Prediction of the Development of Pulmonary Fibrosis Using Serial Thin-Section CT and Clinical Features in Patients Discharged after Treatment for COVID-19 Pneumonia

Objective@#To identify predictors of pulmonary fibrosis development by combining follow-up thin-section CT findings and clinical features in patients discharged after treatment for COVID-19. @*Materials and Methods@#This retrospective study involved 32 confirmed COVID-19 patients who were divided into two groups according to the evidence of fibrosis on their latest follow-up CT imaging. Clinical data and CT imaging features of all the patients in different stages were collected and analyzed for comparison. @*Results@#The latest follow-up CT imaging showed fibrosis in 14 patients (male, 12; female, 2) and no fibrosis in 18 patients (male, 10; female, 8). Compared with the non-fibrosis group, the fibrosis group was older (median age: 54.0 years vs. 37.0 years, p = 0.008), and the median levels of C-reactive protein (53.4 mg/L vs. 10.0 mg/L, p = 0.002) and interleukin-6 (79.7 pg/L vs. 11.2 pg/L, p = 0.04) were also higher. The fibrosis group had a longer-term of hospitalization (19.5 days vs. 10.0 days, p = 0.001), pulsed steroid therapy (11.0 days vs. 5.0 days, p < 0.001), and antiviral therapy (12.0 days vs. 6.5 days, p = 0.012). More patients on the worst-state CT scan had an irregular interface (59.4% vs. 34.4%, p = 0.045) and a parenchymal band (71.9% vs. 28.1%, p < 0.001). On initial CT imaging, the irregular interface (57.1%) and parenchymal band (50.0%) were more common in the fibrosis group. On the worst-state CT imaging, interstitial thickening (78.6%), air bronchogram (57.1%), irregular interface (85.7%), coarse reticular pattern (28.6%), parenchymal band (92.9%), and pleural effusion (42.9%) were more common in the fibrosis group. @*Conclusion@#Fibrosis was more likely to develop in patients with severe clinical conditions, especially in patients with highinflammatory indicators. Interstitial thickening, irregular interface, coarse reticular pattern, and parenchymal band manifested in the process of the disease may be predictors of pulmonary fibrosis. Irregular interface and parenchymal band could predict the formation of pulmonary fibrosis early.

Automatic Detection and Classification of Rib Fractures on Thoracic CT Using Convolutional Neural Network: Accuracy and Feasibility

Objective@#To evaluate the performance of a convolutional neural network (CNN) model that can automatically detect and classify rib fractures, and output structured reports from computed tomography (CT) images. @*Materials and Methods@#This study included 1079 patients (median age, 55 years; men, 718) from three hospitals, between January 2011 and January 2019, who were divided into a monocentric training set (n = 876; median age, 55 years; men, 582), five multicenter/multiparameter validation sets (n = 173; median age, 59 years; men, 118) with different slice thicknesses and image pixels, and a normal control set (n = 30; median age, 53 years; men, 18). Three classifications (fresh, healing, and old fracture) combined with fracture location (corresponding CT layers) were detected automatically and delivered in a structured report. Precision, recall, and F1-score were selected as metrics to measure the optimum CNN model. Detection/diagnosis time, precision, and sensitivity were employed to compare the diagnostic efficiency of the structured report and that of experienced radiologists. @*Results@#A total of 25054 annotations (fresh fracture, 10089; healing fracture, 10922; old fracture, 4043) were labelled for training (18584) and validation (6470). The detection efficiency was higher for fresh fractures and healing fractures than for old fractures (F1-scores, 0.849, 0.856, 0.770, respectively, p = 0.023 for each), and the robustness of the model was good in the five multicenter/multiparameter validation sets (all mean F1-scores > 0.8 except validation set 5 [512 x 512 pixels; F1-score = 0.757]). The precision of the five radiologists improved from 80.3% to 91.1%, and the sensitivity increased from 62.4% to 86.3% with artificial intelligence-assisted diagnosis. On average, the diagnosis time of the radiologists was reduced by 73.9 seconds. @*Conclusion@#Our CNN model for automatic rib fracture detection could assist radiologists in improving diagnostic efficiency, reducing diagnosis time and radiologists’ workload.

Survey on deep learning for pulmonary medical imaging / 医学前沿

As a promising method in artificial intelligence, deep learning has been proven successful in several domains ranging from acoustics and images to natural language processing. With medical imaging becoming an important part of disease screening and diagnosis, deep learning-based approaches have emerged as powerful techniques in medical image areas. In this process, feature representations are learned directly and automatically from data, leading to remarkable breakthroughs in the medical field. Deep learning has been widely applied in medical imaging for improved image analysis. This paper reviews the major deep learning techniques in this time of rapid evolution and summarizes some of its key contributions and state-of-the-art outcomes. The topics include classification, detection, and segmentation tasks on medical image analysis with respect to pulmonary medical images, datasets, and benchmarks. A comprehensive overview of these methods implemented on various lung diseases consisting of pulmonary nodule diseases, pulmonary embolism, pneumonia, and interstitial lung disease is also provided. Lastly, the application of deep learning techniques to the medical image and an analysis of their future challenges and potential directions are discussed.

Effects of recombinant human erythropoietin β injection on levels of SOD, GSH-PX, MDA and Hcy with diabetic peritoneal dialysis / 中国生化药物杂志

Objective To investigate the effects of recombinant human erythropoietin βinjection on levels of superoxide dismutase ( SOD ) , glutathione peroxidase ( GSH-PX ) , malondialdehyde ( MDA ) and homocysteine ( Hcy ) in patients with diabetic peritoneal dialysis.Methods 92 patients of parallel peritoneal dialysis in diabetic nephropathy who received therapy from September 2014 to September 2016 in our hospital were selected and randomly divided into the observation group and the control group with 46 cases in each group.The control group was treated with peritoneal dialysis routine treatment, while the observation group was treated with recombinant human erythropoietin βinjection on this basis.The levels of hemoglobin (Hb), hematocrit (Hct), renal function, SOD, GSH-PX, MDA and Hcy were compared.Results After treatment, the levels of Hb and Hct in the observation group were higher than the control group, the difference was statistically significant (P<0.05), the urinary albumin excretion rate (UAER) and serum creatinine (SCr) in the observation group were lower than the control group, the difference was statistically significant (P<0.05), the levels of SOD and GSH-PX in the observation group were higher than the control group, the levels of MDA and Hcy were lower the control group, the difference was statistically significant (P<0.05).Conclusion The effect of recombinant human erythropoietin βinjection on diabetic nephropathy patients with peritoneal dialysis was significant, which could improve the levels of SOD, GSH-PX, MDA and Hcy.

A preliminary investigation on pulmonary subsolid nodule detection using deep learning methods from chest X-rays / 中华放射学杂志

Objective To evaluate the effectiveness of deep learning methods to detect subsolid nodules from chest X-ray images.Methods The building,training,and testing of the deep learning model were performed using the research platform developed by Infervision,China.The training dataset consisted of 1 965 chest X-ray images, which contained 85 labeled subsolid nodules and 1 880 solid nodules. Eighty-five subsolid nodules were confirmed by corresponding CT exams. We labeled each X-ray image using the corresponding reconstructed coronal slice from the CT exam as the gold standard,and trained the deep learning model using alternate training.After the training,the model was tested on a different dataset containing 56 subsolid nodules,which were also confirmed by corresponding coronal slices from CT exams. The model results were compared with an experienced radiologist in terms of sensitivity,specificity,and test time. Results Out of the testing dataset that contained 56 subsolid nodules, the deep learning model marked 72 nodules,which consisted of 39 true positives(TP)and 33 false positives(FP).The model took 17 seconds.The human radiologist marked 39 nodules,with 31 TP and 8 FP.The radiologist took 50 minutes and 24 seconds. Conclusions Subsolid nodules are prone to mis-diagnosis by human radiologists. The proposed deep learning model was able to effectively identify subsolid nodules from X-ray images.

Image segmentation in tumor CT based on the improved C-V model / 生物医学工程学杂志

Aiming at the shortcomings of slow convergence and inaccuracy segmentation in non-homogeneous images, improvements were made on the traditional C-V model in two aspects. Firstly, using a novel model based on local gradient, the initial contour of the C-V model was quickly moved near the target border, greatly reducing the evolution time. Secondly, combining the characteristics of GVF model from two directions to the target border, an adaptive velocity reconciling item was added for velocity equation of the C-V model to make the model converge to the true border. The segmentation experiments for liver tumors in CT showed that the proposed method could be effective.

Importance of Injury Rehabilitation Tertiary Prevention System / 中国康复理论与实践

@#Injury Rehabilitation Tertiary Prevention System is important to reduce accident, implement medical emergency in time, and recover the function for the people injured in the accident.