The current unsupervised cross‐domain detection methods need source domain data to retrain the detection model in target domain. However, the source domain data may be unavailable due to privacy, decentralization, or computation resource restrictions. A natural idea is to optimize the parameters of the source domain model by self‐supervised learning based on pseudo labels. We propose another approach from the viewpoint of noise perturbation without pseudo‐labeling. It can be assumed that the source and target domains are actually derived from a domain invariant space through domain‐specific perturbations, respectively. A super target domain can be constructed by augmenting more target domain perturbations to the target domain images. The optimal direction of the target domain to the domain invariant space can be approximated as the alignment direction from the super target domain to the target domain. Based on this idea, we propose a novel method called SOAP (SOurce data‐free domain Adaptation through domain Perturbation) which can remove domain perturbation from the target domain. The image‐level, instance‐level, and category consistency regularizations based on Mean Teacher structure are proposed to learn the correct alignment direction. Specifically, the category consistency can also further improve the classification accuracy. Extensive experiments on multiple domain adaptation scenarios demonstrate that SOAP achieves better performance surpassing the baseline (Faster R‐CNN) and multiple state‐of‐the‐art domain adaptation methods which need to access source domain data.
Hyper-graph techniques have been widely investigated in computer vision and medical imaging applications, showing superior performance for modeling complex subject-wise relationships and sufficient flexibility to deal with missing data from multi-modal neuroimaging data. Existing hyper-graph methods, however, are inadequate for two reasons. First, representations are generated only from the observed imaging data, a process that is completely independent of the subsequent data label inference/ classification step. Thus, hyper-graph results constructed in this way may not be consistent with phenotype data such as clinical labels or scores. More critically, it might generate sub-optimal predictions in relation to clinical labels/scores. Second, current hyper-graph inference methods rely on two sequential steps: 1) building the hyper-graph for each individual modality and then predicted latent labels for new subjects upon each constructed hyper-graph and 2) a voting procedure to incorporate inference results across different hyper-graphs. This approach, however, is limited by failing to consider the complex and complementary relationships of multi-modal imaging data with respect to hyper-graph inference procedure. To address these two issues, we propose a novel dynamic hyper-graph inference method supported by a semi-supervised framework. Our method iteratively estimates and adjusts the hyper-graph structures from multi-modal imaging data until consistency between the learned hyper-graph and the observed clinical labels and scores is achieved. This hyper-graph inference framework also eases the integration process of classification (identifying individuals having neurodegenerative disease) and regression (predicting the clinical scores) within the same framework. The experimental results on identifying mild cognition impairment (MCI) subjects and the fine grained recognition of MCI progression stages show improved performance using our proposed hyper-graph inference method compared with conventional methods.
Objective. Computed tomography (CT) scan is a method to predict the progression and prognosis of COVID-19. It is not sufficient merely to measure the prognosis of COVID-19 without other clinical methods. The purpose of this study was to investigate the association between the CT scan and clinical laboratory indicators as well as clinical manifestations. Method. A total of 335 patients were enrolled from January 26, 2020, to February 26, 2020, in Shandong province and Huanggang city. Demographic and clinical characteristics, laboratory variables, and the data from the CT scans were collected for analysis. Scatter plot analysis and correlation analysis were used to calculate the relationship between CT evaluation and other indicators. Multivariable linear regression analysis was used to establish a model for diagnostic and prognostic prediction. Age, CRP, LDH, and lymphocyte counts as independent variables were selected to develop a predictive model, and the results from the CT scans to reflect the degree of lung injury were taken as the dependent variable. Result. The median age was 44 years (IQR: 34–56); among them, 188 (56%) were male. Severe patients were older (56 vs. 40, ). There were statistically significant differences in lymphocyte counts, platelet counts, C-reactive protein (CRP), lactate dehydrogenase (LDH), procalcitonin (PCT), and creatine kinase (CK) between the general patients and severe patients. We found that, without effective antiviral treatment, mild patients had a 6-day interval from symptom onset to CRP elevation, but in severe patients, CRP started to increase from day 2. Lung injury score from a chest CT scan and incidence of acute respiratory distress syndrome (ARDS) were significantly higher in severe patients than in mild patients. Lung injury score from a chest CT scan was closely correlated with CRP (rs = 0.704, ), and they reflected the severity of the disease. The receiver operating curve (ROC) value of the injury score from the chest CT scan was 0.854 (95% CI: 0.808–0.901), and the area under the curve (AUC) value of CRP was 0.823 (95% CI: 0.769–0.878). Conclusion. The results from CRP and chest CT scans were indicators of the severity of COVID-19. Combining patient age, CRP, LDH, and lymphocyte counts, we developed a model that could help to predict lung injury/function of patients with COVID-19.