VGU RESEARCH REPOSITORY
Please use this identifier to cite or link to this item:
https://epub.vgu.edu.vn/handle/dlibvgu/987
Title: | Covid-19 CT images segmentation using deep neural networks | Authors: | Quach Tri Thong | Issue Date: | 2020 | Abstract: | Since the beginning of 2020, the world has been suffering from the COVID-19 pandemic. Many countries have been in lock-down, and the second wave of the pandemic is hitting numerous regions hard. Doctors, nurses, and radiologists on the frontline are overwhelmed by the immense amount of patients. In order to save as many lives as possible, prioritizing patients with severe conditions is extremely crucial. One way of achieving that is by looking at their CT scans and identifying abnormal lung regions. This paper constructed three different UNets to automatically segment these regions, including background, ground-glass opacity, consolidation, and pleural effusion. The three UNets are Three-level UNet, Five-level UNet, and UNet with EfficientNetB0 contracting path. Additionally, due to the limited amount of available data, we use augmentation to generate more data. Moreover, we also post-process the prediction to further improve the model’s performance. The best model has an F1 score of 0.63 and can potentially be utilized in supporting radiologists in segmenting lung CT scans for not only COVID-19 but also other lung diseases. Definitive diagnosis of COVID-19 requires an RT-PCR test, so the model only aims to assess the patients’ severity, not to classify COVID-19 positive patients |
URI(1): | http://epub.vgu.edu.vn/handle/dlibvgu/987 |
Appears in Collections: | Computer Science (CS) Computer Science (CS) |
Files in This Item:
File | Description | Size | Format | Existing users please Login |
---|---|---|---|---|
Covid-19 CT images segmentation using deep neural networks.pdf | 3.15 MB | Adobe PDF |
Page view(s)
112
checked on Apr 26, 2024
Download(s)
60
checked on Apr 26, 2024
Google ScholarTM
Check
Items are protected by © Copyright of Vietnamese - German University Library