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.

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.

To retrospectively validate CT-based radiomics features for predicting the risk of anterior mediastinal lesions. A retrospective study was performed through February 2013 to March 2018 on 298 patients who had pathologically confirmed anterior mediastinal lesions. The patients all underwent CT scans before their treatment, including 130 unenhanced computed tomography (UECT) and 168 contrast-enhanced CT (CECT) scans. The lesion areas were delineated, and a total of 1,029 radiomics features were extracted. The least absolute shrinkage and selection operator (Lasso) algorithm method was used to select the radiomics features significantly associated with discrimination of high-risk from low-risk lesions in the anterior mediastinum. Then, 8-fold and 3-fold cross-validation logistic regression (LR) models were taken as the feature selection classifiers to build the radiomics models for UECT and CECT scan respectively. The predictive performance of the radiomics features was evaluated based on the receiver operating characteristics (ROC) curve. Each of the two radiomics classifiers included the optimal 12 radiomic features. In terms of the area under ROC curve, using the radiomics model in discriminating high-risk lesions from the low-risks, CECT images accounted for 74.1% with a sensitivity of 66.67% and specificity of 64.81%. Meanwhile, UECT images were 84.2% with a sensitivity of 71.43% and specificity of 74.07%. The association of the two proposed CT-based radiomics features with the discrimination of high and low-risk lesions in anterior mediastinum was confirmed, and the radiomics features of the UECT scan were proven to have better prediction performance than the CECT's in risk grading.

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.

Purpose: Thyroid 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.Methods: This 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.Results: The 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.Conclusions: Our 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.

The infant brain undergoes a remarkable period of neural development that is crucial for the development of cognitive and behavioral capacities (Hasegawa et al., 2018). Longitudinal magnetic resonance imaging (MRI) is able to characterize the developmental trajectories and is critical in neuroimaging studies of early brain development. However, missing data at different time points is an unavoidable occurrence in longitudinal studies owing to participant attrition and scan failure. Compared to dropping incomplete data, data imputation is considered a better solution to address such missing data in order to preserve all available samples. In this paper, we adapt generative adversarial networks (GAN) to a new application: longitudinal image prediction of structural MRI in the first year of life. In contrast to existing medical image-to-image translation applications of GANs, where inputs and outputs share a very close anatomical structure, our task is more challenging as brain size, shape and tissue contrast vary significantly between the input data and the predicted data. Several improvements over existing GAN approaches are proposed to address these challenges in our task. To enhance the realism, crispness, and accuracy of the predicted images, we incorporate both a traditional voxel-wise reconstruction loss as well as a perceptual loss term into the adversarial learning scheme. As the differing contrast changes in T1w and T2w MR images in the first year of life, we incorporate multi-contrast images leading to our proposed 3D multi-contrast perceptual adversarial network (MPGAN). Extensive evaluations are performed to assess the qualityand fidelity of the predicted images, including qualitative and quantitative assessments of the image appearance, as well as quantitative assessment on two segmentation tasks. Our experimental results show that our MPGAN is an effective solution for longitudinal MR image data imputation in the infant brain. We further apply our predicted/imputed images to two practical tasks, a regression task and a classification task, in order to highlight the enhanced task-related performance following image imputation. The results show that the model performance in both tasks is improved by including the additional imputed data, demonstrating the usability of the predicted images generated from our approach.