Towards computer-aided severity assessment via deep neural networks for geographic and opacity extent scoring of SARS-CoV-2 chest X-rays

A critical step in effective care and treatment planning for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the cause for the coronavirus disease 2019 (COVID-19) pandemic, is the assessment of the severity of disease progression. Chest x-rays (CXRs) are often used to assess SARS-CoV-2 severity, with two important assessment metrics being extent of lung involvement and degree of opacity. In this proof-of-concept study, we assess the feasibility of computer-aided scoring of CXRs of SARS-CoV-2 lung disease severity using a deep learning system. Data consisted of 396 CXRs from SARS-CoV-2 positive patient cases. Geographic extent and opacity extent were scored by two board-certified expert chest radiologists (with 20+ years of experience) and a 2nd-year radiology resident. The deep neural networks used in this study, which we name COVID-Net S, are based on a COVID-Net network architecture. 100 versions of the network were independently learned (50 to perform geographic extent scoring and 50 to perform opacity extent scoring) using random subsets of CXRs from the study, and we evaluated the networks using stratified Monte Carlo cross-validation experiments. The COVID-Net S deep neural networks yielded R$$^2$$ 2 of $$0.664 \pm 0.032$$ 0.664 ± 0.032 and $$0.635 \pm 0.044$$ 0.635 ± 0.044 between predicted scores and radiologist scores for geographic extent and opacity extent, respectively, in stratified Monte Carlo cross-validation experiments. The best performing COVID-Net S networks achieved R$$^2$$ 2 of 0.739 and 0.741 between predicted scores and radiologist scores for geographic extent and opacity extent, respectively. The results are promising and suggest that the use of deep neural networks on CXRs could be an effective tool for computer-aided assessment of SARS-CoV-2 lung disease severity, although additional studies are needed before adoption for routine clinical use.

Statistical and Machine Learning Models for Classification of Human Wear and Delivery Days in Accelerometry Data

Accelerometers are increasingly being used in biomedical research, but the analysis of accelerometry data is often complicated by both the massive size of the datasets and the collection of unwanted data from the process of delivery to study participants. Current methods for removing delivery data involve arduous manual review of dense datasets. We aimed to develop models for the classification of days in accelerometry data as activity from human wear or the delivery process. These models can be used to automate the cleaning of accelerometry datasets that are adulterated with activity from delivery. We developed statistical and machine learning models for the classification of accelerometry data in a supervised learning context using a large human activity and delivery labeled accelerometry dataset. Model performances were assessed and compared using Monte Carlo cross-validation. We found that a hybrid convolutional recurrent neural network performed best in the classification task with an F1 score of 0.960 but simpler models such as logistic regression and random forest also had excellent performance with F1 scores of 0.951 and 0.957, respectively. The best performing models and related data processing techniques are made publicly available in the R package, Physical Activity.

Artificial intelligence prediction of the effect of rehabilitation in whiplash associated disorder

The active cervical range of motion (aROM) is assessed by clinicians to inform their decision-making. Even with the ability of neck motion to discriminate injured from non-injured subjects, the mechanisms to explain recovery or persistence of WAD remain unclear. There are few studies of ROM examinations with precision tools using kinematics as predictive factors of recovery rate. The present paper will evaluate the performance of an artificial neural network (ANN) using kinematic variables to predict the overall change of aROM after a period of rehabilitation in WAD patients. To achieve this goal the neck kinematics of a cohort of 1082 WAD patients (55.1% females), with mean age 37.68 (SD 12.88) years old, from across Spain were used. Prediction variables were the kinematics recorded by the EBI® 5 in routine biomechanical assessments of these patients. These include normalized ROM, speed to peak and ROM coefficient of variation. The improvement of aROM was represented by the Neck Functional Holistic Analysis Score (NFHAS). A supervised multi-layer feed-forward ANN was created to predict the change in NFHAS. The selected architecture of the ANN showed a mean squared error of 308.07–272.75 confidence interval for a 95% in the Monte Carlo cross validation. The performance of the ANN was tested with a subsample of patients not used in the training. This comparison resulted in a medium correlation with R = 0.5. The trained neural network to predict the expected difference in NFHAS between baseline and follow up showed modest results. While the overall performance is moderately correlated, the error of this prediction is still too large to use the method in clinical practice. The addition of other clinically relevant factors could further improve prediction performance.

662 Statistical learning from clinical and immunogenomic variables to predict response and survival with PD-L1 inhibition in advanced urothelial cancer

BackgroundUrothelial cancer patients treated with immune checkpoint inhibitor (ICI) therapy have varied response and survival.1 Clinical and immunogenomic biomarkers could help predict ICI response and survival to inform decisions about patient selection for ICI treatment.MethodsThe association of clinical metadata and immunogenomic signatures with response and survival was analyzed in a set of 347 urothelial cancer patients treated with the PD-L1 inhibitor atezolizumab as part of the IMVigor210 study.1 Data were divided into a discovery set (2/3 of patients) and validation set (1/3 of patients). We analyzed as potential predictors 70 total variables, of which 16 were clinical metadata and 54 were immunogenomic signatures. Categorical variables were converted to dummy variables (89 total variables: 35 clinical, 54 immunogenomic). Using the discovery set, elastic net regression with Monte Carlo cross-validation was used to build optimal models for response (logistic regression) and survival (Cox proportional-hazards). Model performance was evaluated using the validation set.ResultsIn the optimal model of response, 17 variables (10 clinical, 7 immunogenomic) were selected as informative predictors, including Baseline Eastern Cooperative Oncology Group (ECOG) Score = 0, Neoantigen Burden, Lymph Node Metastases, and Tumor Mutation Burden (figure 1). The final model predicted patient response with good performance (Area Under Curve = 0.828, pAUC = 2.38e-3; True Negative Rate = 91.7%, True Positive Rate = 87.5%, pconfusion matrix = 0.0252). In the optimal model of survival, 32 variables (17 clinical, 15 immunogenomic) were selected as informative predictors, including baseline ECOG Score = 0, IC Level 2+, Race = Asian, and Consensus Tumor Subtype = Neuroendocrine (figure 2). The final model predicted patient survival with good performance (c-indexmodel = 0.652, pc-index = 0.0290).Abstract 662 Figure 1Elastic Net Logistic Regression with Monte Carlo Cross-Validation to Predict Response to Atezolizumab in Urothelial Cancer. (A) Predictive variables with beta coefficient 95% confidence intervals that exclude 0, derived from Monte Carlo cross-validation. (B) Confusion matrix of actual vs. predicted response data in the validation set. (C) Total response proportions of actual and predicted response data in the validation setAbstract 662 Figure 2Elastic Net Cox Proportional-Hazards Regression with Monte Carlo Cross-Validation to Predict Survival. (A) Predictor variables with beta coefficient 95% confidence intervals that exclude 0, derived from Monte Carlo cross-validation. (B) Predictions vs. survival outcomes in the validation set. (C) Loess models of density curves for survival outcomes in the validation set. 95% confidence intervals were generated through bootstrapping with replacement. (D) Loess fit of predictions vs. survival outcomes in the validation set. 95% confidence interval indicates strength of fitConclusionsModels incorporating clinical metadata and immunogenomic signatures can predict response and survival for urothelial cancer patients treated with atezolizumab. Among predictors in those models, baseline performance status is the greatest and most positive predictor of response and survival.ReferenceMariathasan S, Turley S, Nickles D, et al. TGFβ attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells. Nature 2018;554:544–548.

Deep learning based detection of intracranial aneurysms on digital subtraction angiography: A feasibility study

Background Digital subtraction angiography is the gold standard for detecting and characterising aneurysms. Here, we assess the feasibility of commercial-grade deep learning software for the detection of intracranial aneurysms on whole-brain anteroposterior and lateral 2D digital subtraction angiography images. Material and methods Seven hundred and six digital subtraction angiography images were included from a cohort of 240 patients (157 female, mean age 59 years, range 20–92; 83 male, mean age 55 years, range 19–83). Three hundred and thirty-five (47%) single frame anteroposterior and lateral images of a digital subtraction angiography series of 187 aneurysms (41 ruptured, 146 unruptured; average size 7±5.3 mm, range 1–5 mm; total 372 depicted aneurysms) and 371 (53%) aneurysm-negative study images were retrospectively analysed regarding the presence of intracranial aneurysms. The 2D data was split into testing and training sets in a ratio of 4:1 with 3D rotational digital subtraction angiography as gold standard. Supervised deep learning was performed using commercial-grade machine learning software (Cognex, ViDi Suite 2.0). Monte Carlo cross validation was performed. Results Intracranial aneurysms were detected with a sensitivity of 79%, a specificity of 79%, a precision of 0.75, a F1 score of 0.77, and a mean area-under-the-curve of 0.76 (range 0.68–0.86) after Monte Carlo cross-validation, run 45 times. Conclusion The commercial-grade deep learning software allows for detection of intracranial aneurysms on whole-brain, 2D anteroposterior and lateral digital subtraction angiography images, with results being comparable to more specifically engineered deep learning techniques.

Medium-Term Rainfall Forecasts Using Artificial Neural Networks with Monte-Carlo Cross-Validation and Aggregation for the Han River Basin, Korea

In this study, artificial neural network (ANN) models were constructed to predict the rainfall during May and June for the Han River basin, South Korea. This was achieved using the lagged global climate indices and historical rainfall data. Monte-Carlo cross-validation and aggregation (MCCVA) was applied to create an ensemble of forecasts. The input-output patterns were randomly divided into training, validation, and test datasets. This was done 100 times to achieve diverse data splitting. In each data splitting, ANN training was repeated 100 times using randomly assigned initial weight vectors of the network to construct 10,000 prediction ensembles and estimate their prediction uncertainty interval. The optimal ANN model that was used to forecast the monthly rainfall in May had 11 input variables of the lagged climate indices such as the Arctic Oscillation (AO), East Atlantic/Western Russia Pattern (EAWR), Polar/Eurasia Pattern (POL), Quasi-Biennial Oscillation (QBO), Sahel Precipitation Index (SPI), and Western Pacific Index (WP). The ensemble of the rainfall forecasts exhibited the values of the averaged root mean squared error (RMSE) of 27.4, 33.6, and 39.5 mm, and the averaged correlation coefficient (CC) of 0.809, 0.725, and 0.641 for the training, validation, and test sets, respectively. The estimated uncertainty band has covered 58.5% of observed rainfall data with an average band width of 50.0 mm, exhibiting acceptable results. The ANN forecasting model for June has 9 input variables, which differed from May, of the Atlantic Meridional Mode (AMM), East Pacific/North Pacific Oscillation (EPNP), North Atlantic Oscillation (NAO), Scandinavia Pattern (SCAND), Equatorial Eastern Pacific SLP (SLP_EEP), and POL. The averaged RMSE values are 39.5, 46.1, and 62.1 mm, and the averaged CC values are 0.853, 0.771, and 0.683 for the training, validation, and test sets, respectively. The estimated uncertainty band for June rainfall forecasts generally has a coverage of 67.9% with an average band width of 83.0 mm. It can be concluded that the neural network with MCCVA enables us to provide acceptable medium-term rainfall forecasts and define the prediction uncertainty interval.

Convolutional ensembles for Arabic Handwritten Character and Digit Recognition

A learning algorithm is proposed for the task of Arabic Handwritten Character and Digit recognition. The architecture consists on an ensemble of different Convolutional Neural Networks. The proposed training algorithm uses a combination of adaptive gradient descent on the first epochs and regular stochastic gradient descent in the last epochs, to facilitate convergence. Different validation strategies are tested, namely Monte Carlo Cross-Validation and K-fold Cross Validation. Hyper-parameter tuning was done by using the MADbase digits dataset. State of the art validation and testing classification accuracies were achieved, with average values of 99.74% and 99.47% respectively. The same algorithm was then trained and tested with the AHCD character dataset, also yielding state of the art validation and testing classification accuracies: 98.60% and 98.42% respectively.