Method for diagnosing patellar luxation of companion animal on basis of artificial intelligence

Abstract

The present invention relates to patellar luxation diagnosis of a companion animal and, more specifically, to a diagnosis method by which a data set may be constructed by preprocessing a six-axis signal read by means of a harness-type sensor unit capable of being attached to the back of a companion animal, and patellar luxation of the companion animal may be accurately checked in real time through learning using an artificial intelligence model on the basis of the constructed data set.

Description

AI-based method for diagnosing patellar luxation in pets

[0001] The present invention relates to the diagnosis of patellar luxation in companion animals, and more particularly, to a diagnostic method that can accurately check patellar luxation in companion animals in real time using an artificial intelligence model.

[0002] In companion animals like dogs and cats, the patella is a bone located near the kneecap. Patellar luxation occurs when the patella is misaligned over the trochlear groove, protruding inward or outward.

[0003] Patellar luxation is often caused by impact, such as an accident or a fall from a high place. It is common in breeds of dogs with congenital abnormalities in their joints that make them prone to dislocation. It is especially common in small dogs such as Maltese, Pomeranian, and Poodles compared to large dogs. Statistically, patellar luxation is a joint disease that affects approximately 70% of companion dogs in Korea.

[0004] Patellar luxation often goes undetected by pet owners until symptoms become severe. Therefore, accurate diagnosis and prevention are crucial, as patellar luxation can affect a pet's movement, causing irregular hind leg movements and pain or discomfort.

[0005] Various studies have been conducted to prevent patellar luxation in companion animals.

[0006] For example, Korean Patent Publication No. 10-2023-0080828 (published on June 7, 2023) proposes a device and method for diagnosing patellar luxation in companion animals. This method offers the advantage of being able to relatively easily determine the degree of patellar luxation by measuring the angle connecting multiple feature points based on a rear-facing image of the companion animal without the use of a separate, expensive device.

[0007] As a similar technology, Korean Patent Registration No. 10-2255483 (publication date: May 24, 2021) discloses a method for diagnosing joint disease using an artificial intelligence model using an image of a companion animal, and a device using the same.

[0008] In addition, Patent Registration No. 10-2381906 (publication date: April 1, 2022) proposes a sensor-based pet ring for pet management and a total pet care system utilizing the same. This pet ring is useful as a device for managing pets, as it detects the pet's location, posture, etc. through its sensors and transmits information about the pet's unexpected behavior, such as aggression or abnormal behavior, or health status, to a user terminal or veterinary hospital, prompting preemptive action for the pet.

[0009] However, these technologies still fall short of accurately diagnosing and preventing patellar luxation in pets in real time.

[0010] The present invention has been proposed to improve the above-mentioned shortcomings, and its purpose is to provide a method for measuring and accurately diagnosing patellar dislocation in a companion animal in real time using an artificial intelligence model through a harness-type sensor device.

[0011] The present invention for achieving the above object relates to an AI-based method for diagnosing patellar dislocation in a companion animal, comprising the steps of: measuring a signal according to the movement of the companion animal through a harness-type sensor unit that can be detached from the companion animal; constructing an artificial intelligence learning dataset from the measured signal; extracting a feature dataset representing the movement of the joints of the companion animal from the constructed dataset; classifying the condition of the companion animal's patellar bone based on the extracted feature dataset and the companion animal's body information; and determining whether the companion animal has a patellar abnormality based on the classified patellar bone status and indicating the risk of patellar dislocation.

[0012] In the AI-based patellar dislocation diagnosis method of a companion animal of the present invention, it is preferable that the harness-type sensor unit is an MPU (Micro Processor Unit) including a gyro sensor that can be attached to the harness.

[0013] In the AI-based patellar dislocation diagnosis method of a companion animal of the present invention, it is preferable that the signal is an angle, velocity, and acceleration time series signal for six axes including the x-axis, y-axis, z-axis, and the rotation axis of each axis.

[0014] In the AI-based method for diagnosing patellar dislocation in a companion animal of the present invention, it is preferable that the step of constructing the AI ​​learning dataset includes filtering the signal measured by the harness-type sensor unit, normalizing the filtered dataset to convert different scale values ​​into a certain range, labeling the converted dataset, and then inspecting it.

[0015] In the AI-based method for diagnosing patellar dislocation in a companion animal of the present invention, it is preferable that the step of extracting the feature dataset learns the time series dataset through deep learning and extracts the abnormal dataset by comparing it with a normal time series dataset.

[0016] In the AI-based method for diagnosing patellar dislocation in a companion animal of the present invention, it is preferable that the step of classifying the joint condition of the companion animal classifies the patellar condition of the companion animal by learning the feature dataset extracted through deep learning and the body information of the companion animal.

[0017] In the AI-based method for diagnosing patellar dislocation in a companion animal of the present invention, it is preferable to inform the risk of patellar dislocation in the companion animal by classifying it as good when the companion animal's body tilt is between 0 and 3 degrees, as a warning when it is over 3 degrees and within 8 degrees, and as dangerous when it is over 8 degrees.

[0018] According to the AI-based patellar dislocation diagnosis method of the present invention, patellar dislocation of a companion animal can be measured in real time and accurately diagnosed using an artificial intelligence model.

[0019] Figure 1 is a flow chart illustrating the process of diagnosing patellar dislocation in a companion animal of the present invention.

[0020] Figure 2 is a schematic diagram of a device for performing the diagnostic process of Figure 1.

[0021] Figure 3 is a drawing showing a harness-type sensor unit worn on a companion dog.

[0022] The embodiments of the present invention are provided to more clearly explain the present invention to those skilled in the art. The following embodiments may be modified in various ways, and the scope of the present invention is not limited to these embodiments. Rather, the embodiments are provided to further clarify and faithfully explain the present disclosure and to clearly convey the technical concepts of the present invention to those skilled in the art.

[0023] Hereinafter, a specific embodiment of a process for diagnosing patellar dislocation in a companion animal according to the present invention will be described in detail with reference to the attached drawings.

[0024] Although the present invention can be applied to popular pets such as dogs and cats, this embodiment will be described mainly with dogs being the most commonly kept pet.

[0025] Fig. 1 is a flow chart for explaining a process for diagnosing patellar dislocation in a companion animal of the present invention, Fig. 2 is a diagram of the overall configuration of a patellar dislocation diagnosis device in a companion animal of the present invention, and Fig. 3 shows a state in which a harness-type sensor unit is worn on a companion dog.

[0026] To measure a dog's movements, sensors can be attached to various parts of the dog, such as the neck, legs, or back. However, attaching a sensor to the dog's neck using a pet ring, as described in Patent Document 3, has the disadvantage of not being able to detect the entire movement radius of the dog compared to the movement of the entire body.

[0027] Attaching sensors to a dog's hind legs can accurately capture the characteristics of patellar luxation because it directly measures the movement of the dog's hind legs. For example, dogs with patellar luxation may exhibit changes in the angle or speed of their hind legs. Attaching sensors to the leg can accurately measure these changes, making it effective in detecting patellar luxation. However, attaching sensors to the hind legs can restrict the dog's movement, making it difficult to measure overall hind leg movement and potentially hindering the detection of other conditions.

[0028] Attaching the sensor to a dog's back has the advantage of causing less discomfort to the dog than attaching it to the hind legs, and allows for comprehensive measurement of the dog's movements. However, if the sensor is attached to the dog's back and moves independently of the dog's movements or interferes with them, accurate data collection can be difficult. Therefore, this issue needs to be addressed.

[0029] As such a supplementary measure, the present invention measures signals according to the movement of the companion dog through a harness-type sensor unit that can be detached from the companion dog's back.

[0030] The harness mentioned in this specification is a chest strap worn around the dog's chest instead of a collar. It's widely used because it reduces strain on the dog's neck and allows the owner to easily pull the leash. For older dogs or those with physical disabilities who have difficulty walking, a harness supports the dog's body, helping them walk freely.

[0031] If a dog's hind leg movements are irregular or asymmetrical, changes in body movement may also occur. Measuring these changes can help detect patellar luxation. Attaching the harness-type sensor unit according to the present invention to the dog's back allows the sensor to be held more closely to the dog's body, effectively measuring body movement in response to hind leg movement.

[0032] Movement data for companion dogs can be collected in various ways. The harness-type sensor unit (10) is preferably an IoT device including an acceleration sensor, a posture sensor, or a gyro sensor, and a gyro sensor is more preferred. While the acceleration sensor can detect the degree of progress in the x, y, and z directions, the gyro sensor can measure the amount of change in rotation, i.e., angular velocity. The harness-type sensor unit is more preferably an MPU (Micro Processor Unit) including a gyro sensor that can be detached from the harness. In the present embodiment, the nRF52840 Bluetooth communication module from Nordic and the MPU Nano 33 IoT from Arduino were used as the harness-type sensor unit, but the present invention is not limited thereto. Of course, multiple sensor units can be used to more precisely analyze the posture imbalance or behavioral pattern of the companion dog using the harness-type sensor unit of the present invention.

[0033] It is preferable that the signal obtained through the above harness-type sensor unit is a six-axis angle, velocity, and acceleration time series signal including three axes of the x-axis, y-axis, and z-axis and a rotation axis based on each axis. In this case, the format of the dataset can be taken as (timestamp, x, y, z), such as the time at which the data was collected (timestamp) and the angle, velocity, and acceleration values ​​of each axis (x, y, z) measured by the gyro sensor.

[0034] Referring to FIGS. 1 and 2, the signal measured by the harness-type sensor unit (10) may contain noise or distortion, and thus the control unit of the MPU performs filtering, interpolation, normalization, etc. processes to remove this noise or distortion, and then stores the signal in the storage unit. For example, the 6-axis data measured by the gyro sensor can be converted into a Kalman filter to distinguish normal as (-) and caution as (+), and the data can be preprocessed through SVM (Support Vector Machine) model learning.

[0035] In this way, the signal measured from the sensor unit is filtered, the filtered dataset is normalized to convert different scale values ​​into a certain range, and the converted dataset with labeling is transmitted to a cloud server, computer terminal, or mobile phone terminal through the communication unit.

[0036] Because the quality of datasets used in data processing, analysis, and AI learning determines learning outcomes, building and securing high-quality datasets is crucial. In particular, high-quality data labeling, which provides computers with labeled, correct data, is crucial for AI learning.

[0037] The 3D sensor signals measured as above are filtered and labeled as a dataset suitable for an artificial intelligence model, and then a feature dataset representing the movement of the dog's patella is extracted from the dataset.

[0038] AI can learn a dog's movement patterns and identify abnormalities that humans find difficult to detect, potentially preventing patellar luxation. For example, algorithms utilizing machine learning can be developed to detect and learn normal dog gait patterns and abnormal behavior patterns associated with patellar luxation.

[0039] To this end, a feature dataset representing the movement of the patella of the companion dog is extracted from the dataset processed above.

[0040] Monitoring the joint status in real time can help quickly identify changes in status according to the dog's movements. However, in the present invention, the process of extracting a feature dataset can be performed by various independent algorithms or can be performed integrally by an artificial neural network that performs learning.

[0041] Next, the condition of the pet's patella is classified based on the extracted feature dataset and the pet's physical information.

[0042] Patellar dislocation can be divided into stages 1 to 4 depending on the degree and pattern of progression of the dislocation.

[0043] Grade 1 patellar dislocation is a condition in which dislocation does not occur during normal movement, but only occurs when the patella is pushed inward or outward, and the patella returns to its normal position when there is no pushing force.

[0044] Grade 2 patellar dislocation occurs when the patella is artificially pushed, and intermittent dislocation also occurs when the patella is flexed and extended. In Grade 2 patellar dislocation, the patella can be returned to its normal position by moving it back to its normal position or by re-moving the patella.

[0045] In stage 3 patellar dislocation, the patella remains mostly dislocated, and the knee joint can be straightened to artificially return the patella to its normal position, but if the knee joint is moved again, re-dislocation will occur.

[0046] Grade 4 patellar dislocation means that the patella is permanently dislocated and cannot be moved back to its normal position.

[0047] Most medial patellar dislocations are caused by musculoskeletal abnormalities, such as medial displacement of the quadriceps femoris, external twisting and bending of the femur, abnormal formation of the femoral head, rotational instability of the knee joint, and deformities of the tibia.

[0048] Based on feature analysis of the above dataset, a prediction model for patellar luxation in dogs can be trained. Various artificial intelligence algorithms, including machine learning algorithms, various artificial neural networks, and deep learning algorithms such as generative adversarial networks (GANs) and convolutional neural networks (CNNs), can be used to train the prediction model, either supervised or unsupervised.

[0049] Based on the above classification joint status, it determines whether your pet has a patellar abnormality and informs you of the risk of patellar dislocation.

[0050] In fact, when a dog lifts its legs, the left and right tilt of its body changes repeatedly, as it lifts its right leg and then its left leg. This can be detected by the gyro sensor attached to the harness-type sensor part on the dog's back, which can be used to check for patellar dislocation.

[0051] For example, in the present invention, as one embodiment, the risk of patellar dislocation in the companion animal may be determined by classifying the case in which the body tilt of the companion animal is between 0 and 3 degrees as good, a case in which the body tilt is more than 3 degrees and less than 8 degrees repeatedly at least once as a warning, and a case in which the body tilt is more than 8 degrees repeatedly at least three times as a danger.

[0052] Meanwhile, it is necessary to design an architecture for safely storing information about the harness-type IoT device according to the present invention on a cloud server and utilizing it effectively. The cloud server should be built on a stable and scalable cloud platform. It may also include a database server to store the sensor dataset or a web server to handle user-device interactions.

[0053] In this embodiment, the sensor data is time-series data that changes over time, so it is recommended to use a time-series database such as InfluxDB or MongoDB for storage, and to use object storage to safely store information such as user information and device configuration.

[0054] For reference, you can also configure an OAuth or JWT-based authentication system to effectively perform user and device authentication and authorization management.

[0055] Additionally, cloud servers provide APIs for integration with external applications and services, and can design appropriate API endpoints to access user data and device status.

[0056] You can introduce a system to process and analyze device data in real time, or you can utilize tools such as Apache Kafka and Apache Flink that support streaming data processing.

[0057] Additionally, it is necessary to strengthen the security of data transmission and storage, implement policies to comply with privacy regulations such as GDPR, and protect user privacy through security features such as data encryption and access control.

[0058] Furthermore, it would be beneficial to provide users with a dashboard that allows them to monitor their dog's activity in real time via the web or app. It would also be desirable to establish a notification service to effectively notify users of unusual behavior or important events. Furthermore, it would be necessary to implement a mechanism for periodically updating the software and firmware of devices and cloud servers to ensure continuous security and functional improvements.

[0059] This architecture provides a robust and scalable IoT system that helps users effectively manage their pets' health.

[0060] Although the present invention has been described above with reference to one embodiment, it will be understood by those skilled in the art that various modifications and changes can be made to the present invention without departing from the spirit and scope of the present invention as set forth in the claims below.

Claims

1. A method for diagnosing patellar dislocation in a companion animal, A step for measuring signals according to the pet's movements through a detachable harness-type sensor part for the pet. A step of constructing a dataset for artificial intelligence learning from the above measured signals, A step of extracting a feature dataset representing the movement of the joints of the companion animal from the constructed dataset; A step of classifying the patellar condition of the pet based on the extracted feature dataset and the pet's body information; and An AI-based patellar luxation diagnosis method in a companion animal, characterized by including a step of determining whether the companion animal has a patellar abnormality based on the classified patellar condition and informing the companion animal of the risk of patellar luxation.

2. In paragraph 1, An AI-based method for diagnosing patellar dislocation in a companion animal, characterized in that the harness-type sensor unit is an MPU (Micro Processor Unit) including a gyro sensor that can be attached to the harness.

3. In paragraph 1, An AI-based method for diagnosing patellar dislocation in a companion animal, wherein the above signals are time series signals of angle, velocity, and acceleration for six axes including the x-axis, y-axis, z-axis, and the rotation axis of each axis.

4. In paragraph 1, The step of constructing the above artificial intelligence learning dataset is characterized by filtering the signal measured from the harness-type sensor unit, normalizing the filtered dataset to convert different scale values ​​into a certain range, labeling the converted dataset, and then inspecting it. An AI-based method for diagnosing patellar dislocation in a companion animal.

5. In paragraph 1, The step of extracting the above feature dataset is characterized by learning the time series dataset through deep learning, comparing it with a normal time series dataset, and extracting the abnormal dataset. An AI-based method for diagnosing patellar dislocation in a companion animal.

6. In paragraph 1, The step of classifying the patellar condition of the companion animal is an AI-based method for diagnosing patellar dislocation of a companion animal, characterized in that the step of classifying the patellar condition of the companion animal comprises learning the feature dataset extracted through deep learning and the body information of the companion animal to classify the patellar condition of the companion animal.

7. In paragraph 1, The risk of patellar dislocation in the companion animal is determined by an AI-based diagnosis method for patellar dislocation in the companion animal, characterized in that it is categorized as good when the companion animal's body tilt is between 0 and 3 degrees, as a warning when it is over 3 degrees and within 8 degrees, and as dangerous when it is over 8 degrees.

Images