed bug and other invertebrate pests, including arthropods, such as insects (e.g., moths, cockroaches, beetles, silverfish and weevils); arachnids (e.g., ticks); gastropods, such as slugs and snails; and nematodes, there is an urgent need for a discrete, quick and highly sensitive pest detection device that is capable of being trained to automatically detect, distinguish and notify a user of the presence of an infestation of as few as a couple or even a single pest organism, ideally without needing to physically inspect the device for evidence of an infestation. A similar need exists for the detection and differentiation of vertebrate pests, including tetrapods such as mammals (e.g., rodents) and avian pests, particularly with a view to limiting their impact on disease transmission, crop loss and / or structural damage to buildings. We herein disclose a device that addresses these needs, the device comprising a sensing arrangement and an associated controller that together operate to autonomously detect and indicate the presence of said pests and, optionally, distinguish between pests of different species. Statements of Invention The present invention, in its various aspects, is as set out in the accompanying claims. According to a first aspect of the invention there is provided a pest detection device, said device comprising: a. A sensing arrangement configured to detect the presence of one or more volatile organic compounds (VOCs) or other airborne indicators of pest presence, and to generate one or more signals; b. A controller in communication with said sensing arrangement, configured to receive said signals to indicate the presence of a live population of said pests; and c. A power source configured to provide power to said controller and said sensing arrangement, wherein said sensing arrangement comprises one or more VOC gas sensors. Such an arrangement has been shown to be a highly sensitive pest detection device, that is capable of being trained to automatically detect the presence of as few as a couple or even a single pest organism in a test environment. It will be appreciated that the device can be trained to automatically detect the presence of a pest organism based on species specific VOC based indicator signature(s). Therefore, the device can be adapted to detect and, optionally, distinguish between, any of a wide variety of pests whose presence, either current or past, will result in the generation of a measurable indicative VOC signature. The device of the present invention is not, therefore, limited in its broadest embodiments to the detection of any particular pest species. However, in some embodiments, the pest is an invertebrate, and preferably is selected from an arthropod, gastropod and / or nematode. In particularly preferred embodiments the pest is an arthropod, and still more preferably is selected from insects and / or arachnids. In other embodiments, the pest is a vertebrate, and preferably a tetrapod. In particularly preferred embodiments, the pest is selected from mammals, preferably rodents, and / or birds. Moreover, test data has shown that the device provides exceptional sensitivity in the specific detection of insects, in particular bed bugs. Therefore, in particularly preferred embodiments, the pest detection device is an insect detection device, and still more preferably is a bed bug detection device. As used herein, the term “bed bug” denotes insects in the family Cimicidae, and preferably Cimex spp. A particularly preferred Cimex spp. is Cimex lectularius, the common bed bug. However, the device of the present invention is also provided for use in detecting other Cimex spp., including Cimex hemipterus, the tropical bed bug. However, as noted above other pest invertebrates, could also be detected, aggregated, and / or otherwise monitored using the device of the invention, and so the devices can therefore also be used to detect infestations of pest invertebrates other than bed bugs, including other insects, such as pests of the Blattidae family, and preferably of the genus Blatella (e.g., cockroaches); pests of the Lepismatidae family, and preferably of the genus Lepisma (e.g., silverfish); pests of the Tineidae family, and preferably of the genus Tineola (e.g., clothes moths); pests of the Tenebrionidae family, preferably of the genus Tribolium (e.g., flour beetles); pests of the Curculionidae family, preferably of the genus Sitophilus (e.g., grain weevils); arachnids, such as pests of the Ixodidae (e.g., hard ticks) or Argasidae (e.g., soft ticks) families; gastropods, such as pests of the order Stylommatophora, preferably of the Limacidae, Arionidae, Agriolimacidae or Helicidae families (e.g., slugs and snails); and / or nematodes, such as pest of the phylum Nematoda. Moreover, vertebrate pests could also be detected, aggregated and / or otherwise monitored using the device of the invention, and so the devices can therefore also be used to detect infestations of vertebrates, in particular mammals, such as pests of the order Rodentia, which are preferably pests of the Muridae, Cricetidae, Caviidae or Sciuridae families; and / or birds, i.e. vertebrates of the order Aves, which are preferably pests of the Laridae, Corvidae, Columbidae, Accipitridae and Falconidae families. As used herein, the term “pest” refers to any organism that negatively impacts any human concern, including but not limited to health, property, agriculture, and / or environmental issues. Pests may, therefore, contribute to the spreading of diseases throughout human or animal, in particular livestock, populations, consume crops, damage buildings and / or in any other way impact adversely on human objections. As used herein the term “airborne indicators of pest presence” refers to any measurable VOC signal that is generated within an environment as a consequence of, and attributable to, the presence of said pest. In preferred embodiments, said indicators are VOCs that are emitted by a live population of said pests, and / or are VOCs that are emitted as a consequence of the past presence of said pests such as VOCs emitted from faecal matter and / or dead pests. However, suitable indicators may alternatively or additionally include VOCs emitted by the environment as a consequence of the presence of said pests (e.g. as a result of damage to crops caused by feeding). The sensitivity and / or functionality of the device may be further enhanced by the addition of supplementary sensors to the sensing arrangement. Therefore, in preferred embodiments, the sensing arrangement further comprises one or more supplementary sensors selected from optical sensors, movement sensors, environmental sensors, and combinations thereof. As will be appreciated, a broad range of environmental sensors can be incorporated to detect and / or monitor a variety of stimuli and / or indicators of air quality. Examples of such environmental sensors are those that measure Indoor Air Quality (IAQ), and include but are not limited to heat and / or humidity sensors, pollutant detectors and allergen detectors. In order to maximise device sensitivity, the sensing arrangement may be housed within a partially enclosed environment, so that the VOC and / or other airborne indicators of pest presence can flow towards and be spatially restrained in or around the proximity of the VOC gas sensors. Therefore, in preferred embodiments, the sensing arrangement is housed within a chambered structure comprising but not limited to at least one exterior wall defining an internal chamber having at least one inlet in fluid communication with the exterior surface of the chambered structure. Alternatively or additionally, the device may be disposed in an outer housing. Such a housing may advantageously be adapted to include one or more handles to assist with transporting the device, and / or include one or more mounting points. The device of the invention comprises one or more VOC gas sensors, which detect the presence of VOCs or other airborne indicators of pest presence. However, it has been found that not all VOC gas sensors show the necessary sensitivity to distinguish live pest odour profiles from clean air and / or previous pest odour profiles. For example, thermal conductivity, photo-ionization and infrared based sensors have been shown to be unsuitable for the detection of insects, in particular bed bugs. Therefore, in preferred embodiments, the one or more VOC gas sensors are independently selected from electrochemical gas sensors and metal oxide (MOX) gas sensors. In exemplary embodiments, the one or more VOC gas sensors are MOX sensors, which have been surprisingly found to show the best accuracy, both in single and multiple sensor configurations. As will be readily appreciated, a MOX sensor refers a sensor comprising a heatable metal oxide surface that changes its electrical resistance depending on the oxygen content on its surface, with reducing gases such as VOCs (which would consume oxygen by being combusted on the heated metal oxide surface) reducing resistance. A wide variety of MOX sensors are available commercially and / or could be easily manufactured, any of which can be incorporated into the device of the present invention. Examples of suitable, commercially available MOX sensors include the following: SGP41 (Sensiron); ZMOD4510 (Renesas); ZMOD4410 (Renesas); ZMOD4450 (Renesas); BME668 (Bosch); ENS160 (Sciosense); and MiCS-6814 (SGX Sensortech). In exemplary devices, the MOX sensors are independently selected from SGP41 and ZMOD4510 MOX sensors. The detection accuracy of the device of the invention can be increased by including multiple VOC gas sensors. Therefore, in preferred embodiments, the device comprises at least 2 VOC gas sensors, and more preferably at least 2 MOX sensors. However, it is believed that, if the total number of sensors increases, the risk of false positive results may also increase due to, e.g., signal noise. Therefore, in preferred embodiments, the device preferably comprises no more than 10, more preferably no more than 8 and still more preferably no more than 4 of said VOC gas sensors. As noted above, the invention provides a highly sensitive pest detection device, that can automatically detect the presence of as few as a couple, or even a single, pest organism in a test environment. To achieve automated detection, the device preferably comprises a controller that is configured to receive signals from the VOC gas sensors and apply one or more pest specific detection algorithm to determine the current or past presence of one or more populations of said pest. In some examples, the controller is configured to apply at least two different pest specific detection algorithms to determine, and distinguish between, the current or past presence of populations of at least two different pest species. In particularly preferred embodiments, the controller is configured to apply one or more pest specific detection algorithms to determine the presence of one or more live or dead populations of said pests, or to determine contamination or damage caused by said pests. As will be readily appreciated, the invention is not limited to any particular pest specific detection algorithm, which will depend upon multiple factors including, but not limited to, the pest to be identified along with sensor type, positioning and heating rates. Therefore, the pest specific detection algorithm(s) is / are derived by implementing a device training method wherein any suitable machine learning or mathematically derived model is applied to training data obtained in the presence and / or absence of said pest(s). Therefore, this device training method forms a separate aspect of the invention, and is described elsewhere in more detail. However, in an exemplary embodiment of the first aspect, the device of the invention includes a sensing arrangement comprising two or more MOX sensors, and said one or more algorithm is a machine learning model or a mathematically derived model to determine and / or distinguish between the presence of one or more live populations of said pests, evidence of contamination or damage caused by said pests, or no evidence of current or past pest presence. As will be appreciated, the controller and sensing arrangement do not necessarily require a physical electrical connection between them. Indeed, the controller and sensing arrangement need only be configured to enable the effective transfer of information therebetween. It is envisaged that this can be accomplished through either wireless or physical, e.g. wired, means. In preferred embodiments, however, the sensing arrangement is positioned remote from the controller and, in particularly preferred embodiments, is configured to connect to said controller via wireless communication such as by Bluetooth, Wi-Fi, ZigBee or radio. Wireless communication of this kind enables the storage and / or analysis of VOC data and pest detection algorithms on a centralised network, which in turn permits the remote modification of said algorithms without needing to physically access disassemble the device. In addition, those modifications may be simultaneously transmitted to multiple pest detection devices by virtue of their collective wireless connection to the centralised network. However, in an alternative embodiment, the sensing arrangement is configured to connect to the controller via wired, e.g. electrical or optical fibre connection. The device of the present invention may also comprise at least one indicator that is configured to provide an output signal to a user of the device in response to the controller determining the current or past presence of a live population of said pests. As will be readily appreciated, such an indicator may produce a visual, audio and / or haptic alert signal. Therefore, in preferred embodiments, said indicator comprises one or more lights, speakers and / or motors that are configured to produce a visual, audio and / or haptic alert signal. Such indicators may include device specific hardware or may be in form of a multi-purpose device such as a computer or telephone, that produce an alert signal in the form of a notification delivered through an application or message. Preferably, the at least one indicator is positioned remote from the controller and, in particularly preferred embodiments, is configured to connect to said controller via wireless communication such as by Bluetooth, Wi-Fi, Zigbee or radio. However, in an alternative embodiment, the controller indicator is configured to connect to the controller via wired, e.g. electrical or optical fibre, connection. Alternatively, the indicator may be positioned on or within the device. According to a second aspect, the invention provides a method for training a pest detection device to detect the current or past presence of a live population of pests, wherein said method comprises: a. providing a pest detection device according to the first aspect of the invention; b. obtaining one or more first training data sets by recording said one or more sensor signal outputs when said live pests are not present; c. obtaining one or more second training data sets by recording said one or more sensor signal outputs in the presence of a live population of said pests, or in an environment damaged or contaminated by the past presence of said pests; d. implementing a machine learning model or mathematically derived model using said first and second training data sets obtained in steps b) and c) as inputs to generate a pest specific detection algorithm; and e. configuring said controller to apply, when the device is in use, said pest specific detection algorithm to said one or more sensor signals to determine the presence or absence of a live population of said pests; and / or contamination or damage caused by said pests, and, based on the determined presence of said live pests, or on said contamination or damage caused by said pests, provide an output signal confirming the need for corrective or control measures. It will be appreciated that the device of the invention may on occasion be used together with one or more pest attractant VOC compounds, and that such attractant compounds may mask the VOC signature of the target pest. Therefore, in preferred embodiments, the first training data sets and / or the said second training data sets are obtained in the presence of a pest attractant chemical lure. However, in alternative embodiments, the first training data sets and said second training data sets are obtained in the absence of a pest attractant chemical lure, particularly if the device will not be used together with pest attractant VOC compounds. In some embodiments, the method trains the device to detect the presence of invertebrates, preferably arthropods, and more preferably still an arachnid or an insect. In preferred examples of these embodiments, the method trains the device to detect the presence of (i) an insect selected from the Cimicidae, Blattidae, Lepismatidae, Tineidae, Tenebrionidae and Curculionidae families; and / or (ii) an arachnid selected from the Ixodidae and Argasidae families. In a particularly preferred embodiment, the method trains the device to detect the presence of beg bugs. Preferred pests of the Blattidae family are of the genus Blatella (e.g., cockroaches); preferred pests of the Lepismatidae family, are of the genus Lepisma (e.g., silverfish); preferred pests of the Tineidae family are of the genus Tineola (e.g., clothes moths); preferred pests of the Tenebrionidae family are of the genus Tribolium (e.g., flour beetles); preferred pests of the Curculionidae family are of the genus Sitophilus (e.g., grain weevils); and preferred arachnid pests, are of the Ixodidae (hard ticks) or Argasidae (soft ticks) families (e.g., ticks). In another embodiment, the method trains the device to detect the presence of (i) gastropods; and / or (ii) nematodes. Preferred gastropod pests are slugs and / or snails selected from the Limacidae, Arionidae, Agriolimacidae or Helicidae families. Alternatively or additionally, the method trains the device the detect the presence of pest vertebrates. In preferred examples of these methods, the method trains the device to detect the presence of (i) a mammal; and / or (ii) a bird. Preferred mammalian pests are of the order Rodentia, and more preferably are pests of the Muridae, Cricetidae, Caviidae or Sciuridae families; preferred avian pests are pests of the Laridae, Corvidae, Columbidae, Accipitridae and Falconidae families. Preferably, the first training data sets comprise at least one data set collected when no pests of any species are present, and at least one data set collected when a population of pests of a species other than that to be detected is present. Such a combination of negative control data helps to develop a detection algorithm that is capable of distinguishing between pest infestations from different species. Alternatively or additionally, the said second training data sets preferably comprises at least one data set collected on the presence of a live population of said pests, and at least one data set collected in the presence of faeces, urine and / or exuviae indicative of a past infestation of said pest to be detected. Such a combination of test data sets helps to develop a detection algorithm that is capable of distinguishing between current and past pest infestations. According to a third aspect, the invention provides a pest detection device that is trained to detect the presence of a population of pests according to the method of the second aspect of the invention. According to a fourth aspect, the invention provides a method for detecting the presence of a live population of pests, said method comprising placing one or more pest detection devices according to the first or third aspect of the invention in an environment to be monitored. In preferred embodiments, the device is coupled to a dedicated communication hub that allows for a remote alert to pest presence to be sent securely to a user. As will be readily appreciated, the pest detection device may be disposed in any environment in which pest detection is required. Therefore, in some embodiments, the device is placed in an enclosed space such as room or vehicle and is optionally mounted to a piece of furniture such as a bed. Suitable enclosed spaces include, but are not limited to, any lodgings or sleeping quarters, as well as shipping containers, livestock barns, supermarket stockrooms, shopfloors, and commercial vehicles. Alternatively, the device may be placed in an outdoor environment, and can be used, e.g., to detect forestry and crop damaging pests. Preferred features of each aspect of the invention may be as described in connection with any of the other aspects. Throughout the description and claims of this specification, except where the context requires otherwise due to express language or necessary implication, the words “comprise” and “contain” and variations of the words, for example “comprising” and “comprises” mean “including but not limited to” and do not exclude other moieties, additives, components, integers or steps. Throughout the description and claims of this specification, the term “and / or” includes any and all combinations of one or more of the associated listed elements. Throughout the description and claims of this specification, the singular encompasses the plural unless the context otherwise requires. In particular, where the indefinite article is used, the specification is to be understood as contemplating plurality as well as singularity, unless the context requires otherwise. All references, including any patent or patent application, cited in this specification are hereby incorporated by reference. No admission is made that any reference constitutes prior art. Further, no admission is made that any of the prior art constitutes part of the common general knowledge in the art. Other features of the present invention will become apparent from the following examples. Generally speaking, the invention extends to any novel one, or any novel combination, of the features disclosed in this specification (including the accompanying claims and drawings). Thus, features, integers, characteristics, compounds or chemical moieties described in conjunction with a particular aspect, embodiment or example of the invention are to be understood to be applicable to any other aspect, embodiment or example described herein, unless incompatible therewith. Moreover, unless stated otherwise, any feature disclosed herein may be replaced by an alternative feature serving the same or a similar purpose. An embodiment of the present invention will now be described by way of example only with reference to the following, wherein: Figure 1. Single Sensor Odour Sampling Test Chamber; Figure 2. Customised Mount for Gas Sensor in Odour Sampling Test Chamber (inverted orientation); Figure 3. Multi-Sensor Odour Sampling Test System; Figure 4. Bespoke air flow system within outer shell of the Multi-Sensor Odour Sampling Test System of Figure 3; Figure 5. Heating cycles utilised during collection of Single Sensor training data sets: HP-301 (A); HP-354 (B); HP-411 (C); and HP-501 (D); Figure 6. Single sensor Odour Sampling Test Chamber Training Workflow; Figure 7. Scores plot for PCA on all 14 sensors; Figure 8. Loading plot for PCA with data from all 14 sensors; Figure 9. PCA Scores plot for a thermal conductivity based sensor (A); an infrared based sensor (B); and a photoionisation based sensor (C); Figure 10. PCA Scores plot for an electrochemical sensor; Figure 11. PCA Scores plot for single metal oxide (SGP41) sensor (A); and a combination of 2 metal oxide (SGP41 and ZMOD4510) sensors; Figure 12. PCA Scores plots for a combination of seven metal oxide sensors: Blatella germanica (A); Tineola bisselliella (B); and Lepisma saccharinum (C); Figure 13. PCA Scores plot showing the response of fourteen sensors, including metal oxide and electrochemical based sensors, to: Cavia porcellus odours compared to air (A); Ixodes ricinus odours compared to air (B); and Sitophilus granarius odours compared to clean food substrate (C). Figure 14. PCA Scores plot showing the response of six sensors, including metal oxide and electrochemical based sensors, to Fratercula arctica odours compared to air. Figure 15. PCA Scores plot showing the response of fourteen sensors, including metal oxide and electrochemical based sensors, to three mixed-species populations, comprising: Tineola bisselliella and Cimex lectularius (A); Lepisma saccharinum and Cimex lectularius (B); and Sitophilus granarius and Tribolium castaneum (C). Table 1. Sensors Provided in Multi-Sensor Air Flow System of Figure 4; Table 2. Bed Bug Detection Results of AI derived model challenge using the Single Sensor Odour Sampling Test Chamber of Figure 1. Table 3. Predicted outcomes from a ML-based model of gas sensors detection of uncontaminated food, and food with flour beetles. Experimental As a proof of concept, and to demonstrate the effective training of prototype smart pest detection devices, single and multi-sensor systems were prepared, and detection sensitivity was assessed using bed bugs (Cimex lectularius) as a model pest organism. It has been shown that the devices can be trained for the detection of other pests, in particular, birds, (exemplified as a proof of concept using data obtained for Atlantic puffins (Fratercula arctica); mammals, such as guinea pigs (Cavia porcellus); and / or other arthropods, such as cockroaches (Blattella germanica), silverfish (Lepisma saccharinum), clothes moths (Tineola bisselliella), flour beetles (Tribolium castaneum), grain weevils (Sitophilus granarius) and ticks (Ixodes ricinus). Additional proof of concept experiments were conducted using the multi-sensor system to assess detection specificity toward individual species within a population comprising more than one species. In particular, species-specific detection was investigated on three mixed-species populations, comprising: (i) clothes moths and bed bugs; (ii) silverfish and bed bugs; or (iii) grain weevils and flour beetles. MATERIALS Sensors Single Sensor System For initial proof of principle experiments, a commercial (Bosch Sensortec BME688 AI Gas Sensor) gas sensor, which allows users to train sensors to detect VOC odour profiles and, using the associated software (BME AI-Studio), build algorithms to automatically determine the presence of specific pests based on detected odour profiles, was incorporated into a still air odour detection apparatus (Figure 1) that was designed to allow full cleaning between uses. Specifically, a customised mounting method was designed and fabricated to repeatedly locate the development board and a single sensor board (which comprises eight copies of the same BME688 sensor to allow simultaneous data collection using different heating profiles) to the test chamber lid (Figure 2). The mounting brackets were manufactured from aluminium, as this material can be odour neutralised with IPA, unlike tape and / or plastic brackets. Further, a range of interchangeable spacer tubes were prepared to allow the vertical offset distance of the sensor boards relative to the lid (and so the depth of penetration into the detection apparatus) to be altered as required. Multi-Sensor System A bespoke air sampling system was designed and developed to assess the sensitivity of combinations of gas sensor types (Figure 3). In this apparatus, air is pumped into a bespoke air flow chamber (Figure 4), linked with 14 different sensors (Table 1). Data was collected as text files which permitted separate analysis of the data for each individual sensor. Test organisms Cimex lectularius, the common bed bug, is a nocturnal blood-feeding parasite of humans and animals. Male and female, Sweden Field Strain, bed bugs (Cimex lectularius) were obtained from CimexStore. They were stored separately sexed in plastic colony pots (60 x 40 mm) in a controlled environment (27°C ± 2°C, 70% humidity ± 10%, 12h:12hr Light:Dark cycle) inside an incubator until used in testing. Similarly, Blatella germanica (cockroaches), Lepisma saccharinum (Silverfish) and Tineola bisselliella (clothes moths) were obtained from commercial sources and stored, separately sexed, according to the supplier’s recommendations until used in testing. In addition, Tribolium castaneum (flour beetles), Sitophilus granarius (grain weevils) and Ixodes ricinus (ticks) were obtained from commercial sources and stored, separately sexed, according to the supplier’s recommendations until used in testing. For the vertebrate experiments, Fratercula arctica (Atlantic puffin) odour data was collected from inhabited burrows and compared to odours collected from empty burrows. Six sensors were exposed to these odours for 2 minutes. Cavia porcellus (guinea pigs) were obtained from a commercial source, stored and separately sexed according to the supplier’s recommendations until used in testing. METHODS Single Sensor System The system was trained to detect bed bugs by collecting a series of training data sets or models. A single bed bug, a bed bug exposed paper (to simulate a previous bed bug infestation) and a commercial, VOC-based, lure (BugscentsTM, Arctech Innovation) were sampled using the prototype single sensor system for 30 minutes each. Individual data sets were collected using four of the seventeen different sensor pre-set heating cycles, denoted HP-301, HP-354, HP-411 and HP501, respectively. The specific heating cycles used in these training data sets / models are shown in Figure 5(A) – 5(D). Following data collection, the BME-AI studio was utilised to automatically conduct all sensor training, including internal training, to produce a classification algorithm specific for each of the four tested heat cycle data sets. Each of the classification algorithms produced by the BME-AI studio could then be downloaded as a BSEC Config file, describing a combination of sensor presets and data processing, for subsequent use in “trained” BME688 sensor-based detection systems. A schematic of the sensor training and testing workflows is provided in Figure 6. Multi-Sensor System Initially, the multi-Sensor system was also trained by collecting the following training data sets: A single bed bug; a bed bug exposed paper (to simulate a previous bed bug infestation); and air. Sampling was conducted using the multi-sensor system for 30 seconds each (10 replicates). Subsequently, the same system was trained by collecting the following training data sets: (i) a single cockroach vs. air; (ii) a single silverfish vs. air; and (iii) a single clothes moth vs. air. Sampling was again conducted using the multi-sensor system for 30 seconds each (16-20 replicates). As additional proof of concept experiments, the same system was trained by collecting the following training data sets: (i) guinea pig faeces vs. air (NB it should be appreciated that other byproducts, e.g urine, that are indicative of pest presence can be used as a data set, either as an alternative or in addition to faeces; (ii) five ticks vs. air; (iii) a single grain weevil vs. air; (iv) a single Atlantic puffin vs. air (see, e.g., Figure 13 and Figure 14). Sampling was again conducted using the multi-sensor system for 30 seconds each (16-20 replicates). In yet further proof of concept experiments, the same system was trained by collecting the following training data sets involving mixed-species populations: (i) clothes moths vs. bed bugs; (ii) silverfish vs. bed bugs; and (iii) grain weevil vs. flour beetle (see, e.g., Figure 15). Sampling was again conducted using the multi-sensor system for 30 seconds each (16-20 replicates). Statistical Analysis For the single sensor system data, BME AI-Studio software was utilized to develop heat profile specific pest detection / classification algorithms. Data were classified using the proprietary neural net architecture (ADAM optimizer). The samples were exposed to the sensors for a total of 30 minutes, depending on the individual heat profile, allowing between 391 and 783 individual heat cycles to run per data set. These data were then split 30:70, with 70% of the data used to train the algorithm and 30% used to validate the algorithm. By comparing the model predictions of the validation data set with the actual categories, accuracy rates were calculated using the formula: ^^^^^^^^^^^^^^^^ =^^^^^^ ^^^^ ^^^^^^ ^^^^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^ ^^^^ ^^^^^^ ^^^^^^^^^^^^^^ ^^^^^^ ^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ In the multi-sensor system, there were 31 data outputs from the 14 test sensors (with variable temperature settings available for each sensor). These data were analysed using a three component Principle Component Analysis on Multisens Analyzer (Standard 2.1.0.54, JLM).

[0002] Example 1: Single Sensor System Four algorithms, one for each of the tested heat cycles, were assessed mathematically, and detection accuracy varied from 87% to 94%. The data for the most accurate model, i.e. HP-411, is shown in Table 2. Example 2: Multi-Sensor System (Cimex lectularius) Initially, a PCA was run with all available sensors and data outputs. The resultant scores plot showed good although not ideal resolution between the three odour profiles, i.e. single bed bug; a bed bug exposed paper; and air (Figure 7), while the loadings plot indicated that several sensors were of no importance to the accuracy of the model (Figure 8). Subsequent PCA analysis, using subsets of available sensors and data outputs, confirmed that sensitivity was dependent on sensor type. Notably, thermal conductivity, photo-ionization and infrared based sensors did not show any sensitivity to distinguishing live bed bug odours from air or bed bug associated odours (Figure 9(A), 9B, 9(C)). Electrochemical sensors performed better, with the best performing examples of such sensors showing some resolution, but not full separation of the odour signals (Figure 10). However, metal oxide-based sensors were found to show the best lone function, being able to distinguish the bed bug and bed bug associated odours from an air sample (Figure 11(A)). Further, a combination of two such metal oxide sensors provided enhanced discrimination between the bed bug and bed bug associated odours (Figure 11(B)). Example 3: Multi-Sensor System (Blatella germanica, Lepisma saccharinum, Tineola bisselliella) For each pest insect test, a PCA was run with seven available metal oxide sensors and data outputs. The resultant scores plots showed, for each insect, good resolution between the two tested odour profiles (Figure 12), with full separation of the odour signals observed in both the clothes moth (Figure 12(B)) and silverfish (Figure 12(C)) tests. Example 4: Multi-Sensor System (Cavia porcellus, Ixodes ricinus, Sitophilus granarius, Fratercula arctica and Tribolium castaneum) In the pest rodent test, fourteen available sensors, including metal oxide and electrochemical based sensors, where exposed to odours from Cavia porcellus (guinea pig) faeces, and to ambient air, for 2 minutes respectively. The PCA Scores plot generated from the raw rodent sensing data shows full separation between Cavia porcellus odours and air along the first and second principal components, indicating the sensing system’s ability to accurately detect Cavia porcellus odours in air (Figure 13(A)). In the bird trial, odours were collected from inhabited puffin (Fratercula arctica) burrows and compared to odours collected from empty burrows. Six sensors were exposed to these odours for 2 minutes. The PCA Scores plot generated from the raw bird sensing data shows some separation between Fratercula arctica odours and air along the first and second principal components, indicating moderate sensing performance (Figure 14). Additional arthropod experiments were performed, involving other insects, specifically grain weevils (Sitophilus granarius) as well as arachnids, specifically ticks (Ixodes ricinus). Individual grain weevils (Sitophilus granarius) were placed in 250ml glass bottles and left overnight to allow the headspace to develop. PCA showed some separation between the odours of food infested with grain weevils and uninfested food, respectively, along the principal components, demonstrating moderate sensing performance (Figure 13(C)). In tick trials, five specimens were placed in glass bottles, and the headspace sampled. Fourteen sensors were exposed to these odours and to ambient air for 2 minutes. Eight sensors including metal oxide, and electrochemical based sensors, were found to show strong responses to tick-related odours. Subsequent PCA analysis shows full separation between tick odours and air, confirming the system’s ability to accurately detect ticks in air (Figures 13(B)). In additional trials, a machine learning-based model was created using gas sensor response to uninfested food (grain), and flour beetle (Tribolium castaneum) infested food odours (Table 3). The model predicted the response outcomes with an accuracy of over 85%. Example 5: Multi-Sensor System (detection of individual species within mixed- species populations: (i) clothes moths vs. bed bugs; (ii) silverfish vs. bed bugs; and (iii) grain weevils vs. flour beetles) As part of a series of proof of concept experiments to demonstrate the ability of the multi-sensor system to detect the presence of individual species in the presence of others, odour response trials were performed on mixed-species populations of insects. The PCA scores plot of the response of multi-sensor system’s response to a mixture of clothes moth and bed bug-related odours (Figure 15(A)) shows some separation, whilst the responses to mixtures of silverfish and bed bugs (Figure 15(B)), and grain weevil and flour beetle exhibit full separation (Figure 15(C)), indicating effective interspecies detection within mixed-species samples using the multi-sensor system according to the present invention. SUMMARY Proof of principle experiments have successfully demonstrated that, when incorporated into a smart pest detection device, VOC gas sensors, in particular metal oxide VOC sensors, are capable of detecting pest insect, in particular bed bug, odours, and distinguishing them from VOC-based chemical lures and / or from odours relating to previous infestations. In contrast, alternative VOC gas sensors based on alternative technology, e.g. thermal conductivity, photo-ionization, infrared or electrochemical based sensors, were shown to provide vastly inferior detection sensitivity. Further, it has been found that system sensitivity can be improved by the inclusion of more than one, e.g. at least two, such metal oxide VOC sensors. In this regard, it is believed that the inclusion of a combination of several (e.g. more than 10, 8 or 4), such sensors may have a deleterious effect on overall sensitivity of the system, not least due to the generation of false positive results due to signal noise. In addition, and whilst the proof of principle work provided herein has been completed primarily using bed bugs as the pest subject it is expected, and has been shown that, following exposure to appropriate pest-specific training data, the smart detection device would be able to likewise detect and distinguish between different target pest organisms. Moreover, proof of concept work on interspecies detection within mixed-species populations confirms that, following exposure to multiple pest-specific training datasets, the smart detection device is able to detect and distinguish between different target pest organisms within the same samples.

[0003] Table 1 Table 2 Table 3 Actual Food contaminated with Uncontaminated food redflour beetles Uncontaminated 101 13 food Predicted Food contaminated with redflour 15 103 beetles

Claims

Claims 1) A pest detection device, comprising: a. A sensing arrangement configured to detect the presence of one or more volatile organic compounds (VOCs) or other airborne indicators of pest presence, and to generate one or more signals; b. A controller in communication with said sensing arrangement, configured to receive said signals to indicate the presence of a live population of said pests; and c. A power source configured to provide power to said controller and said sensing arrangement, wherein said sensing arrangement comprises, but is not limited to one or more VOC gas sensors. 2) The pest detection device according to claim 1, wherein said sensing arrangement further comprises one or more supplementary sensors selected from optical sensors, movement sensors, environmental sensors e.g. heat sensors, humidity sensors, pollutant detectors and / or allergen detectors, and combinations thereof. 3) The pest detection device according to claim 1 or claim 2, wherein the sensing arrangement is housed within a chambered structure comprising at least one exterior wall defining an internal chamber having at least one inlet in fluid communication with the exterior surface of the chambered structure. 4) The pest detection device according to any of the preceding claims, wherein the sensing arrangement is disposed in an outer housing. 5) The pest detection device according to any of the preceding claims, wherein said one or more VOC gas sensors are independently selected from metal oxide (MOX) gas sensors and electrochemical gas sensors.6) The pest detection device according to claim 5, wherein said one or more VOC gas sensors are MOX gas sensors. 7) The pest detection device according to any of the preceding claims, wherein said sensing arrangement comprises at least 2, and preferably no more than 10, VOC gas sensors. 8) The pest detection device according to any of the preceding claims, wherein said controller is configured to receive said one or more sensor signals and apply one or more pest specific detection algorithms to determine the current or past presence of one or more populations of said pest. 9) The pest detection device according to claim 8, wherein said controller is configured to apply at least two different pest specific detection algorithms to determine, and distinguish between, the current or past presence of populations of at least two different pest species. 10) The pest detection device according to claim 8 or claim 9, wherein said controller is configured to apply one or more pest specific detection algorithms to determine the presence of one or more live or dead populations of said pests, or to determine contamination or damage caused by said pests. 11) The pest detection device according to claim 10, wherein said sensing arrangement comprises two or more MOX sensors, and said one or more algorithm is a machine learning model or a mathematically derived model to determine and / or distinguish between the presence of one or more live populations of said pests, evidence of contamination or damage caused by said pests, or no evidence of current or past pest presence. 12) The pest detection device according to any preceding claim, wherein said sensing arrangement is positioned remote from the controller.13) The pest detection device according to claim 12, wherein said sensing arrangement is configured to connect to said controller via wireless communication such as by Bluetooth, Wi-Fi, Zigbee or radio. 14) The pest detection device according to any one of claims 1 to 11, wherein said sensing arrangement is configured to connect to the controller via wired, e.g. electrical or optical fibre connection. 15) The pest detection device according to any of the preceding claims, further comprising at least one indicator configured to provide an output signal to a user of the device in response to said controller determining the current or past presence of one or more live populations of said pests. 16) The pest detection device according to claim 15, wherein said indicator comprises one or more lights, speakers and / or motors that are configured to produce a visual, audio and / or haptic alert signal. 17) The pest detection device according to claim 16, wherein said indicator is positioned remote from the controller and wherein said indicator is optionally configured to connect to said controller via wireless communication such as by Bluetooth, Wi-Fi, Zigbee or radio. 18) The pest detection device according to claim 15 or claim 16, wherein said indicator is positioned on or within the device. 19) The pest detection device according to any one of the preceding claims, wherein said pests are invertebrates, and are optionally selected from arthropods; gastropods; and / or nematodes. 20) The pest detection device according to claim 19, wherein said arthropods are: a. of the Cimicidae family, and are preferably of the genus Cimex;b. of the Blattidae family, and are preferably of the genus Blatella; c. of the Lepismatidae family, and are preferably of the genus Lepisma; d. of the Tineidae family, and are preferably of the genus Tineola; e. of the Ixodidae family, and are preferably of the genus Ixodes; f. of the Curculionidae family, and are preferably of the genus Sitophilus; and / or g. of the Tenebrionidae family, and are preferably of the genus Tribolium. 21) The pest detection device according to any one of claims 1 to 18, wherein said pests are vertebrates, and are optionally selected from mammals, and preferably rodents; and / or birds. 22) A method for training a pest detection device to detect the current or past presence of a live population of pests, said method comprising: a. providing a pest detection device according to any of claims 1 to 21; b. obtaining one or more first training data sets by recording said one or more sensor signal outputs when said live pests are not present; c. obtaining one or more second training data sets by recording said one or more sensor signal outputs in the presence of a live population of said pests, or in an environment damaged or contaminated by the past presence of said pests; d. implementing a machine learning model or mathematically derived model using said first and second training data sets obtained in steps b) and c) as inputs to generate a pest specific detection algorithm; and e. configuring said controller to apply, when the device is in use, said pest specific detection algorithm to said one or more sensor signals to determine the presence or absence of a live population of said pests; and / or contamination or damage caused by said pests, and, based on the determined presence of said live pests, or on said contamination ordamage caused by said pests, provide an output signal confirming the need for corrective or control measures. 23) The method according to claim 22, wherein said first training data sets and said second training data sets are obtained in the presence of a pest attractant chemical lure. 24) The method according to claim 22, wherein said first training data sets and said second training data sets are obtained in the absence of a pest attractant chemical lure. 25) The method according to any of claims 22 to 24, wherein said pests are of the Cimicidae family, and are preferably of the genus Cimex. 26) The method according to an of claims 22 to 25, wherein said one or more first training data sets comprises at least one data set collected when no pests of any species are present, and at least one data set collected when a population of pests of a species other than that to be detected are present. 27) The method according to any of claims 22 to 26, wherein said second training data sets comprises at least one data set collected in the presence of faeces, urine and / or exuviae indicative of a past infestation of said pest to be detected. 28) A pest detection device trained to detect the presence of a population of pests according to the method of any of claims 22 to 27. 29) A method for detecting the presence of a live population of pests, said method comprising placing one or more pest detection device(s) according to any of claims 1 to 21 or 28 in an environment to be monitored, optionally wherein saiddevice is coupled to a dedicated communication hub that allows for a remote alert to pest presence to be sent securely to a user. 30) The method according to claim 29, wherein said device is placed in an enclosed space such as room or vehicle, and is optionally mounted to a piece of furniture such as a bed. 31) The method according to claim 29, wherein said device is placed in an outdoor environment.

Images