MPS-Net: Multi-Point Supervised Network for CT Image Segmentation of COVID-19

Abstract

The new coronavirus, which has become a global pandemic, has confirmed more than 88 million cases worldwide since the first case was recorded in December 2019, causing over 1.9 million deaths. Since COIVD-19 lesions have clear imaging features on CT images, it is suitable for the auxiliary diagnosis and treatment of COVID-19. Deep learning can be used to segment the lesions areas of COVID-19 in CT images to help monitor the epidemic situation. In this paper, we propose a multi-point supervision network (MPS-Net) for segmentation of COVID-19 lung infection CT image lesions to solve the problem of a variety of lesion shapes and areas. A multi-scale feature extraction structure, a sieve connection structure (SC), a multi-scale input structure and a multi-point supervised training structure were implemented into MPS-Net. In order to increase the ability to segment various lesion areas of different sizes, the multi-scale feature extraction structure and the sieve connection structure will use different sizes of receptive fields to extract feature maps of various scales. The multi-scale input structure is used to minimize the edge loss caused by the convolution process. In order to improve the accuracy of segmentation, we propose a multi-point supervision training structure to extract supervision signals from different upsampling points on the network. Experimental results showed that the dice similarity coefficient (DSC), sensitivity, specificity and IOU of the segmentation results of our model are 0.8325, 0.8406, 09988 and 0.742, respectively. The experimental results demonstrated that the network proposed in this paper can effectively segment COVID-19 infection on CT images. It can be used to assist the diagnosis and treatment of new coronary pneumonia.

An Unsupervised Learning-Based Multi-Organ Registration Method for 3D Abdominal CT Images

“…Moreover, the dice coefficient (DICE), Haustorff distance (HD), contour mean distance (CMD), intersection over union (IOU), sensitivity (SS), and specificity (SC) are also included [ 37 ]: where and are the multi-organ segmentation masks of the fixed and warped images, respectively; and are the boundaries for and , respectively , , , and is the true positive, true negative, false negative, and false positive voxels in the multi-organ segmentation masks of the fixed and warped images, respectively. is the distance from to , and the distance from to .…”

Robust chest CT image segmentation of COVID-19 lung infection based on limited data

“…The 3D transformation of these architectures and the integration into our pipeline would be an interesting experiment to evaluate improvement possibilities. Other high-performance 2D approaches like Saood et al [ 37 ] and Pei et al [ 38 ] were difficult to compare due to these models are purely trained and evaluated on 2D slices with COVID-19 presence [ 66 ].…”

“… Related Work Training Dataset Validation/Testing Performance Author Model Architecture Source Dim. Sample Size Control DSC – COVID-19 Sample Size Amyar et al [ 5 ] U-Net (Standard) Amyar et al [ 5 ] 2D 1219 Yes 0.78 150 Fan et al [ 41 ] Inf-Net (Attention U-Net) Fan et al [ 41 ] 2D 1650 Yes 0.764 50 Qiu et al [ 43 ] MiniSeg (Attention U-Net) Qiu et al [ 43 ] 2D 3558 Yes 0.773 3558 Saood et al [ 37 ] U-Net (Standard) SIRM [ 66 ] 2D 80 No 0.733 20 Saood et al [ 37 ] SegNet SIRM [ 66 ] 2D 80 No 0.749 20 Pei et al [ 38 ] MPS-Net (Supervision U-Net) …”

“…Nonetheless, multiple approaches try to solve these problems with different methods. The most popular network models for COVID-19 segmentation are variants of the U-Net which achieved reasonable performance on sufficiently sized 2D datasets [ 5 , 10 , [33] , [34] , [35] , [36] , [37] , [38] , [39] , [40] ]. In order to compensate limited dataset sizes, more attention has been drawn to semi-supervised learning pipelines [ 10 , 41 , 42 ].…”

Review and classification of AI-enabled COVID-19 CT imaging models based on computer vision tasks

“…To accurately segment COVID-19 infected regions, and to solve the problem of the diversity of lesion shapes and areas, Pei et al [ 96 ] presented an architecture referred to as the multi-point supervision network (MPS-Net). Multi-scale feature extraction and sieve connection were used to extract features of various sizes.…”