Europe Union

AI modeling and machine learning in detecting OA disease among dogs

AI modeling, machine learning and deep neural network in detecting OA disease among dogs
The idea was to create machine learning algorithms for a high-tech sensor, implemented into a collar, capable of detecting dog's Osteoarthritis without the need of spending time and money on the veterinarian's human expertise or the usual medical imaging modality like canine thoracic radiographs, magnetic resonance imaging or ultrasound images.

Successful implementation of the idea could also improve veterinary medicine, as it could be widely used in clinical practice. And most importantly, to get the best possible sensitivity to correctly diagnose all the affected dogs. Osteoarthritis is the most common form of arthritis in dogs and is a degenerative joint disorder that leads to a permanent deterioration of the cartilage that provides cushioning for the joints. This painful condition is estimated to affect about a quarter of all dogs throughout their lifetime. Thus we wanted to employ computer science, AI models and and computer-aided detection to recognise the patterns and enable more accurate diagnosis in predictive analytics.

What are deep neural networks and deep learning model?

Deep learning systems are a part of machine learning methods, more specifically, a broader family based on artificial neural networks with representation learning. There are supervised, semi-supervised or unsupervised machine learning models.

When we consider what are deep neural networks in deep learning algorithms, the simple explanation would be that it is a neural network with more complexity, i.e. more than two layers. They employ sophisticated mathematical modelling to process data in complex ways.

We distinguish three main types of deep neural networks in machine learning:

  • Convolutional Neural Network (CNN)
  • Recurrent Neural Network (RNN)
  • Artificial Neural Network (ANN)
  • Deep neural network can mean any learning neural network with multiple hidden layers. Deep convolutional neural networks are especially useful in image recognition.

    Recurrent neural networks create a circular connection to allow some nodes to affect the inputs of other nodes in the network.

    Artificial deep neural network models can aid in classification, regression problems, and sentiment analysiss and are designed to replicate the behavours observed in a human brain or an animal one. They allow the machine learning to process increasingly complex data via a simplified reflexion of the animal or human brain to mimic the behaviour of a nervous system and neurons which can aid in fields like medical imaging (e.g. magnetic resonance images).

    How to distinguish OA disease among dogs using AI?

    How to distinguish OA disease among dogs using AI

    Our task was to conduct data analysis based on the provided data sets and input data, with the help of computing resources, artificial intelligence and machine learning model, with measurements of dog movements and to distinguish between dogs with and without OA based on the collected sensor’s raw data.

    We’ve decided to employ deep learning models and deep convolutional neural network for imaging data to recognize patterns based on collected raw data.

    We wanted to prove the universality of solutions of the deep learning model for different breeds. At first, we relied on the previously collected training data provided to us by data scientists.

    The problem emerged at the level of data collection in the data sets provided, as it was improperly annotated. There was also an issue with logistic regression model and overall technical trouble with computer-aided detection and data collection from the gyroscope. Our solution involved the development of the right processing artificial intelligence algorithms and AI models of the data from accelerometers and gyroscopes, then creating classification algorithms that operate on the collected data sets, as well as adjusting the software/hardware interfacing to obtain stable results.

    Use of deep machine learning models and deep learning algorithms to detect OA in dogs.

    Based on the training data provided by data scientists, we concluded that due to the importance of high precision of the deep machine learning and digital imaging, we started by monitoring dogs’ activity over a long period. After analysing different data points and input and output variables, our aim was to introduce significant improvements to the solutions regarding logistic regression model, machine learning models and AI models. We were looking for a set of discriminative features and data points which would be crucial for determining dogs’ illnesses. Based on the results, using analytical methods, we were able to determine whether the dog had the OA.

    Our team proposed a deep neural network machine learning algorithm to increase the level of classified OA severity, as well as the diagnostic accuracy of the device and AI models. The results of our efforts allowed us to sort through all the data and extract the deep features. We were able to implement artificial intelligence solutions and deep neural networks to distinguish between dogs with OA symptoms. We proposed high-quality deep learning model and other deep-learning technologies to generate the results that could work with more data in the future.

    Summary

    We are extremely pleased with the results of our work on the deep learning and data analysis, as we were able to extract all the necessary data and implement further measures. We are positively impressed how the use of different computing resources, artificial neural networks and machine learning techniques can impact different aspects of life aid human intelligence, as well as data scientists. Every new project shows us how the tools we use, such as AI, machine learning or deep neural network in computer science, allow us to support and help in everyone’s daily life and improving human expertise on daily basis.

    We invite you to follow our further developments and check our recent projects.

    Tech stack:

    • Inertial measurement unit
    • Deep neural network
    • AI modelling