Improving The Accuracy of Osteoarthritis Classification Using Deep Learning

Project Title

Improving The Accuracy of Osteoarthritis Classification Using Deep Learning

Project Area of Specialization Artificial IntelligenceProject Summary

In this project we will train the Convolutional Neural Network (CNN) for Multi Osteoarthritis Study (MOST) dataset and then we will use this model’s weights and transfer the information for training of another dataset Korea National Health and Nutrition Examination Survey (KNHANES) so that Transfer learning could be achieved. Therefore, this technique would help us to improve the classification accuracy of the network

Project Objectives

Objective 1: Getting acquainted with Deep Neural Networks (DNN)

The first objective is to develop an understanding of basic and fundamental concepts about the artificial intelligence and its one of its major field (in essence, DNN). The project will help an individual in post graduate level as well, as one would already know the implementation of artificial intelligence in image recognition technology

Objective 2:To train the first Neural Network for MOST Dataset

The second objective will be to train the CNN for Multi Osteoarthritis Study (MOST) dataset.

Objective 3:Utilizing the weights from 1st dataset (MOST) to train KNHANES (Korea National Health and Nutrition Examination Survey)

The 3rd objective is to ensure that the feature learning of the pre trained model of the dataset (MOST) has been done and then this information (features and patterns) from the 1st dataset will be used to train another neural network for second dataset (KNHANES) and check its accuracy.

Objective 4: Training the second Neural Network

The dataset (KNHANES) will be used to train the second network so that it could first analyse its pattern and distribution sequence and then remembers it for the future if same input is given.

Project Implementation Method

In this project we will train the Convolutional Neural Network (CNN) for Multi Osteoarthritis Study (MOST) dataset and then we will use this model’s weights and transfer the information for training of another dataset Korea National Health and Nutrition Examination Survey (KNHANES) so that Transfer learning could be achieved.

All the experiments will be conducted using a rich ecosystem of tools and libraries such as PyTorch.

Benefits of the Project

• This approach will provide benefit to the health care sector such as hospital.
• Osteoarthritis (OA) is the most common diseases in the world. It has been reported that the estimated overall costs per patient for OA treatments reach 19,000 ?/year [4]. So a machine learning based classification model would help in the detection of the disease and also it will be cheaper as compared to the detection by magnetic resonance imaging (MRI).

Technical Details of Final Deliverable

Osteoarthritis is a very common disease. Common ways of OA detection include MRI which is effective but expensive and not readily available. Another method is Radiography which is quite cheap but degeneration of cartilage cannot be detected through it. Our goal is to using Machine Learning Techniques to develop a program that can detect this disease quite accurately and which is affordable for everyone.

Final Deliverable of the Project Software SystemCore Industry ITOther Industries Education , Medical Core Technology Artificial Intelligence(AI)Other Technologies Artificial Intelligence(AI)Sustainable Development Goals Good Health and Well-Being for PeopleRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	80000
MSI Geforce GTX 1660 Armor 6G OC Video Graphics Card 6GB GDDR5	Equipment	1	32000	32000
Intel Core i5-9400F Desktop Processor - LGA1151	Equipment	1	25000	25000
Gigabyte B365M GAMING HD Intel B365 Gaming Motherboard	Equipment	1	13000	13000
Printing	Miscellaneous	1	5000	5000
Stationery	Miscellaneous	1	5000	5000