An Effective Deep Neural Network for Lung Lesions Segmentation From COVID-19 CT Images

Abstract

Automatic segmentation of lung lesions from COVID-19 Computed Tomography (CT) images can help to establish a quantitative model for diagnosis and treatment. For this reason, this work provides a new segmentation method to meet the needs of CT images processing under COVID-19 epidemic. The main steps are as follows: Firstly, the proposed Region of Interest (ROI) extraction implements patch mechanism strategy to satisfy the applicability of 3D network and remove irrelevant background. Secondly, 3D network is established to extract spatial features, where 3D Attention model promotes network to enhance target area. Then, to improve convergence of network, a combination loss function is introduced to lead gradient optimization and training direction. Finally, data augmentation and Conditional Random Field (CRF) are applied to realize data resampling and binary segmentation. This method was assessed with some comparative Experiment. By comparison, the proposed method reached the highest performance. Therefore, it has potential clinical applications.

Automatic Detection of Colorectal Polyps Using Transfer Learning

“…Since this is a medical application, the meaning of each metric must be explained. The interpretation of the results differs according to the intended purpose [42][43][44]. (e) F1 score: is considered a better benchmark than accuracy when the aim is to compare different models [12].…”

“…Since this is a medical application, the meaning of each metric must be explained. The interpretation of the results differs according to the intended purpose [ 42 , 43 , 44 ]. Sensitivity: represents the cases correctly classified as positive relative to the actual number of positive cases: Specificity: represents the ratio of classified negative cases relative to truly negative cases: Precision: means the ratio of those correctly classified as positive to the total number of positive cases: Accuracy: means the ratio of those correctly classified as positive to the total number of positive cases: F1 score: is considered a better benchmark than accuracy when the aim is to compare different models [ 12 ].…”

Learning to learn by yourself: Unsupervised meta‐learning with self‐knowledge distillation for COVID‐19 diagnosis from pneumonia cases

“…Most patients have an air bronchogram 60 . The distribution characteristics of the abnormalities on X‐ray images about these five types of pneumonia are similar to those of CT images (slices) 52,61‐73 . Although the collected 2D data (e.g., X‐ray images) in our proposed data set misses lots of information (original intensity level, spacing, etc.)…”

“…60 The distribution characteristics of the abnormalities on X-ray images about these five types of pneumonia are similar to those of CT images (slices). 52,[61][62][63][64][65][66][67][68][69][70][71][72][73] Although the collected 2D data (e.g., X-ray images) in our proposed data set misses lots of information (original intensity level, spacing, etc.) than original volume data, considering the usage of our proposed 2D-oriented algorithm, we have tried our best to keep the original size of the images while avoiding the problem of image distortion.…”

Analyzing Groups of Inpatients’ Healthcare Needs to Improve Service Quality and Sustainability

“…Several studies [16][17][18][19] applied deep neural networks to detect coronavirus diseases from the X-ray and computed tomography (CT) images, and some evaluated multiple convolutional neural network (CNN) models. Chen et al [20] proposed an efficient deep learning model for removing irrelevant backgrounds, extracting spatial features, and automatically segmenting lung lesions from CT images. Vidal [21] et al proposed a multi-stage transfer learning approach to obtain a robust system able to segment lung regions from portable X-ray devices despite the lack of samples.…”