The main contributions of our work can be summarized as follows:
• Development of a novel connectivity-promoting regularization loss function for an image segmentation framework detecting pathologic COVID-19 regions in pulmonary CT images.
• Quantitative validation showing improved performance attributable to our new TV-UNet approach compared to published state-of-the-art segmentation approaches.
we used some different datasets:
-
all available data from the COVID-19 CT segmentation dataset [1], consisting of 929 CT slices from 49 patients. Out of these, 473 CT-image slices are labeled as including COVID-19 pathologies with Ground-Glass pathology regions identified by expert tracing. The remaining 456 CT image slices are labeled as COVID-19 pathology free. CT-slice sizes were either 512×512 or 630×630. While a small subset of the 929 CT images also have regions of additional pathologies identified and labeled as Consolidation and/or Pleural Effusion, this work focuses on the Ground Glass mask.
-
One such dataset with semi-supervised COVID-19 segmentations (COVID-SemiSeg) was recently reported in [2]. The COVID-SemiSeg dataset consists of two sets. The first one contains 1600 pseudo labels generated by Semi-InfNet model and 50 labels by expert physicians. The second set includes 50 multi-class labels. Overall, there are 48 images that can be used for performance-comparison assessment and these CT data were used to compare our TV-Unet approach with other methods.
[1] COVID-19 CT segmentation dataset, link: https://medicalsegmentation.com/covid19/, accessed: 2020-04-11.
[2] COVID-SemiSeg Dataset, link: https://arxiv.org/pdf/2004.14133.pdf
- First dataset:
To evaluate the effect of radically different training/testing set composition and to demonstrate the robustness of the obtained results, two different splits of training, validation, and testing sets are selected from this dataset.
In Split 1, data from a relatively large number of 46 training/validation-set patients and a small number of only 3 testing-set patients were used. 729 CT image slices formed the training and validation sets, and 200 images the testing set. The number of slices in this split can be seen in main_TV_Unet_Split1.py code.
In Split 2, a more balanced distribution of patient numbers was used with 654 CT image slices from 35 patients included in the training and validation sets, and 275 images from 14 patients formed the testing set.
- The second dataset:
To compare TV-UNet model with the Inf-Net and other promising image segmentation models trained on COVID-SemiSeg dataset, including UNet++, Semi-Inf-Net, DeepLab-v3, FCN8s, and Semi-Inf-Net+FCN8s, training and testing sets are downloaded from here.
For this comparison just run main_TV_Unet_inf.py code.
This work is done by Narges Saeedizadeh, Shervin Minaee, Rahele Kafieh, Shakib Yazdani, and Milan Sonka (the previous editor in chief of IEEE TMI).
The Arxiv version of the paper can be downloaded from here.
If you find this work useful, you can refer our work as:
@article{ title={COVID TV-UNet: Segmenting COVID-19 Chest CT Images Using Connectivity Imposed U-Net},
author={Saeedizadeh, Narges and Minaee, Shervin and Kafieh, Rahele and Yazdani, Shakib and Sonka, Milan},
journal={arXiv preprint arXiv:2007.12303},
year={2020} }