Detecting covid-19 using medical imaging
Research in medical as well as IT field are expeditiously happening around the globe, it is still hard for the doctors to detect and confirm and the disease at an early stage and immediately isolate an infected patient from the healthy population. In current times all over the world the use of&
2025-06-28 16:32:01 - Adil Khan
Detecting covid-19 using medical imaging
Project Area of Specialization Artificial IntelligenceProject SummaryResearch in medical as well as IT field are expeditiously happening around the globe, it is still hard for the doctors to detect and confirm and the disease at an early stage and immediately isolate an infected patient from the healthy population. In current times all over the world the use of reverse transcriptase polymerase chain reaction (RT-PCR) is the standard. This molecular biology technique detects genetic material that is specific for the SARS-CoV-2 virus. Yet, RT-PCR is not 100% accurate, and some experts have raised questions around false-positive and false-negative test results.
In this project, we introduced a detection of Covid-19 using CT scan. The main contribution of this project is given as follows.
First, an image processing algorithm is used to detect empty spaces from a CT scan of the patient. The algorithm processes the CT scan image, extracts disease information concentrating spots, and their positions thereof. The system also reports if individual patient has lung cancer or infection or the novel coronavirus disease. Our main focus remains on the covid-19 detection. Detection information is made available to the patient. This CT scan detection of covid-19 reduces the time taken by the RT-PCR test and then the sample collection. The swab that’s supposed to be pushed into the back of the nose (often painfully) and then curve down into the throat sometimes doesn’t reach far enough, or doesn’t remain in place long enough, to collect a decent sample, and even then, the test is not fully reliable.
Project Objectives- To develop an application of deep learning-based framework, where the proposed approach will be trained using a CT image dataset that consisted of COVID19, lung cancer, and normal person CT scans. The goal was to differentiate between COVID19, lung cancer, and normal cases.
- To develop front end application, preferably a user-friendly interface to communicate patient’s disease using his chest CT scan image sequence.
- Reduce the enormous false positive rate by increasing the efficiency and accuracy of the diagnostic procedure.
- Cut down the overall cost and create a better environment.
- Lessen the duration of diagnosis and prognosis of the disease
We aim to introduce a fully automated method for detecting COVID-19 cases from the output files(images) of the lung CT scan device. Our system would not need any medical expert for system configuration and will take all the CT scans of a patient and clarifies if that patient is infected to COVID-19 or not.
In our system, at the first stage of our work, this system runs our proposed image processing algorithm to discard those CT images that inside the lung is not properly visible in them. This action helps to reduce the processing time and false detections. In this way, we speed up the process because the network does not have to analyze all the images. Also, we improve the accuracy by giving the network only the proper images(most efficient CT scan among all).
Then we will train and test three deep convolutional neural networks for classifying the selected images. One of them is enhanced convolutional model, which is designed to improve classification accuracy. We’ll train, evaluate, and compare three different deep convolutional networks: Xception, ResNet50V2, and our own model. At the final stage, we evaluate our fully automated system in two different ways. The first way is single image classification, and the second is the fully automated system that we evaluated more patient with images.
We can also investigate the infected areas of the COVID-19 classified images by segmenting the infections using a feature visualization algorithm.
We may be using existing deep learning networks, they have identified lung cancer on chest CT scan images and introduced the network with high accuracy and will be able to diagnose normal patients, lung cancer, and COVID-19, with an overall accuracy.
Benefits of the ProjectOur project aims to give more accurate and quick results as compared to RT-PCR.
Usually radiologist take much more time to analyse the given CT scans while our system aims to give result in 5-6 minutes.
Expert radiologist guidance is required for accurate and rapid COVID-19 detection using a CT scan of the chest. Timely and accurate treatment of COVID-19 disease is a challenging task for the healthcare givers. But the limited availability of conventional COVID-19 detection kits is a major issue. Thus, our automatic diagnosis model is required for COVID-19 detection using imaging modality to reduce the manual involvement in disease detection using a CT scan of chest images.
If a patient has been ill for a short duration then it is the CT scan which is of greater utility because it is more likely to detect small subpleural haze along the hidden areas of say mediastinal reflections of pleura which might be missed on an x-ray altogether. It may also detect the mediastinal lymph nodes which have a strong negative prediction against the COVID-19 disease, or detect alternate pathological diagnosis of pneumonias or pulmonary infection of other etiology. CT scan is good enough for diagnosis.
For a severely ill patient presenting at this stage, there is practically no practical advantage of either the x-rays or the ultrasound. It should, in fact, must be the CT scan to be the imaging modality of choice. CT scan objectively assesses the geographic spread of involvement and change from ground glass opacity to consolidation or fibrosis. CT scores are also developed along the scale of 0-25 based on these evolution patterns, which can further help in the stratification of the severity.
Technical Details of Final DeliverableOur final deliverable would be a software detecting covid-19 and lung cancer from CT scans using ML.
Three deep convolutional networks used are Xception, ResNet50V2 and our own model.
Xception introduced new inception modules constructed of depth-wise, separable convolution layers (depth-wise convolutional layers followed by a point-wise convolution layer) and gave best results on ImageNet dataset.ResNet50V2, is a upgraded version of ResNet50.
Feature pyramid network(FPN) was utilized in RetinaNet for enhancing object detection. FPN helps the network better learning and detecting objects at different scales that exist in an image. FPN solves this problem when there are objects with various scales in the image so FPN generates a bottom-up and a top-down feature hierarchy with lateral connections from the network’s generated features at different scales. This helps the network generate more semantic feature, increasing accuracy.
Here we use FPN for image classification as COVID-19 infections exist in different scales, so using FPN helps to extract various semantic features of the input image. Our model can detect COVID infections even when they are tiny and, more importantly, detect COVID false positives fewer because it learns better about the infection points.
The architecture of the proposed network is as follows: We will use ResNet50V2 as the backbone network and compare our model with ResNet50V2 and Xception. In FPN we will use concatenation layers instead of adding layers inside the feature pyramid network.
FPN needs to extract five final features that each one presents the input image features on a different scale. After that, we will implement dropout layers ( to avoid overfitting), followed by the first classification layers. First classification layers will be feed to the final classification layer, Relu activation function will be used here.
Finally, at end of the architecture, we will concatenate the five classified layers (each consisting of two neurons) and made a ten neurons dense layer. Then we connected this layer to the final classification layer, which applies the softmax function. With running this procedure, the network will utilize different classification results based on various scales features. As a result, the network would become able to classify the images better.
Final Deliverable of the Project Software SystemCore Industry MedicalOther IndustriesCore Technology Artificial Intelligence(AI)Other TechnologiesSustainable 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) | 65000 | |||
| Real samples of Covid-19 CT scans | Equipment | 1 | 35000 | 35000 |
| Matlab toolbox | Equipment | 1 | 30000 | 30000 |