Search

While deep neural networks (DNNs) and other machine learning models often have higher accuracy than simpler models like logistic regression (LR), they are often considered to be "black box" models and this lack of interpretability and transparency is considered a challenge for clinical adoption. In healthcare, intelligible models not only help clinicians to understand the problem and create more targeted action plans, but also help to gain the clinicians' trust. One method of overcoming the limited interpretability of more complex models is to use Generalized Additive Models (GAMs). Standard GAMs simply model the target response as a sum of univariate models. Inspired by GAMs, the same idea can be applied to neural networks through an architecture referred to as Generalized Additive Models with Neural Networks (GAM-NNs). In this manuscript, we present the development and validation of a model applying the concept of GAM-NNs to allow for interpretability by visualizing the learned feature patterns related to risk of in-hospital mortality for patients undergoing surgery under general anesthesia. The data consists of 59,985 patients with a feature set of 46 features extracted at the end of surgery to which we added previously not included features: total anesthesia case time (1 feature); the time in minutes spent with mean arterial pressure (MAP) below 40, 45, 50, 55, 60, and 65 mmHg during surgery (6 features); and Healthcare Cost and Utilization Project (HCUP) Code Descriptions of the Primary current procedure terminology (CPT) codes (33 features) for a total of 86 features. All data were randomly split into 80% for training (n = 47,988) and 20% for testing (n = 11,997) prior to model development. Model performance was compared to a standard LR model using the same features as the GAM-NN. The data consisted of 59,985 surgical records, and the occurrence of in-hospital mortality was 0.81% in the training set and 0.72% in the testing set. The GAM-NN model with HCUP features had the highest area under the curve (AUC) 0.921 (0.895-0.95). Overall, both GAM-NN models had higher AUCs than LR models, however, had lower average precisions. The LR model without HCUP features had the highest average precision 0.217 (0.136-0.31). To assess the interpretability of the GAM-NNs, we then visualized the learned contributions of the GAM-NNs and compared against the learned contributions of the LRs for the models with HCUP features. Overall, we were able to demonstrate that our proposed generalized additive neural network (GAM-NN) architecture is able to (1) leverage a neural network's ability to learn nonlinear patterns in the data, which is more clinically intuitive, (2) be interpreted easily, making it more clinically useful, and (3) maintain model performance as compared to previously published DNNs.

Prediction models of post-liver transplant mortality are crucial so that donor organs are not allocated to recipients with unreasonably high probabilities of mortality. Machine learning algorithms, particularly deep neural networks (DNNs), can often achieve higher predictive performance than conventional models. In this study, we trained a DNN to predict 90-day post-transplant mortality using preoperative variables and compared the performance to that of the Survival Outcomes Following Liver Transplantation (SOFT) and Balance of Risk (BAR) scores, using United Network of Organ Sharing data on adult patients who received a deceased donor liver transplant between 2005 and 2015 (n = 57,544). The DNN was trained using 202 features, and the best DNN's architecture consisted of 5 hidden layers with 110 neurons each. The area under the receiver operating characteristics curve (AUC) of the best DNN model was 0.703 (95% CI: 0.682-0.726) as compared to 0.655 (95% CI: 0.633-0.678) and 0.688 (95% CI: 0.667-0.711) for the BAR score and SOFT score, respectively. In conclusion, despite the complexity of DNN, it did not achieve a significantly higher discriminative performance than the SOFT score. Future risk models will likely benefit from the inclusion of other data sources, including high-resolution clinical features for which DNNs are particularly apt to outperform conventional statistical methods.

What we already know about this topicWHAT THIS ARTICLE TELLS US THAT IS NEW: BACKGROUND:: The authors tested the hypothesis that deep neural networks trained on intraoperative features can predict postoperative in-hospital mortality.MethodsThe data used to train and validate the algorithm consists of 59,985 patients with 87 features extracted at the end of surgery. Feed-forward networks with a logistic output were trained using stochastic gradient descent with momentum. The deep neural networks were trained on 80% of the data, with 20% reserved for testing. The authors assessed improvement of the deep neural network by adding American Society of Anesthesiologists (ASA) Physical Status Classification and robustness of the deep neural network to a reduced feature set. The networks were then compared to ASA Physical Status, logistic regression, and other published clinical scores including the Surgical Apgar, Preoperative Score to Predict Postoperative Mortality, Risk Quantification Index, and the Risk Stratification Index.ResultsIn-hospital mortality in the training and test sets were 0.81% and 0.73%. The deep neural network with a reduced feature set and ASA Physical Status classification had the highest area under the receiver operating characteristics curve, 0.91 (95% CI, 0.88 to 0.93). The highest logistic regression area under the curve was found with a reduced feature set and ASA Physical Status (0.90, 95% CI, 0.87 to 0.93). The Risk Stratification Index had the highest area under the receiver operating characteristics curve, at 0.97 (95% CI, 0.94 to 0.99).ConclusionsDeep neural networks can predict in-hospital mortality based on automatically extractable intraoperative data, but are not (yet) superior to existing methods.

Summary: Background: Data on oral vancomycin for primary sclerosing cholangitis (PSC)‐associated inflammatory bowel disease (IBD) are limited. Aims: Using data from the Paediatric PSC Consortium, to examine the effect of vancomycin on IBD activity. Methods: In this retrospective multi‐centre cohort study, we matched vancomycin‐treated and untreated patients (1:3) based on IBD duration at the time of primary outcome assessment. The primary outcome was Physician Global Assessment (PGA) of IBD clinical activity after 1 year (±6 months) of vancomycin. We used generalised estimating equations (GEE) to examine the association between vancomycin and PGA remission, adjusting for IBD type, severity and medication exposures. Secondary outcomes included serum labs and endoscopic remission (global rating of no activity) among those with available data and also analysed with GEE. Results: 113 PSC‐IBD patients received vancomycin (median age 12.7 years, 63% male). The matched cohort included 70 vancomycin‐treated and 210 untreated patients. Vancomycin was associated with greater odds of IBD clinical remission (odds ratio [OR] 3.52, 95% CI 1.97–6.31; adjusted OR [aOR] 5.24, 95% CI 2.68–10.22). Benefit was maintained in sensitivity analyses restricted to non‐transplanted patients and those with baseline moderate–severe PGA. Vancomycin was associated with increased odds of endoscopic remission (aOR 2.76, 95% CI 1.002–7.62; N = 101 with data), and with lower CRP (p = 0.03) and higher haemoglobin and albumin (both p < 0.01). Conclusion: Vancomycin was associated with greater odds of IBD clinical and endoscopic remission. Additional, preferably randomised, controlled studies are needed to characterise efficacy using objective markers of mucosal inflammation, and to examine safety and define optimal dosing. [ABSTRACT FROM AUTHOR]

Nonalcoholic fatty liver disease (NAFLD) is the most common chronic liver disease worldwide. It represents a spectrum of disease from simple hepatic steatosis to steatohepatitis that may develop into progressive hepatic fibrosis and even cirrhosis. NAFLD is the most rapidly increasing indication for liver transplantation in adults. In children, the incidence of NAFLD has also increased over the past decade. Although the majority of children with NAFLD are overweight or obese, there is an increasing subset of children with normal body mass index with so-called lean NAFLD. NAFLD in children is associated with several extrahepatic manifestations, including hyperlipidemia, insulin resistance, and obstructive sleep apnea. The pathogenesis of NAFLD in children involves a multifactorial interaction among genetics, in utero exposures, early childhood exposures, and ongoing nutritional exposures. Although there are some similarities between pediatric NAFLD and adult NAFLD, liver biopsies in children show histologic differences between the two. The current standard-of-care treatment of NAFLD in children is lifestyle change to decrease caloric intake and increase physical activity. There are no medications currently approved for the treatment of NAFLD in children. This article aims to summarize the current understanding of pediatric NAFLD and future directions for intervention and therapeutic aims. [ABSTRACT FROM AUTHOR]

Perianeurysmal cysts are a rare and poorly understood finding in patients both with treated and untreated aneurysms. While the prior literature suggests that a minority of perianeurysmal cysts develop 1-4 years following endovascular aneurysm treatment, this updated review demonstrates that nearly half of perianeurysmal cysts were diagnosed following aneurysm coiling, with the other half diagnosed concurrently with an associated aneurysm prior to treatment. 64% of perianeurysmal cysts were surgically decompressed, with a 39% rate of recurrence requiring re-operation. We report a case of a 71-year-old woman who presented with vertigo and nausea and was found to have a 3.4 cm perianeurysmal cyst 20 years after initial endovascular coiling of a ruptured giant ophthalmic aneurysm. The cyst was treated with endoscopic fenestration followed by open fenestration upon recurrence. The case represents the longest latency from initial aneurysm treatment to cyst diagnosis reported in the literature and indicates that the diagnosis of perianeurysmal cyst should remain on the differential even decades after treatment. Based on a case discussion and updated literature review, this report highlights proposed etiologies of development and management strategies for a challenging lesion.

The aggressive primary brain tumor glioblastoma (GBM) is characterized by aberrant metabolism that fuels its malignant phenotype. Diverse genetic subtypes of malignant glioma are sensitive to selective inhibition of the NAD + salvage pathway enzyme nicotinamide phosphoribosyltransferase (NAMPT). However, the potential impact of NAD + depletion on the brain tumor microenvironment has not been elaborated. In addition, systemic toxicity of NAMPT inhibition remains a significant concern. Here we show that microparticle-mediated intratumoral delivery of NAMPT inhibitor GMX1778 induces specific immunologic changes in the tumor microenvironment of murine GBM, characterized by upregulation of immune checkpoint PD-L1, recruitment of CD3 + , CD4 + , and CD8 + T cells, and reduction of M2-polarized immunosuppressive macrophages. NAD + depletion and autophagy induced by NAMPT inhibitors mediated the upregulation of PD-L1 transcripts and cell surface protein levels in GBM cells. NAMPT inhibitor modulation of the tumor immune microenvironment was therefore combined with PD-1 checkpoint blockade in vivo , significantly increasing the survival of GBM-bearing animals. Thus, the therapeutic impacts of NAMPT inhibition extended beyond neoplastic cells, shaping surrounding immune effectors. Microparticle delivery and release of NAMPT inhibitor at the tumor site offers a safe and robust means to alter an immune tumor microenvironment that could potentiate checkpoint immunotherapy for glioblastoma. SIGNIFICANCE: Microparticle-mediated local inhibition of NAMPT modulates the tumor immune microenvironment and acts cooperatively with anti-PD-1 checkpoint blockade, offering a combination immunotherapy strategy for the treatment of GBM., (©2020 American Association for Cancer Research.)