International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 10 Issue: 01 | Jan 2023
p-ISSN: 2395-0072
www.irjet.net
A Review on Covid Detection using Cross Dataset Analysis Komal Bangarwa1, Najme Zehra Naqvi2 1PG Student, Computer Science and Engineering, Indira Gandhi Delhi Technical University for Women, New Delhi,
India
2Professor, Dept. of Computer Science and Engineering, Indira Gandhi Delhi Technical University for Women, New
Delhi, India ---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - In the history of the medical field, Covid-19 is one
one who already have some inherent medical conditions like diabetes, heart disease, chronic respiratory disease or cancer were at high risk of reaching a serious condition. Many fast covid detection techniques have been used to check the presence of viruses inside the human body like the RT-PCR test[12][13]. But when cases go to a serious level and the oxygen level started to drop below a certain level (~92%), doctors started to advise CT-scan results for the patients infected by Covid. A computerized tomography scan (i.e., CT scan) provides highly precise images of internal body parts, muscles and blood arteries. This allows doctors to recognize the internal function and analyse their shape, thickness and texture. CT scans are better than X-rays because CT scan provides a set of a portion of the particular region without overlapping different body structures. That’s why CT scans are better because it gives more detailed information through images which helps in extracting the exact problem and its allocation. A numerous DL methods have evolved and are used for examining Covid-19 by using tomography scans.[24][25][26]. With time, the amount of CT-scan images begins to increase in number and that results in having a good amount of data that can be used for training models that help in better results. Models trained with one dataset should give good accuracy to similar types of datasets which is one of the main factors in determining how well our model learned. Two different datasets are used for cross-dataset analysis in which first is alluded to source dataset which is utilized for training the model & the second is the target dataset which is unseen to the model. Their sources are different but they should address the same task. Many different attributes have been seen in cross-dataset over the years. A short annotation related to its attributes [19] is explained below.
of the deadliest viruses which affected the whole world. The Covid-19 pandemic caused disastrous loss of the human race and unpredictive challenges for community health and even affected the professional world. Various covid detection approaches have been used for fast detection. One of them is medical imaging, specifically Computed Tomography plays a decisive role in examining and monitoring of infected cases. Machine Learning (ML) and Deep Learning (DL) methods also have a significant part in developing different models that can help in diagnosing. This paper will give an overview of deep neural learning approaches used for corona-virus detection by utilizing CT scans in cross-dataset analysis. Even though with ongoing research, model performance has improved with time but still there are areas to cover in cross-dataset analysis. Some limitations observed were the generalization problem, dataset bias problem and robustness which is the capability to work with difficult images that occur due to the variation in image technology. The paper also gives an overview of the methods used so far for pre-processing of medical images and transfer learning techniques. Cross-dataset analysis works on improving the model accuracy by handling the abovementioned knowledge gaps and providing a model that works more practically by handling different datasets from different sources of the same task and tests how well it adapts to the model. The result might help in understanding the approaches used to overcome the limitations and identify further opportunities in cross-dataset analysis.
Key Words: Adaption, Generalization, covid-19 dataset, Computed Tomography
1. INTRODUCTION
1) Attributes on data:
World Health Organization named corona-virus as Covid-19 which is considered as one of the deadliest contagious diseases that originated from the SARS-CoV virus [14]. The last quarter of year 2019 was the beginning phase of this novel virus which has affected the whole world in every sector either its mass-production, shipment or hospitality. Coronaviruses mostly attack respiratory organs [15]. Most people who were infected by this virus encounter mild to moderate level respiratory discomfort but many recover without requiring any especial medical procedure. However, there exist many cases where the condition was severe and needed more medical surveillance. Geriatric people and the
© 2023, IRJET
|
Impact Factor value: 7.529
|
Data accessibility: This refers to the accessibility and ampleness of both source data and target data. Balanced between data: Do the provided datasets have a balanced number of data samples? Continuous data: Is the provided data continuous or online and is it progressing with time? Feature space: Consistency of identified features in data.
ISO 9001:2008 Certified Journal
|
Page 888