OBJECTIVE:In this retrospective multicentre cohort study, we aimed to develop and validate an interpretable machine learning (ML) model for early post-stroke epilepsy (PSE) prediction in patients who underwent mechanical thrombectomy (MT). We hypothesised that an ML model developed using electronic medical record data can accurately predict the risk of PSE in patients with acute ischaemic stroke treated with MT. METHODS:We collected data from 1185 adult patients who underwent MT between 2017 and 2022 at two hospitals for model development and used data from three external centres for validation. The key variables were selected using Cox regression, least absolute shrinkage and selection operator regression and Boruta. Eight ML algorithms were used for model development. The model performance was evaluated using the concordance index (C-index) and integrated Brier scores. Additionally, the 1-year, 3-year and 5-year PSE risk prediction accuracies were assessed using the areas under the receiver operating characteristic curves (AUCs) and calibration curves. Decision curve analysis was used to evaluate the clinical utility, and the feature importance was quantified. RESULTS:An optimal combination comprising nine features was used for model development. Of the eight ML models, the random survival forest (RSF) model demonstrated optimal performance and maintained robust discriminative accuracy across the three external validation cohorts (C-index: 0.812; AUC: 0.748-0.837; C-index: 0.822; AUC: 0.795-0.839 and C-index: 0.763; AUC: 0.763-0.833). Finally, we developed an online calculator (https://pseshiny.shinyapps.io/shinydashboard_sa_1model/) based on the RSF model. SIGNIFICANCE:This prediction model could help identify patients at high risk of developing PSE following MT and support individualised treatment strategies.

BACKGROUND:Acute ischemic stroke (AIS) represents a significant etiology of acquired epilepsy in adults. However, there remains a paucity of evidence concerning predictive factors for poststroke epilepsy (PSE) among patients undergoing mechanical thrombectomy (MT). We aimed to evaluate the association between poor collateral circulation and the risk of PSE. METHODS:This multicenter cohort study included AIS patients who underwent MT (2017-2022) at three tertiary centers. Adjusted Cox proportional hazard models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for PSE risk. Subgroup analyses, sensitivity analyses, and competing risk analyses were also performed. RESULTS:We included 1167 patients (mean age: 62 years; 46.5% female). PSE frequency was higher in the poor collateral circulation group compared to the control group (13.0% vs. 4.2%, p < 0.001). Patients with inadequate collateral circulation (ASITN/SIR < 2) showed significantly increased risks of both 12-month PSE and 60-month PSE development (HR > 1, p < 0.05) compared to those with adequate collateral status (ASITN/SIR ≥ 2) following MT for AIS. CONCLUSIONS:Poor collateral circulation is associated with PSE and may help identify individuals at high risk for PSE.

Purpose:During the acute phase of obstructive pulmonary disease (COPD), completing a standard pulmonary function test may be challenging for some patients. The goal of this experiment is to develop a machine learning model that uses chest CT images for automated diagnosis and grading of COPD patients, aiming to enhance diagnostic efficiency and accuracy. Patients and Methods:The study retrospectively included 173 COPD patients and 176 healthy controls from December 2017 to June 2023. Deep learning segmentation modules were used to automatically segment the obtained chest CT images for lung parenchyma, airway, pulmonary artery, and vein. Imaging features were extracted from these segmented regions. The most reliable and relevant features were selected using Mann-Whitney U-test with a significant p-value of 0.05 and the least absolute shrinkage and selection operator (LASSO) method. Machine learning models were established through support vector machine (SVM) classifier in the training set and further tested in the internal testing set. Additional tests were performed on an external testing set with 68 individuals. Results:In the machine learning model for COPD diagnosis, the image model achieved AUC values of 0.981 and 0.977 in the training and testing sets, with corresponding accuracies of 0.949 and 0.956 respectively. For COPD severity grading, the image model obtained AUC values of 0.889 and 0.796 in the training and testing sets, along with accuracies of 0.784 and 0.719. Conclusion:The machine learning model based on chest CT images can accurately predict lung function, which can assist in the diagnosis and severity grading of COPD.

BackgroundNon-motor symptoms (NMS) in Parkinson’s disease (PD) often precede motor manifestations and are challenging to detect with conventional MRI. This study investigates the use of the Multi-Flip-Angle and Multi-Echo Gradient Echo Sequence (MULTIPLEX) in MRI to detect previously undetectable microstructural changes in brain tissue associated with NMS in PD.MethodsA prospective study was conducted on 37 patients diagnosed with PD. Anxiety and depression levels were assessed using the Hamilton Anxiety Scale (HAMA) and Hamilton Depression Scale (HAMD), respectively. MRI techniques, including 3D T1-weighted imaging (3D T1WI) and MULTIPLEX - which encompasses T2*-mapping, T1-mapping, proton density-mapping, and quantitative susceptibility mapping (QSM)—were performed. Brain subregions were automatically segmented using deep learning, and their volume and quantitative parameters were correlated with NMS-related assessment scales using Spearman’s rank correlation coefficient.ResultsCorrelations were observed between QSM and T2* values of certain subregions within the left frontal and bilateral temporal lobes and both anxiety and depression (absolute r-values ranging from 0.358 to 0.480, p < 0.05). Additionally, volume measurements of regions within the bilateral frontal, temporal, and insular lobes exhibited negative correlations with anxiety and depression (absolute r-values ranging from 0.354 to 0.658, p < 0.05). In T1-mapping and proton density-mapping, no specific brain regions were found to be significantly associated with the NMS of PD under investigation.ConclusionQuantitative parameters derived from MULTIPLEX MRI show significant associations with clinical evaluations of NMS in PD. Multiparametric MR neuroimaging may serve as a potential early diagnostic tool for PD.

ObjectivesBy employing deep learning-based automatic whole-brain region segmentation technology, we aim to investigate the cross-sectional associations between regional brain volumes and disease duration in patients with Parkinson’s disease (PD).MethodsA retrospective study design was implemented on 83 patients diagnosed with idiopathic PD who had complete clinical and imaging data. Cranial magnetic resonance images (MRI) were imported into the uAI platform for automated regional segmentation of brain tissue. Volumetric data from five major brain regions and 80 subregions were extracted to explore their potential associations with disease progression in PD patients. Statistical analysis was conducted using a multiple linear regression model within the framework of linear regression analysis, with statistical significance defined as p < 0.05.ResultsCross-sectional analysis revealed that in PD patients, volume ratios of multiple brain regions—including the bilateral precentral gyrus, right medial frontal gyrus, bilateral postcentral gyrus, bilateral superior and inferior parietal lobules, bilateral precuneus, right cuneus, right lingual gyrus, bilateral lateral occipital gyrus, and right globus pallidus—were negatively associated with disease duration (p < 0.05). In contrast, the right hippocampus, right inferior temporal gyrus, and left superior temporal gyrus showed positive correlations (p < 0.05). The combined volume ratios of these brain regions also decreased with longer disease duration (p < 0.05). Furthermore, absolute volume differences in the hippocampus, fusiform gyrus, isthmus of the cingulate gyrus, and cerebellar white matter increased as the disease progressed (p < 0.05).ConclusionIn PD patients, volume ratios and absolute volume differences in specific brain subregions associated with lateralized intracranial changes may serve as potential biomarkers for assessing brain tissue alterations during disease progression.

BackgroundDendritic cells (DCs) play a crucial role in anti-tumor immunity by capturing, processing, and presenting tumor antigens to T cells, making DC-based immunotherapy a promising approach for cancer treatment. However, the most commonly used clinical strategy still relies on inducing DCs in vitro using granulocyte-macrophage colony-stimulating factor (GM-CSF) and interleukin-4 (IL - 4) (GM/IL4-DCs), which often results in a heterogeneous cell population with suboptimal anti-tumor function. Here, we compared DCs generated by co-stimulating with FMS-like tyrosine kinase 3 ligand (FLT3L) and GM-CSF (FL/GM-DCs) with the conventional GM/IL4-DCs.MethodTo compare the functional differences of DCs induced by different methods, we conducted a comprehensive study. Mouse bone marrow cells were continuously cultured for 9 days in a FLT3L/GM-CSF-containing medium. After cell collection, we analyzed the composition, subpopulations, and status of FL/GM-DCs using flow cytometry and scRNA-seq. Flow cytometry was also used to assess their antigen presentation and ability to stimulate T cells. In vivo experiments were performed to examine their distribution, anti-tumor effects, and therapeutic responses in tumor models. Finally, combining scRNA-seq and scTCR-seq, we explored the mechanisms by which FL/GM-DCs reshape the tumor microenvironment.ResultsThe results showed that FL/GM-DCs exhibited a unique subpopulation distribution, characterized by an abundance of conventional cDC subpopulations, and demonstrated enhanced cross-antigen presentation capabilities. Notably, FL/GM-DCs were able to induce a broader and more tumor-specific CD8+ T cell response, effectively reshaping the tumor microenvironment by promoting the infiltration of cytotoxic T lymphocytes (CTLs) and reducing immunosuppressive components. In contrast, GM/IL4-DCs contained fewer cDC subpopulations, eliciting a weaker initial CD8+ T cell response and yielding relatively inferior anti-tumor effects.ConclusionIn summary, FLT3L combined with GM-CSF induced DCs, through their unique subpopulation composition and functional state, can more effectively expand tumor-specific CD8+ T cells and reshape the tumor microenvironment, thereby achieving superior immunotherapy outcomes. This study highlights the potential of FL/GM-DCs as a next-generation DC platform, paving the way for improved clinical translation of DC-based adoptive cancer immunotherapies.

PURPOSE:Magnetic resonance imaging (MRI) refers to one of the critical image modalities for diagnosis, whereas its long acquisition time limits its application. In this study, the aim was to investigate whether deep learning-based techniques are capable of using the common information in different MRI sequences to reduce the scan time of the most time-consuming sequences while maintaining the image quality. METHOD:Fully sampled T1-FLAIR, T2-FLAIR, and T2WI brain MRI raw data originated from 217 patients and 105 healthy subjects. The T1-FLAIR and T2-FLAIR sequences were subsampled using Cartesian masks based on four different acceleration factors. The fully sampled T1/T2-FLAIR images were predicted from undersampled T1/T2-FLAIR images and T2WI images through deep learning-based reconstruction. They were qualitatively assessed by two senior radiologists in accordance with the diagnosis decision and a four-point scale image quality score. Furthermore, the images were quantitatively assessed based on regional signal-to-noise ratios (SNRs) and contrast-to-noise ratios (CNRs). The chi-square test was performed, where p < 0.05 indicated a difference with statistical significance. RESULTS:The diagnosis decisions from two senior radiologists remained unchanged in accordance with the accelerated and fully sampled images. There were no significant differences in the regional SNRs and CNRs of most assessed regions (p > 0.05) between the accelerated and fully sampled images. Moreover, no significant difference was identified in the image quality assessed by two senior radiologists (p > 0.05). CONCLUSION:Deep learning-based image reconstruction is capable of significantly expediting the brain MR imaging process and producing acceptable image quality without affecting diagnosis decisions.

In prior research, both miRNA-125b and BLZ945 have shown potential in effectively inhibiting M2 macrophage polarization and producing antitumor effects. Nevertheless, their physicochemical characteristics present significant challenges for efficient in vivo delivery. Ionizable cationic lipid nanoparticles (LNPs), recognized for their superior biocompatibility and drug-loading capacity, serve as a novel carrier for nucleic acid-based therapeutics. In our study, we successfully encapsulated both agents within LNPs and conducted a thorough characterization. Subsequently, we investigated their potential to repolarize M2 macrophages in vitro and evaluated their in vivo distribution, biosafety, and antitumor efficacy. The findings revealed that the LNPs maintained excellent drug-loading efficiency, consistent particle size, and stable zeta potential. All formulations effectively inhibited M2 macrophage polarization in vitro. Upon administration in vivo, the LNPs not only demonstrated favorable biosafety profiles but also accumulated efficiently in tumor tissues, substantially reducing tumor burden, particularly notable in co-loaded LNPs. Our results affirm that LNPs are an effective carrier for miRNA-125b and BLZ945, highlighting this encapsulation approach as promising for the treatment of solid tumors and meriting further investigation. Practitioner points: (i) Ionizable cationic nanoparticles provide high and stable encapsulation rates to efficiently load nucleic acid polymers into the LNP, avoiding the rapid accumulation of circulating macrophages, which can lead to reduced penetration of the LNP into target tissues. Therefore, it can be used as a novel drug delivery method to benefit clinical patients. (ii) miRNA-125b LNP/BLZ945 LNP attenuated the depleting effect of BLZ945 on macrophages and significantly inhibited macrophage M2 polarization. It could be effectively distributed in tumors and showed good biosafety while exerting antitumor effects, bringing hope to clinical pancreatic tumor patients.

臂丛神经损伤(BPI)是一种严重的上肢神经损伤,伤后上肢尤其手部运动的恢复极为困难.恢复程度主要取决于伤后治疗是否及时,故BPI的早期诊断就显得尤为重要[1].早期的臂丛神经重建术对于损伤神经的恢复具有重要意义,然而,对损伤的早期诊断、分级以及手术方法的选择依旧困难重重[2].

Background and aimsDyslipidemia is known to contribute to arterial stiffness, while the inverse association remains unknown. This study aimed to explore the association of baseline arterial stiffness and its changes, as determined by brachial-ankle pulse wave velocity (baPWV), with dyslipidemia onset in the general population.MethodsThis study enrolled participants from Beijing Health Management Cohort using measurements of the first visit from 2012 to 2013 as baseline, and followed until the dyslipidemia onset or the end of 2019. Unadjusted and adjusted Cox proportional regression models were used to evaluate the associations of baseline baPWV and baPWV transition (persistent low, onset, remitted and persistent high) with incident dyslipidemia.ResultsOf 4362 individuals (mean age: 55.5 years), 1490 (34.2%) developed dyslipidemia during a median follow-up of 5.9 years. After adjusting for potential confounders, participants with elevated arterial stiffness at baseline had an increased risk of dyslipidemia (HR, 1.194; 95% CI, 1.050-1.358). Compared with persistent low baPWV, new-onset and persistent high baPWV were associated with a 51.2% and 37.1% excess risk of dyslipidemia.ConclusionThe findings indicated that arterial stiffness is an early risk factor of dyslipidemia, suggesting a bidirectional association between arterial stiffness and lipid metabolism.

Three-Dimensional (3D) sequence of magnetic resonance imaging (MRI) plays a critical role in musculoskeletal joints imaging; however, its long acquisition time limits the clinical application. In such condition, Compressed sensing (CS) is introduced to accelerate MRI during clinical practice. We aimed to investigate the feasibility of an isotropic 3D variable-flip-angle fast spin echo (FSE) sequence with CS technique (CS-MATRIX) compared to conventional 2D sequences on knee imaging.

Purpose:To evaluate and compare the image quality and diagnostic accuracy of Artificial Intelligence-assisted Compressed Sensing (ACS) sequences for lumbar disease, as an acceleration method for MRI combining parallel imaging, half-Fourier, compressed sensing and neural network and routine 2D sequences for lumbar spine. Methods:We collected data from 82 healthy subjects and 213 patients who used 2D ACS accelerated sequences to examine the lumbar spine while 95 healthy subjects and 234 patients used routine 2D sequences. Acquisitions included axial T2WI, sagittal T2WI, T1WI, and T2-fs sequences. All obtained images of these subjects were analyzed in the light of calculating image quality factors such as signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) for selected regions of interest. The lumbar image quality, artifacts and visibility of lesion structure were assessed by two radiologists independently. Differences between the evaluation values above were tested for statistical significance by the Wilcoxon signed-ranks test. Inter-observer agreements of image quality between two radiologists were measured using Cohen's kappa correlation coefficient. Results:The ACS accelerated sequences not only reduced the scanning time by 18.9%, but also retained basically the same image quality as the routine 2D sequences in both healthy subjects and patients. Artifacts are less produced on ACS accelerated sequences compared with routine 2D sequences (p < 0.05). Apart from this, there were no significant differences in quantitative SNR, CNR measurements and qualitative scores within reviewing radiologists for each group (p > 0.05). Moreover, inter-observer agreement between two radiologists in scoring image quality was substantial consistently for ACS accelerated sequences and routine sequences (kappa = 0.622-0.986). Conclusion:Compared with routine 2D sequences, ACS accelerated sequences allow for faster lumbar spine imaging with similar imaging quality and present reliable diagnostic accuracy, which can potentially improve workflow and patient comfort in musculoskeletal examinations.

MRI is widely used in the diagnosis of spinal diseases, but due to the limitation of long scanning time, it is often difficult for patients to tolerate. This study aims at exploring whether the ACS (AI-assisted Compressed Sensing) accelerated sequences can assure similar image quality and diagnostic accuracy of lumbar spine diseases compared to 2D conventional sequences. Results suggested that 2D ACS sequence has become a well-anticipated technique for routine examination of lumbar spine and lumbar spine lesions diagnosis, while ensuring high image quality and reduced scan time effectively than traditional 2D sequences in MRI.

PurposeThyroid hormones sensitivity is a newly proposed clinical entity closely related with metabolic health. Prior studies have reported the cross-sectional relationship between thyroid hormones sensitivity and diabetes; however, the longitudinal association is unclear to date. We aimed to explore the relationship between impaired thyroid hormone sensitivity at baseline and diabetes onset using a cohort design.MethodsThis study enrolled 7283 euthyroid participants at the first visit between 2008 and 2009, and then annually followed until diabetes onset or 2019. Thyrotropin (TSH), free triiodothyronine (FT3) and free thyroxine (FT4) were measured to calculate thyroid hormone sensitivity by thyroid feedback quantile-based index (TFQI), Chinese-referenced parametric thyroid feedback quantile-based index (PTFQI), thyrotropin index (TSHI), thyrotroph thyroxine resistance index (TT4RI) and FT3/FT4 ratio. Cox proportional hazard model and cross-lagged panel analysis were used.ResultsThe mean baseline age was 44.2 ± 11.9 years, including 4170 (57.3%) male. During a median follow-up of 5.2 years, 359 cases developed diabetes. There was no significant association between thyroid hormones sensitivity indices and diabetes onset, and adjusted hazard ratios per unit (95% CIs) were 0.89 (0.65-1.23) for TFQI, 0.91 (0.57-1.45) for PTFQI, 0.95 (0.70-1.29) for TSHI, 0.98 (0.70-1.01) for TT4RI and 2.12 (0.17-5.78) for FT3/FT4 ratio. Cross-lagged analysis supported the temporal association from fasting glucose to impaired thyroid hormones sensitivity indices.ConclusionsOur findings could not demonstrate that thyroid hormones sensitivity status is a predictor of diabetes onset in the euthyroid population. Elevated fasting glucose (above 7.0 mmol/L) appeared to precede impaired sensitivity indices of thyroid hormones.

Purpose: The three-dimensional (3D) sequence of magnetic resonance imaging (MRI) plays a critical role in the imaging of musculoskeletal joints; however, its long acquisition time limits its clinical application. In such conditions, compressed sensing (CS) is introduced to accelerate MRI in clinical practice. We aimed to investigate the feasibility of an isotropic 3D variable-flip-angle fast spin echo (FSE) sequence with CS technique (CS-MATRIX) compared to conventional 2D sequences in knee imaging. Methods: Images from different sequences of both the accelerated CS-MATRIX and the corresponding conventional acquisitions were prospectively analyzed and compared. The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of the structures within the knees were measured for quantitative analysis. The subjective image quality and diagnostic agreement were compared between CS-MATRIX and conventional 2D sequences. Quantitative and subjective image quality scores were statistically analyzed with the paired t-test and Wilcoxon signed-rank test, respectively. Diagnostic agreements of knee substructure were assessed using Cohen's weighted kappa statistic. Results: For quantitative analysis, images from the CS-MATRIX sequence showed a significantly higher SNR than T2-fs 2D sequences for visualizing cartilage, menisci, and ligaments, as well as a higher SNR than proton density (pd) 2D sequences for visualizing menisci and ligaments. There was no significant difference between CS-MATRIX and 2D T2-fs sequences in subjective image quality assessment. The diagnostic agreement was rated as moderate to very good between CS-MATRIX and 2D sequences. Conclusion: This study demonstrates the feasibility and clinical potential of the CS-MATRIX sequence technique for detecting knee lesions The CS-MATRIX sequence allows for faster knee imaging than conventional 2D sequences, yielding similar image quality to 2D sequences.

BackgroundStroke is a major disease with high morbidity and mortality worldwide. Currently, there is no quantitative method to evaluate the short-term prognosis and length of hospitalization of patients.PurposeWe aimed to develop nomograms as prognosis predictors based on imaging characteristics from non-contrast computed tomography (NCCT) and CT perfusion (CTP) and clinical characteristics for predicting activity of daily living (ADL) and hospitalization time of patients with ischemic stroke.Materials and methodsA total of 476 patients were enrolled in the study and divided into the training set (n = 381) and testing set (n = 95). Each of them owned NCCT and CTP images. We propose to extract imaging features representing as the Alberta stroke program early CT score (ASPECTS) values from NCCT, ischemic lesion volumes from CBF, and TMAX maps from CTP. Based on imaging features and clinical characteristics, we addressed two main issues: (1) predicting prognosis according to the Barthel index (BI)–binary logistic regression analysis was employed for feature selection, and the resulting nomogram was assessed in terms of discrimination capability, calibration, and clinical utility and (2) predicting the hospitalization time of patients–the Cox proportional hazard model was used for this purpose. After feature selection, another specific nomogram was established with calibration curves and time-dependent ROC curves for evaluation.ResultsIn the task of predicting binary prognosis outcome, a nomogram was constructed with the area under the curve (AUC) value of 0.883 (95% CI: 0.781–0.985), the accuracy of 0.853, and F1-scores of 0.909 in the testing set. We further tried to predict discharge BI into four classes. Similar performance was achieved as an AUC of 0.890 in the testing set. In the task of predicting hospitalization time, the Cox proportional hazard model was used. The concordance index of the model was 0.700 (SE = 0.019), and AUCs for predicting discharge at a specific week were higher than 0.80, which demonstrated the superior performance of the model.ConclusionThe novel non-invasive NCCT- and CTP-based nomograms could predict short-term ADL and hospitalization time of patients with ischemic stroke, thus allowing a personalized clinical outcome prediction and showing great potential in improving clinical efficiency.SummaryCombining NCCT- and CTP-based nomograms could accurately predict short-term outcomes of patients with ischemic stroke, including whose discharge BI and the length of hospital stay.Key ResultsUsing a large dataset of 1,310 patients, we show a novel nomogram with a good performance in predicting discharge BI class of patients (AUCs > 0.850). The second nomogram owns an excellent ability to predict the length of hospital stay (AUCs > 0.800).

Objective To develop a deep learning-based model using esophageal thickness to detect esophageal cancer from unenhanced chest CT images. Methods We retrospectively identified 141 patients with esophageal cancer and 273 patients negative for esophageal cancer (at the time of imaging) for model training. Unenhanced chest CT images were collected and used to build a convolutional neural network (CNN) model for diagnosing esophageal cancer. The CNN is a VB-Net segmentation network that segments the esophagus and automatically quantifies the thickness of the esophageal wall and detect positions of esophageal lesions. To validate this model, 52 false negatives and 48 normal cases were collected further as the second dataset. The average performance of three radiologists and that of the same radiologists aided by the model were compared. Results The sensitivity and specificity of the esophageal cancer detection model were 88.8% and 90.9%, respectively, for the validation dataset set. Of the 52 missed esophageal cancer cases and the 48 normal cases, the sensitivity, specificity, and accuracy of the deep learning esophageal cancer detection model were 69%, 61%, and 65%, respectively. The independent results of the radiologists had a sensitivity of 25%, 31%, and 27%; specificity of 78%, 75%, and 75%; and accuracy of 53%, 54%, and 53%. With the aid of the model, the results of the radiologists were improved to a sensitivity of 77%, 81%, and 75%; specificity of 75%, 74%, and 74%; and accuracy of 76%, 77%, and 75%, respectively. Conclusions Deep learning-based model can effectively detect esophageal cancer in unenhanced chest CT scans to improve the incidental detection of esophageal cancer.

The worldwide spread of coronavirus disease (COVID-19) has become a threat to global public health. It is of great importance to rapidly and accurately screen and distinguish patients with COVID-19 from those with community-acquired pneumonia (CAP). In this study, a total of 1,658 patients with COVID-19 and 1,027 CAP patients underwent thin-section CT and were enrolled. All images were preprocessed to obtain the segmentations of infections and lung fields. A set of handcrafted location-specific features was proposed to best capture the COVID-19 distribution pattern, in comparison to the conventional CT severity score (CT-SS) and radiomics features. An infection size-aware random forest method (iSARF) was proposed for discriminating COVID-19 from CAP. Experimental results show that the proposed method yielded its best performance when using the handcrafted features, with a sensitivity of 90.7%, a specificity of 87.2%, and an accuracy of 89.4% over state-of-the-art classifiers. Additional tests on 734 subjects, with thick slice images, demonstrates great generalizability. It is anticipated that our proposed framework could assist clinical decision making.

The coronavirus disease, named COVID-19, has become the largest global public health crisis since it started in early 2020. CT imaging has been used as a complementary tool to assist early screening, especially for the rapid identification of COVID-19 cases from community acquired pneumonia (CAP) cases. The main challenge in early screening is how to model the confusing cases in the COVID-19 and CAP groups, with very similar clinical manifestations and imaging features. To tackle this challenge, we propose an Uncertainty Vertex-weighted Hypergraph Learning (UVHL) method to identify COVID-19 from CAP using CT images. In particular, multiple types of features (including regional features and radiomics features) are first extracted from CT image for each case. Then, the relationship among different cases is formulated by a hypergraph structure, with each case represented as a vertex in the hypergraph. The uncertainty of each vertex is further computed with an uncertainty score measurement and used as a weight in the hypergraph. Finally, a learning process of the vertex-weighted hypergraph is used to predict whether a new testing case belongs to COVID-19 or not. Experiments on a large multi-center pneumonia dataset, consisting of 2148 COVID-19 cases and 1182 CAP cases from five hospitals, are conducted to evaluate the prediction accuracy of the proposed method. Results demonstrate the effectiveness and robustness of our proposed method on the identification of COVID-19 in comparison to state-of-the-art methods.

Chest computed tomography (CT) becomes an effective tool to assist the diagnosis of coronavirus disease-19 (COVID-19). Due to the outbreak of COVID-19 worldwide, using the computed-aided diagnosis technique for COVID-19 classification based on CT images could largely alleviate the burden of clinicians. In this paper, we propose an Adaptive Feature Selection guided Deep Forest (AFS-DF) for COVID-19 classification based on chest CT images. Specifically, we first extract location-specific features from CT images. Then, in order to capture the high-level representation of these features with the relatively small-scale data, we leverage a deep forest model to learn high-level representation of the features. Moreover, we propose a feature selection method based on the trained deep forest model to reduce the redundancy of features, where the feature selection could be adaptively incorporated with the COVID-19 classification model. We evaluated our proposed AFS-DF on COVID-19 dataset with 1495 patients of COVID-19 and 1027 patients of community acquired pneumonia (CAP). The accuracy (ACC), sensitivity (SEN), specificity (SPE), AUC, precision and F1-score achieved by our method are 91.79%, 93.05%, 89.95%, 96.35%, 93.10% and 93.07%, respectively. Experimental results on the COVID-19 dataset suggest that the proposed AFS-DF achieves superior performance in COVID-19 vs. CAP classification, compared with 4 widely used machine learning methods.