System and method for detecting acute abdomen syndrome in horses

Abstract

The present invention relates to a system and method for the automatic and early detection of acute abdomen syndrome in horses by analysing audio recordings. The system comprises a capturing device, which is attached to the abdomen of a horse by means of a fastening means, to obtain an audio recording. The analysis is carried out by means of a processing device, with an artificial intelligence system, which identifies and assesses the peristaltic sounds of the horse to classify the condition of its intestinal motility and provide a diagnosis.

Description

[0001] SYSTEM AND METHOD FOR DETECTING ACUTE ABDOMEN SYNDROME IN EQUINES

[0002] Background of the Invention

[0003] The present invention belongs to the field of veterinary medicine, and focuses specifically on the detection of acute abdomen syndrome in horses by means of a system that allows the capture and analysis of the peristaltic sounds of these animals.

[0004] Acute abdominal syndrome, also known as equine colic, is a leading cause of acute abdominal pain in horses and a frequent reason for emergency veterinary care in these animals. Acute abdominal syndrome is particularly prevalent in stabled sport horses and has been identified not only as a serious animal health problem but also as a cause of forced confinement of trained athletes and, ultimately, as a significant cause of economic losses for horse breeding operations worldwide.

[0005] Due to the high morbidity and mortality associated with acute abdominal syndrome, the average cost of treating a single case of colic in a horse is not only US$215.60, but also a concerning recurrence rate of 48.2% after treatment, and a 11.9% chance of developing laminitis within one year. Thus, when a horse suffers from colic, in addition to the animal's health problems, there is a significant risk of recurrence and complications, as well as economic losses for the horse's owner, estimated at an average annual cost of approximately US$36,806 per horse, not including losses resulting from the animal's death.

[0006] It is important to note that horses instinctively conceal symptoms when they are uncomfortable, which presents a significant challenge for the timely diagnosis of acute abdominal syndrome. This characteristic of horses adds further complexities for veterinarians due to the rapid progression and potential for the condition to worsen, thus impacting effective treatment. Additionally, in controlled environments such as stables or paddocks, sport horses are particularly susceptible to colic, with a higher incidence during nighttime hours when human supervision is minimal or nonexistent.

[0007] In advanced stages, acute abdomen syndrome can lead to severe complications, including multiple organ failure and sudden death from shock. In addition to the health implications for the horse, owners and caretakers face emotional and financial impacts, as well as a significant loss of time.

[0008] While surgery is an option, it is not always the best solution for treating acute abdominal syndrome due to its low success rate. Furthermore, potential postoperative complications and high costs must be considered.

[0009] As an example, the cost of colic surgery is estimated to start at CLP 3,750,000, not including additional postoperative expenses. Long-term complications include a significant increase in the incidence of colic after surgery. Studies indicate that horses that have undergone abdominal surgery for colic are between 2.8 and 7.6 times more likely to experience colic again, during a risk period that extends up to 100 days after the surgical procedure.

[0010] Approximately 85% of horses are able to return to their pre-operative activities or reach the expected performance level, which corresponds to 80% of their performance prior to their acute abdominal syndrome condition. In some cases, the recovery period can extend beyond the estimated 6 months, reaching up to a year.

[0011] Surgeries for acute abdominal syndrome have a significant impact on the market value and sale of horses. A history of colic surgery is often a deciding factor for potential buyers, considering the risks associated with complications and the possibility of recurrence. Thus, the market value of a horse that has undergone surgery can be negatively affected, reducing its sales opportunities in the equine market. The conventional diagnosis of acute abdominal syndrome is based on the evaluation and observation of various factors in the horse, such as its behavior, heart rate, body temperature, and monitoring of peristaltic sounds of abdominal motility using a stethoscope. This practice requires a significant investment of time and effort from veterinary professionals.

[0012] The present invention provides a diagnosis of acute abdomen syndrome in horses, which is based on listening to and analyzing the sounds emitted in the animal's cecum.

[0013] The cecum is a fundamental anatomical structure in the equine digestive system, located at the junction of the small intestine and the initial portion of the large intestine, the colon. The cecum is sac-like and plays a crucial role in the fermentation of fibrous foods, characteristic of the equine diet.

[0014] Within the cecum, microorganisms facilitate the breakdown of plant fiber, an essential process for the horse to extract and utilize nutrients from its feed. The cecum acts as a fermentation center in the gastrointestinal tract, playing a vital role in the effective digestion of certain nutrients.

[0015] During a horse's digestive process, certain cycles and relevant sounds can be detected. The present invention identifies the cecal cycles and the different sounds produced during these cycles.

[0016] The cecal cycle in horses encompasses the complete passage of the food bolus from its entry into the cecum to its elimination. In other words, the cecal cycle includes the entry of digestive material into the cecum, through fermentation and peristaltic movements, to its final release from the cecum.

[0017] During the cecum cycle, two particularly relevant sounds can be detected: the borborygmus and the sound emitted by the discharge stage.

[0018] At the beginning of the cecal cycle, a sound called borborygmi occurs, caused by the movement of fluids and gases in horses. This sound is generated when the digestive contents, a mixture of liquid and air, pass through the ileocecal valve, the junction between the small and large intestines, and enter the cecum. This sound is indicative of both the normal muscular activity of the cecum, which helps process and ferment the material, and the passage of the digestive contents through the gastrointestinal tract. Borborygmi is a sign of healthy gastrointestinal activity and is especially noticeable during the fermentation phase and in preparation for the discharge of processed material from the cecum.

[0019] The cecal cycle ends with the unloading stage, in which the digestive contents are transferred from the cecum into the colon, specifically through the cecocolic valve. Unloading occurs after the material has undergone a period of fermentation in the cecum, where the fibrous components of the horse's diet are broken down. The unloading stage produces a sound that can be identified.

[0020] Since the sounds of borborygmi and gurgling are emitted during the cecal cycle, auscultation of the cecal cycle is an integral part of the clinical examination performed by veterinarians to assess the horse's digestive health, allowing them to identify potential abnormalities such as equine colic. To identify the cecal cycles during auscultation, the veterinarian must listen for and accurately identify these sounds.

[0021] It is important to add that during auscultation, mild or small borborygmi and / or incomplete discharges (borborygmi and partial discharges) may occur. These are not indicative of the beginning or end of a cecal cycle and may be found within or outside of a cycle. Therefore, to properly identify cecal cycles, it is essential that the veterinary specialist distinguish between borborygmi and complete discharges, which effectively mark the beginning and end of a cycle.

[0022] Another crucial aspect during auscultation is identifying the frequency and regularity of abdominal sounds. Significant variations or the complete absence of borborygmi, the sound emitted during the discharge phase, or cecal cycles can be indicative of problems in the horse's digestive system and the presence of acute abdomen syndrome. These changes can suggest anything from an alteration in the normal function of the cecum to more serious conditions.

[0023] In particular, the complete absence of sounds during auscultation is cause for concern, as it may indicate a decrease or cessation of activity in the cecum, which could be a symptom of an obstruction or blockage in the digestive system. This situation requires immediate veterinary attention, as it could compromise the horse's health and life.

[0024] The practice of auscultation requires extensive knowledge and experience on the part of the veterinarian, as he must identify, at specific points on the horse's abdomen, particular and characteristic sounds of the horse's digestive activity.

[0025] During auscultation, there are several potential errors that can impact the accuracy of a veterinarian's diagnosis. One of the most significant is the misinterpretation of sounds, as the ability to distinguish between normal and abnormal sounds depends heavily on experience and knowledge. Furthermore, sounds can vary depending on the horse's diet, physical activity, and overall health. A misinterpretation of sounds can lead to a misdiagnosis. Time constraints and the inability to perform prolonged auscultations present additional obstacles to accurate and early diagnosis.

[0026] Finally, depending on the environment, external noises can significantly interfere with the veterinarian's ability to hear and identify abdominal sounds, which can lead to errors in assessment.

[0027] In the state of the art, there are developments that attempt to provide an alternative to the aforementioned veterinary diagnosis, for example, through the use of various sensors, and the analysis of the data collected by said sensors.

[0028] Within the state of the art is document US10561365, which discloses a system and method for monitoring anomalies in horses, such as changes in their health condition, abnormal behavior, diseases, allergies, colic, alterations in sleep patterns, and the identification of risk situations such as being trapped. This system includes various sensors to collect data on several equine parameters, a database to store this information, and a processor to analyze the data and detect abnormal behaviors.

[0029] Also noteworthy is patent US7335168B2, which describes a system for monitoring the health and behavior of animals, particularly in livestock settings. The system utilizes various sensors, including motion, temperature, heart rate, respiratory rate, and acoustic sensors. It also includes a processor to identify conditions of interest based on a database of stored patterns, threshold tables, and heuristic engines.

[0030] However, in the state of the art there is no system and method that allows the capture and analysis of a horse's peristaltic sounds, specifically the sounds emitted during the cecal cycle, and that provides an early diagnosis of the state of its intestinal motility.

[0031] In light of the above, a system and method for the automatic and early detection of acute abdomen syndrome in horses is required, which is non-invasive, uses cecal auscultation as its principle, is easy for a user to operate, and overcomes the problems mentioned.

[0032] Brief description of the invention

[0033] The present invention relates to a system and method for the automatic detection of acute abdominal syndrome in horses through the recognition of peristaltic sounds, corresponding to borborygmus and the sound of the discharge stage, and the identification and analysis of cecal cycles.

[0034] The system of the invention comprises a capture device, which includes a sound recorder, a microphone, and a bell. This capture device is attached to the abdomen of a horse by means of a fastening device.

[0035] The recording device captures the horse's peristaltic sounds for subsequent processing and analysis by a processor incorporating an artificial intelligence system. This system then provides a diagnosis of acute abdominal syndrome in the horse. The diagnosis can be promptly communicated to the horse's owner and veterinarian.

[0036] The analysis of the processing device automatically identifies, from an audio recording, the horse's blind cycles, and the existence, duration and frequency of the sounds of borborygmus and discharge, and with this information issues a diagnosis.

[0037] The invention also comprises a display device, in which the processed audios, the areas of interest of each audio recording made, and relevant information related to the analysis performed are displayed in detail.

[0038] Brief description of the figures

[0039] The relevant figures of the invention are indicated below, which should not be considered as limitations of the invention.

[0040] Figure 1. This corresponds to a basic flow diagram of the invention.

[0041] Figure 2. Corresponds to a perspective view of the capture device of the invention.

[0042] Figure 3. This corresponds to a front view of the fixing cover of the capture device.

[0043] Figure 4. This corresponds to a bottom view of the fixing cover of the capture device.

[0044] Figure 5. This corresponds to a front view of the insulating basin of the capture device.

[0045] Figure 6. This corresponds to a front view of the microphone bell of the capture device.

[0046] Figure 7. This corresponds to a cross-sectional view of the capture device.

[0047] Figure 8. This corresponds to a photograph showing the capture device positioned on a horse.

[0048] Figure 9. This corresponds to a flowchart of the diagnostic classification. Figure 10. This corresponds to a screenshot of an audio recording analyzed with its labels.

[0049] Detailed description of the invention

[0050] The system and method for automatic, early and non-invasive detection and diagnosis of acute abdomen syndrome in horses of the present invention comprises different elements, which allow the capture, analysis and processing of the sounds emitted in the cecum cycle.

[0051] Figure 1 shows a flowchart of the invention. In the first stage, using a capture device (10), at least one audio recording is made of the peristaltic sounds emitted by a horse. These recordings are transferred to a processing device (20), which processes and analyzes them using an artificial intelligence system to generate diagnoses. The diagnoses and recordings can be stored in a database (30) or in the cloud. The information from the recordings and diagnoses can be viewed using a display device (40).

[0052] If a diagnosis is negative, the system can send an alert to the animal's owner and veterinarian, or to one or more external devices. The veterinarian can then evaluate the diagnosis for validation and recommend a treatment plan for the animal.

[0053] The main components of the invention correspond to the capture device (10) and the processing device (20).

[0054] The capture device (10) consists primarily of an insulating basin (11) containing a bell (13). The bell (13) houses a microphone (14) and a diaphragm (131). The upper part of the insulating basin (11) has magnets (111) (see Figure 5). In a preferred configuration, the insulating basin (11) has three magnets (111).

[0055] For improved sound insulation, the insulating basin (11) comprises a base sponge (112) and an inner sponge, the latter being located in the empty space inside the insulating basin (11). The inner sponge (not shown in the figures) provides sound insulation around the bell (13). On the other hand, the base sponge (112) provides insulation at the base of the capture device (10).

[0056] The capture device (10) has a lid (12) which, in conjunction with the insulating basin (11), comprises pins (122) located in internal grooves (121). In a preferred embodiment, the lid (12) has three pins (122). Figures 3 and 4 show the lid (12) of the invention. Figure 2 shows the insulating basin (11) with the lid (12) and the base sponge (112).

[0057] Figure 7 shows the main elements of the capture device (10) in a cross-sectional view. Specifically, the bell (13), which includes a diaphragm (131) and a microphone (14), is visible. The bell (13) is located and installed within the insulating basin (11). Figure 7 also shows the cover (12), a slot (121) with magnets (111, 122), and the base sponge (112).

[0058] The capture device (10) also includes a means of securing it to the horse, which corresponds to a restraint layer (16). This layer is an elastic fabric band, made of, for example, Lycra or Spandex, or a similar material, that adjusts and adapts to the horse's abdomen. For securing it to the horse, the restraint layer (16) includes a fastening mechanism, which may be of the Velcro and / or zipper type, and / or made of a similar material. The restraint layer (16) covers all areas relevant to the auscultation process for detecting equine colic.

[0059] When making audio recordings, the insulating basin (11), with the bell (13) and microphone (14) inside, is placed between the horse's skin and the support layer (16). For proper fixation, the cap (12) is placed over the support layer (16), directly over the insulating basin (11). In this way, the pins (122) on the inside of the cap (12) align with the pins (111) on the top of the insulating basin (11).

[0060] The above features allow the insulating basin (11) and the capture device (10) to be securely, but not permanently, fixed in the desired recording position, even if the animal makes sudden movements. Additionally, since the insulating basin (11) is secured by the cap (12) with handles, it can be repositioned quickly and easily. Simply remove the cap, place the insulating basin (11) in the new desired position, and secure it again with the cap (12). This allows the user to easily make multiple audio recordings in different locations.

[0061] Once the capture device (10) is installed and fixed, the base sponge (1 12) makes contact with the horse's coat, isolating unwanted noises and sounds, which may come, for example, from the contact of the capture device (10) with the animal's skin and due to small movements of the insulating basin

[0062] (1 1 ) relating to the skin and fur of the animal, for example, by means of its breathing movements, even though it is fixed with the lid (12).

[0063] The base sponge (1 12) is disc-shaped, with a circular cutout in the center, allowing the diaphragm (131) to be in direct contact with the animal.

[0064] During the recording of peristaltic sounds, the base sponge (1 12) manages to considerably reduce unwanted noise produced between the insulating basin (1 1 ) and the fur and skin of the animal, thus acting as an insulator of unwanted noises and sounds.

[0065] Figure 8 shows the capture device (10) fixed to the horse, thanks to the cover (12) and the fastening layer (16).

[0066] For better listening to the cecal cycle and accurate detection of peristaltic sounds, the capture device (10) should be placed on the horse's abdomen, preferably the upper abdomen near the rear of the horse.

[0067] Notwithstanding the foregoing, thanks to the fact that the capture device can be relocated to multiple positions, thanks to the clamping layer (16) and the cover

[0068] (12), the user can easily take multiple samples or audio recordings in different areas of the horse's abdomen.

[0069] Once the capture device (10) is attached to the animal, at least one audio recording is acquired using the microphone (14), which is connected to a sound recorder. The audio recording can have varying durations, with a minimum duration of four minutes being optimal.

[0070] In one configuration, the sound recorder has internal memory for storing recordings. The recorder may also have external storage, such as an SD or micro SD card, for later transfer of recordings to a processing device (20). Optionally, the sound recorder can transmit at least one audio recording to the processing device (20) via wired or wireless connection. In this way, the processing device (20) receives at least one audio recording from the sound recorder via wired, wireless, or external storage.

[0071] The processing device (20) can be, but is not limited to, a computer, a smartphone, a server and / or a small computer (e.g., a Raspberry Pi) or a laptop computer, among others.

[0072] Regardless of the type of connection between the sound recorder and the processing device (20), the audio recordings can be stored in a database (30) or in a cloud.

[0073] The audio recordings are analyzed by the processing device (20), and through an artificial intelligence system it provides a diagnosis of the equine's intestinal motility and the possible existence of colic.

[0074] In a first stage, the artificial intelligence system of the processing device (20) receives an audio recording and identifies and labels the sounds corresponding to the borborygmi and discharges.

[0075] In a second stage, the artificial intelligence system of the processing device (20) performs an analysis of the labels corresponding to the borborygmi and discharges, and determines and labels the various blind cycles of the audio recording.

[0076] In a third stage, the artificial intelligence system of the processing device (20), based on the cecal cycles, determines and classifies the state of intestinal motility of a horse, and based on this result delivers a diagnosis.

[0077] The artificial intelligence system of the invention integrates different machine learning methods, specifically a neural network, a support vector machine, and a decision tree to issue the diagnosis.

[0078] The artificial intelligence system of the processing device (20) is based on a pre-trained convolutional neural network known as PANNs (Pretrained Audio Neural Networks). This network is pre-trained with large audio datasets, allowing it to be adapted to different types of audio applications.

[0079] The application of pre-trained neural networks has the advantage of its versatility, since they can be adapted to different tasks and contexts, and being already pre-trained they do not require the same level of training as a new neural network, generating a reduction in training time and resources.

[0080] The neural network of the artificial intelligence system comprising the processing device (20) corresponds to a fourteen-layer convolutional neural network.

[0081] To detect and label the sounds or events corresponding to borborygmus and discharge, the aforementioned neural network underwent fine-tuning. For this purpose, a large number of audio recordings were acquired from various healthy horses. These audio recordings were successfully made using the capture device (10) of the invention.

[0082] The neural network was trained using supervised learning, where the sounds or events of borborygmi and discharge were labeled in each of the audio recordings, so that the neural network could later predict the labels of these events in new audio recordings.

[0083] The audio recordings were separated into two sets or samples, corresponding to training audio and test audio. From these recordings, 293 borborygmi events and 190 discharge events were labeled, of which 70% were used for training and the remainder for testing. Each of the training audio tracks was segmented, and, as previously mentioned, the events corresponding to borborygmi and discharges, as well as the blind cycles, were identified and labeled. To ensure proper training of the artificial intelligence system and the neural network, each of these labels was validated by a veterinary specialist.

[0084] In the sample of training audios, it was found that the average duration of the borborygmus is 3.5 seconds, and the average duration of the sound emitted in the discharge is 3.6 seconds.

[0085] Preprocessing was performed to clean, prepare, and transform the training and test audio files into a format suitable for analysis by the neural network. Audio cleaning consisted of removing noise from the sample using a signal averaging window, eliminating noise below a desired threshold. These audio files were then represented in log-Mel spectrograms and waveforms, which served as the inputs for the neural network.

[0086] Log-Mel spectrograms are an audio representation that combines the characteristics of spectrograms with the Mel scale. Spectrograms can be described as visual representations of frequency and intensity over time. Applying a logarithmic scale or transformation to spectrograms highlights important features and minimizes less relevant ones, thus providing a format more suitable for processing by neural networks. The Mel scale, in turn, aims to better align sounds with human perception.

[0087] On the other hand, audio waves represent the direct and / or raw representation of sound, as they correspond to the waveform that shows the variations in sound amplitude over time. The waveform representation of audio is used for training neural networks, especially those involving time sequences.

[0088] In this way, the preprocessing of the training audios allows them to be converted into a graphic format (histograms or two-dimensional arrays), so that they can be efficiently processed by neural networks, for example, convolutional neural networks, since they take advantage of the techniques developed in the field of image processing.

[0089] The neural network was trained using four RTX5000 graphics cards, reaching a total of 100,000 iterations. This training was successful, using the audio files for training in a graphical format (histograms or two-dimensional arrays), along with their borborygm and discharge labels.

[0090] The processing device (20) also performs the aforementioned preprocessing of the input audios, which are analyzed in the indicated graphical representations by the neural network, which detects and labels the borborygmus and discharge events in an automatic manner.

[0091] Notwithstanding the foregoing, the results or predictions provided by the neural network can be subjected to a further refinement process, which allows for a higher quality identification of borborygmus and discharge events.

[0092] Specifically, a moving average window is used to smooth the output values ​​of the neural network. This technique involves taking a window or set of consecutive values ​​and calculating the average of the prediction. The window is then shifted over time, updating the average. This smooths the values ​​delivered by the neural network, filtering out abrupt changes in the audio recordings and / or errors in the neural network's predictions.

[0093] The moving average window also eliminates outliers, or atypical values—that is, data that differs significantly from the rest of the data. In other words, it removes data from recordings that are not aligned with the overall trend of the data.

[0094] The results provided by the moving average window process are also transformed into a binary value, corresponding to whether a sound is identified—either a borborygmus or a discharge—or not. The neural network also identifies and labels borborygmus and partial discharge events, which, although not indicative of the beginning or end of a cycle, can have significant clinical value for a veterinary specialist.

[0095] In this way, the neural network provides the various labels for borborygmi and discharge events, and also the presence or absence of sounds.

[0096] In order to determine the possible blind cycles, the processing device (20) performs an iteration on the labels of the borborygmi and discharges identified by the neural network, and generates different combinations between pairs of borborygmi and discharges, proposing candidates for the blind cycles.

[0097] For the identification of blind cycles, the artificial intelligence system of the processing device (20) uses a support vector machine, which evaluates the designated candidates of blind cycles.

[0098] The support vector machine was trained with “cycle” labels, which comprise a set of borborygmi and discharges that effectively form a cycle and are indicative of the start and end of a cycle, and with “non-cycle” labels, which comprise a set of borborygmi and discharges that do not belong to a cycle or are not indicative of the start and end of a cycle. Both labels, “cycle” and “non-cycle”, indicate the start and duration times of these events.

[0099] Since it is not a deep learning algorithm, the support vector machine did not require a large amount of data for training. It used 33 cycle labels and 30 non-cycle labels, with 80% used for training and the remaining 20% ​​for testing.

[0100] The support vector machine classifies whether the blind loop candidates correspond to a "loop" or a "non-loop." In this way, the support vector machine provides the blind loops identified in an audio recording for further analysis.

[0101] The artificial intelligence system of the processing device (20) uses the predictions provided by the neural network and the support vector machine, and classifies the intestinal motility state using a decision tree. If no sounds are detected, the artificial intelligence system of the processing device (20) classifies it as atony.

[0102] Conversely, if the artificial intelligence system of the processing device (20) identifies a sound, it evaluates whether it detects or corresponds to a blindness cycle. If at least one blindness cycle is not detected, the artificial intelligence system of the processing device (20) classifies it as atonia.

[0103] In the event that the artificial intelligence system of the processing device (20) detects the presence of sound and at least one cycle of the blind person, it subsequently identifies the number of cycles of the blind person in a four-minute time window.

[0104] If it identifies fewer than two cycles in four minutes, the processing device's artificial intelligence system (20) classifies it as hypomotile. If it identifies between two and four cycles in four minutes, the processing device's artificial intelligence system (20) classifies it as normomotile. Finally, if it identifies more than four cycles in four minutes, the processing device's artificial intelligence system classifies it as hypermotile.

[0105] The selection of the time window was based primarily on empirical evidence and scientific literature. The four-minute time window allows the system to identify multiple cecal cycles, ensuring that this identification is representative of the horse's health.

[0106] The classification of atony, normomotile, hypomotile, and hypermotile by the processing device (20) corresponds to the diagnoses that the invention can provide. Figure 9 shows the flowchart for the diagnostic classification of the present invention.

[0107] A normomotile diagnosis is considered desirable or normal. However, diagnoses of atony, hypomotility, and hypermotility are considered undesirable, and the presence of colic or acute abdominal syndrome is highly probable.

[0108] The test audios were used to validate the artificial intelligence system model, and a 72.2% accuracy was obtained in predicting the labels corresponding to the borborygmus and discharge sounds, a 74.1% accuracy in predicting the cecum cycles, and a 93.3% accuracy in predicting the diagnoses.

[0109] The diagnosis, like all data and information obtained from the capture and analysis, can be stored in a database (30) or cloud, allowing for later consultation. In the case of a negative diagnosis, the information can also be automatically sent to the animal's caretaker or owner and / or veterinarian.

[0110] The system of the invention also includes a display device (40), which can be a computer, a smartphone, a tablet, a laptop, etc.

[0111] The display device (40) is in communication, for example, wired or wirelessly, to the database (30) or cloud, to access the information and diagnosis issued.

[0112] The display device (40), through a user interface, shows the relevant system information and the issued diagnosis.

[0113] Specifically, the display device (40) can show a graphical representation of an audio recording obtained from the capture device (10), and the analysis performed by the processing device (20), by means of a sound wave representation.

[0114] The display device (40) also allows the visualization of the labels and duration of the sounds associated with borborygmi and discharges in the different sections of the analyzed audio recording. Similarly, the various cycles of the blind in the analyzed audio recording are displayed in the different sections of the audio.

[0115] Figure 10 shows a screenshot of an analyzed audio recording labeled with borborygmus, discharge, and cycle. The figure also shows that the artificial intelligence system identified and labeled the presence of a borborygmus and a partial discharge within the cycle; that is, they are not indicative of the beginning or end of the cycle. The display device (40) can also show a box with the labels of the analyzed audio recording, a box with contextual information about the horse, and a box with the issued diagnosis.

[0116] The label chart includes information on the durations of borborygmus sounds, discharge, and blind cycles.

[0117] The contextual information box displays information about the recorded and analyzed audio, for example, the audio name, recording date, recording start time, recording duration, location, horse name, horse age, horse sex, horse status or activity at the time of recording, time since last meal, time since last training, location of the capture device (10) at the time of recording, ambient noise level, etc.

[0118] Finally, the diagnostic panel displays the diagnosis issued by the system.

[0119] The above features provide a complete view to a user or veterinarian of the diagnosis and classification of the audio recordings analyzed by the system.

[0120] In a preferred mode, the user interface of the display device (40), along with contextual information and audio recordings, is hosted on a web server, making it accessible from any web browser. This allows access to the user interface from any display device (40) that supports an internet connection.

[0121] In a preferred mode, if the processing device (20) issues a negative diagnosis, i.e., of atony, hypomotility, and hypermotility, the user interface issues a visual and / or audible alarm, alerting the user of said diagnosis.

[0122] Additionally, in the event of a negative diagnosis, the system can send an alert to at least one external device (computer, smartphone, etc.), for example, the horse's owner and the veterinarian, so they can take the appropriate action. This alert could be an email, text message, phone call, message via an app, etc.

[0123] The present invention also includes its associated method, which includes all the features and steps of the artificial intelligence system already described, as well as the features and steps of the system described. The method comprises the following main steps:

[0124] Position the capture device (10) in direct contact with the horse's abdomen, then fix the capture device (10) in the selected location using the fastening means to maintain its stability during recording.

[0125] Once the capture device (10) is positioned, the microphone (13), which is connected to a recorder, is used to make an audio recording of the horse's internal sounds.

[0126] Subsequently, and according to the characteristics described above, the audio recording is transferred to the processing device (20) where the analysis will be performed. This transfer can be done via wired connection, wirelessly, or using external storage.

[0127] Once the aforementioned transfer has been carried out, in order to identify and predict the different specific sound events, one of the audio recordings is analyzed using the artificial intelligence system of the processor device (20) of the invention.

[0128] According to the description already provided, the first part of this analysis consists of identifying and labeling, from the audio recording, the sounds or events corresponding to borborygmi and discharges using the artificial intelligence system of the invention, and specifically, using its neural network as previously described. Subsequently, the blind cycles of the audio recording are identified and labeled using the artificial intelligence system of the invention, and specifically, using the support vector machine described. As already indicated, it is also possible to use the moving average window for better sound identification. Once the events are labeled, the artificial intelligence system of the invention classifies the state of the horse's intestinal motility based on the analysis performed, in order to subsequently provide a diagnosis of the horse's health.As previously described, this classification and diagnosis can be performed using a decision tree.

[0129] Both the audio recording, with the indicated labels, and the resulting diagnosis can be stored in a database (30), allowing for later consultation, monitoring of the animal's health, or any future useful use.

[0130] Finally, the information from the analyzed recorded audio and the provided diagnosis are displayed on the display device (40), thus facilitating the interpretation of the results by the users.

[0131] The method may also include the step of attaching the capture device (10) to the clamping layer (16) using the fasteners (122) and (111), as previously described herein. This allows the insulating basin (11) to be fixed in the desired position in a firm but non-permanent manner.

[0132] Additionally, and according to the characteristics already described, the processing device (20), before the analysis, can perform the preprocessing already described of the input audios, that is, of the audio recordings.

[0133] As previously described, the method provides a diagnosis of atonia when sounds are not identified or detected, or when at least one cycle of the cecum is not identified or detected in the analyzed audio recording. The diagnosis is hypomotile when fewer than two cycles are identified or detected in four minutes, normomotile when between two and four cycles are identified or detected in four minutes, or hypermotile when more than four cycles are identified or detected in four minutes in the analyzed audio recording.

[0134] According to the description, the system method can issue an alarm, either through the user interface of the display device (40) or to external devices, in the case of delivering a negative diagnosis.

[0135] The descriptions of the system and method should not be considered as limitations of the invention, since there are various modifications, variations and alternatives that, although not explicitly described, are included within the spirit and scope of the present invention.

[0136] List of elements

[0137] 10 Capture Device

[0138] 11 Insulating basin

[0139] 111 Insulating Basin Magnets

[0140] 112 Base sponge

[0141] 12 Lid

[0142] 121 Slots

[0143] 122 Lid Magnets

[0144] 13 Bell

[0145] 131 Diaphragm

[0146] 14 Microphone

[0147] 16 Fastening layer

[0148] 20 Processor device

[0149] 30 Display device

[0150] 40 User Interface

Claims

CLAIMS 1. An automatic and early detection system for acute abdomen syndrome in horses by listening to and analyzing the peristaltic sounds of a horse's cecum, CHARACTERIZED in that it comprises: - a capture device (10), comprising a microphone (14) connected to a recorder to obtain an audio recording of the horse's abdominal sounds; - a means of attaching the capture device (10) to the horse's abdomen; - a processing device (20) with an artificial intelligence system that analyzes the audio recording and identifies the sounds or events corresponding to borborygmi and discharges, cecal cycles and classifies the state of intestinal motility of the horse to provide a diagnosis; - a database (30) to store the audio recording and the diagnosis; - a display device (40) for displaying information from the analyzed audio recording, and the diagnosis.

2. The system of claim 1, CHARACTERIZED in that the capture device (10) comprises: - an insulating basin (11 ) with manes (11 1 ) on its upper part; - a bell (13), located inside the insulating basin (1 1 ), which houses the microphone (14) and a diaphragm (131 ); - a lid (12) with handles (122) located in internal slots (121).

3. The system of claim 2, CHARACTERIZED in that the insulating basin (11) comprises a base sponge (112), with a disc shape, having a circular cutout in the center, and an inner sponge.

4. The system of claim 3, CHARACTERIZED in that the fastening means is a fastening layer (16) that fits and adapts to the horse's abdomen.

5. The system of claim 4, CHARACTERIZED in that the fastening layer is made of lycra or spandex or other similar material.

6. The system of claim 5, CHARACTERIZED in that the fastening layer comprises a Velcro and / or zipper and / or similar type closure mechanism.

7. The system of claim 6, CHARACTERIZED in that the insulating basin (11) is located between the horse's skin and under the restraint layer (16), in turn, the lid (12) is located over the restraint layer (16), such that the pins (122) on the inside of the lid (12) coincide with the pins (111) on the top of the insulating basin, firmly and non-permanently fixing the capture device (10) in a desired position.

8. The system of claim 7, CHARACTERIZED in that the sound recorder has a memory for storing audio recordings.

9. The system of claim 8, CHARACTERIZED in that the sound recorder has an external storage memory.

10. The system of claim 9, CHARACTERIZED in that the processing device (20) receives, in a wired, wireless or external memory manner, the audio recordings from the sound recorder.

1. The system of claim 12, CHARACTERIZED in that the processing device (20) can be a computer, a smartphone, a server and / or a small or portable computer.

12. The system of claim 11, CHARACTERIZED in that the artificial intelligence system of the processing device (20) comprises a network neuronal that identifies the sounds or events corresponding to borborygmi and discharges from the audio recording.

13. The system of claim 12, CHARACTERIZED in that the neural network of the artificial intelligence system is a pre-trained convolutional neural network of fourteen layers.

14. The system of claim 13, CHARACTERIZED in that the artificial intelligence system uses a moving average window for better identification of borborygmus and discharge sounds or events.

15. The system of claim 14, CHARACTERIZED in that the artificial intelligence system of the processing device (20) comprises a support vector machine that identifies and labels the blind cycles of the audio recording.

16. The system of claim 15, CHARACTERIZED in that the artificial intelligence system of the processing device (20) classifies the state of intestinal motility of the horse to provide a diagnosis by means of a decision tree.

17. The system of claim 16, CHARACTERIZED in that the artificial intelligence system of the processing device (20) provides a diagnosis of atony by not identifying or detecting sounds in the analyzed audio recording.

18. The system of claim 17, CHARACTERIZED in that the artificial intelligence system of the processing device (20) provides a diagnosis of atony by not identifying or detecting at least one blind cycle in the analyzed audio recording.

19. The system of claim 18, CHARACTERIZED in that the artificial intelligence system of the processing device (20) provides a diagnosis of hypomotility by identifying or detecting less than two cycles in four minutes.

20. The system of claim 19, CHARACTERIZED in that the artificial intelligence system of the processing device (20) provides a normomotile diagnosis by identifying or detecting between two and four cycles in four minutes.

21. The system of claim 20, CHARACTERIZED in that the artificial intelligence system of the processing device (20) provides a diagnosis of hypermobility by identifying or detecting more than four cycles in four minutes.

22. The system of claim 21, CHARACTERIZED in that the database (30) also stores the labels of the analyzed audio recording, contextual information about the recorded and analyzed audio, such as the audio name, recording date, recording start time, recording duration, location, horse name, horse age, horse sex, horse status or activity at the time of recording, time since last meal, time since last training, location of the capture device (10) at the time of recording and / or ambient noise level.

23. The system of claim 22, CHARACTERIZED in that the display device (40) can be a computer, a smartphone, a tablet and / or a laptop computer.

24. The system of claim 23, CHARACTERIZED in that the display device (40) is in wired or wireless communication to the database (30), to access the audio recording information and the provided diagnosis.

25. The system of claim 24, CHARACTERIZED in that the display device (40) by means of a user interface displays a box with the labels of the analyzed audio recording, a box of contextual information of the equine and a box of the issued diagnosis.

26. The system of claim 25, CHARACTERIZED in that the user interface is hosted on a web server.

27. The system of claim 26, CHARACTERIZED in that if the processing device (20) delivers a negative diagnosis, the user interface emits a visual and / or audible alarm, alerting the user of said diagnosis.

28. The system of claim 26 or 27, CHARACTERIZED in that if the processing device (20) delivers a negative diagnosis it emits an alarm to at least one external device.

29. The system of claim 28, CHARACTERIZED in that the alarm is an email, a text message, a telephone call or a message through an application.

30. A method for the automatic and early detection of acute abdomen syndrome in horses by listening to and analyzing the peristaltic sounds of a horse's cecum, CHARACTERIZED in that it comprises the steps of: - position a capture device (10) in contact with the abdomen of a horse; - fix the capture device (10) in the desired location using a fastening means; - make an audio recording using a microphone (13), connected to a recorder, which is inside the capture device (10); - transfer the recording to a processing device (20); - analyze, using an artificial intelligence system, the audio recording on the processing device (20); - determine and label the sounds or events corresponding to borborygmi and discharges in the audio recording using the artificial intelligence system; - determine and label the blind cycles of the audio recording using the artificial intelligence system; - to classify, using the artificial intelligence system, the state of the horse's intestinal motility and provide a diagnosis; - store the audio recording, with its labels, and the diagnosis in a database (30); - display on a display device (40) the information from the audio recording and analysis, and the diagnosis provided.

31. The method of claim 30, CHARACTERIZED in that the capture device (10) comprises an insulating basin (11) with handles (111) on its upper part and a lid (12) with handles (122) on its inner part, and the fastening means is a fastening layer (16), and wherein the step of fixing the capture device (10) comprises placing the insulating basin (11) between the horse's skin and the fastening layer (16), and the lid (12) on the fastening layer (16), so that the handles (111) and (122) coincide and fix the position of the capture device (10).

32. The method of claim 31, CHARACTERIZED in that the step of transferring the audio recording to the processing device (20) is performed by wired, wireless, or external memory.

33. The method of claim 32, CHARACTERIZED in that the step of determining and labeling the sounds or events corresponding to borborygmi and discharges of the audio recording is performed with a neural network.

34. The method of claim 33, CHARACTERIZED in that the step of determining and labeling the sounds or events corresponding to borborygmi and discharges from the audio recording also uses a moving average window for better identification of the sounds.

35. The method of claim 34, CHARACTERIZED in that the step of determining and labeling the blind cycles of the audio recording uses a support vector machine.

36. The method of claim 35, CHARACTERIZED in that the step of classifying the intestinal motility status of the horse and providing the diagnosis is performed by means of a decision tree.

37. The method of claim 36, CHARACTERIZED in that the step of providing a diagnosis delivers a diagnosis of atonia by not identifying or detecting sounds or by not identifying or detecting at least one blind cycle in the audio recording.

38. The method of claim 37, CHARACTERIZED in that the step of providing a diagnosis delivers a diagnosis of the hypomotile type by identifying or detecting less than two cycles in four minutes.

39. The method of claim 38, CHARACTERIZED in that the step of providing a diagnosis delivers a normomotile type diagnosis by identifying or detecting between two and four cycles in four minutes.

40. The method of claim 39, CHARACTERIZED in that the step of providing a diagnosis delivers a diagnosis of the hypermotile type by identifying or detecting more than four cycles in four minutes.

41. The method of claim 40, CHARACTERIZED in that if the processing device (20) delivers a negative diagnosis it emits an alarm to at least one external device or user interface.

Images