Classification Of COVID-19 Disease By Using Explainable Artificial Intelligence With X-ray Images
The eruption of COVID-19 caused in excess of 100,000 deaths so far in the USA. It is important to direct initial screening of patients with the side effects of COVID-19 to control the spread of this virus. In any case, it is getting difficult to conduct the tests with limited testing units because o
2025-06-28 16:30:48 - Adil Khan
Classification Of COVID-19 Disease By Using Explainable Artificial Intelligence With X-ray Images
Project Area of Specialization Artificial IntelligenceProject SummaryThe eruption of COVID-19 caused in excess of 100,000 deaths so far in the USA. It is important to direct initial screening of patients with the side effects of COVID-19 to control the spread of this virus. In any case, it is getting difficult to conduct the tests with limited testing units because of the developing number of patients. A few investigations proposed chest X-beam pictures are very helpful in detecting this disease. Accordingly, it is fundamental to utilize each accessible asset, chest X-beam to lead an enormous number of tests at the same time. Accordingly, this investigation plans to build up a learning-based model that can identify Coronavirus patients with better precision on chest X-beam picture dataset. In this work, two distinctive deep learning approaches such as Darknet-53 and Mobilenet-v2 have been implemented on given dataset. By using these approaches or techniques we will train the model by providing the dataset of COVID-19 affected and non-affected X-beam images which can predict the results by using Explainable Artificial Intelligence. It can classify each and everything about this disease in the output.
Project ObjectivesTo be proposed an explainable AI approach in which we will assign more than one information for each image during the training process. Based on this step, it is a high chance of improved accuracy for the correct classification. And we will proposed a feature selection approach to select the best features for final classification. Also we will publish a search paper.
Project Implementation MethodWe are using Matlab for the proposed project implementation. We acquired dataset from the Kaggle website .the dataset is the combination of multiple classes. (For example, 1000 chest X-ray images and 1000 Normal patients images were collected from kaggle source), then we trained our dataset on two different deep learning techniques (Darknet-53 and Mobilenet-v2) and then applied explainable AI on the results with the help of LIME ( Local Interpretable Model-Agnostic Explanations).
Benefits of the Project- As we know the world is suffering from swear pandemic situation.
- Beside health issues many regions around the world badly effected economically. Due to these reasons mostly people are not able to bear the expenses of COVID 19 test.
- So our first motivation for this project is to propose a system that can be cost efficient and affordable by common person. Because in this project we use AI which use X-Ray image comparison with the data set we used to train our system and generate results which is very cheap process.
- Moreover in these circumstances we need test result on urgent bases so that experts made some conclusion. But remote techniques like physical sample collecting techniques requires a lot of time. But as our system is X-Ray comparison based so it provide outputs result within no time.
- On the other hand tube testing techniques involves physical interaction between patient and physician, this system do not require patient present on spot it just his/her X-rays that can be read by our system for result.
- We are using Explainable AI instead of simple AI because we are targeting the well-defined results for users. The user can trust on the system without any confusion with the help of Explainable AI. The Explainable AI always justify the results to the users.
The final details will include:
- A fully trained model to classify the COVID-19 and Normal chest x-ray images with accuracies of 95% and above.
- A prototype for the model.
- FYP documentation including (6-8) chapters and complete details related to the project.
- Publish Research paper on it.
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 69700 | |||
| NVIDIA Tesla GPU Computing Processor Graphic Cards 900-22081-2250- | Equipment | 1 | 59700 | 59700 |
| Documentation printing | Miscellaneous | 1 | 5000 | 5000 |
| Research paper publish | Miscellaneous | 1 | 5000 | 5000 |