Search for: All records

Background: Type 1 diabetes (T1D) is a devastating autoimmune disease, and its rising prevalence in the United States and around the world presents a critical problem in public health. While some treatment options exist for patients already diagnosed, individuals considered at risk for developing T1D and who are still in the early stages of their disease pathogenesis without symptoms have no options for any preventive intervention. This is because of the uncertainty in determining their risk level and in predicting with high confidence who will progress, or not, to clinical diagnosis. Biomarkers that assess one’s risk with high certainty could address this problem and will inform decisions on early intervention, especially in children where the burden of justifying treatment is high. Single omics approaches (e.g., genomics, proteomics, metabolomics, etc.) have been applied to identify T1D biomarkers based on specific disturbances in association with the disease. However, reliable early biomarkers of T1D have remained elusive to date. To overcome this, we previously showed that parallel multi-omics provides a more comprehensive picture of the disease-associated disturbances and facilitates the identification of candidate T1D biomarkers. Methods: This paper evaluated the use of machine learning (ML) using data augmentation and supervised ML methods for the purpose of improving the identification of salient patterns in the data and the ultimate extraction of novel biomarker candidates in integrated parallel multi-omics datasets from a limited number of samples. We also examined different stages of data integration (early, intermediate, and late) to assess at which stage supervised parametric models can learn under conditions of high dimensionality and variation in feature counts across different omics. In the late integration scheme, we employed a multi-view ensemble comprising individual parametric models trained over single omics to address the computational challenges posed by the high dimensionality and variation in feature counts across the different yet integrated multi-omics datasets. Results: the multi-view ensemble improves the prediction of case vs. control and finds the most success in flagging a larger consistent set of associated features when compared with chance models, which may eventually be used downstream in identifying a novel composite biomarker signature of T1D risk. Conclusions: the current work demonstrates the utility of supervised ML in exploring integrated parallel multi-omics data in the ongoing quest for early T1D biomarkers, reinforcing the hope for identifying novel composite biomarker signatures of T1D risk via ML and ultimately informing early treatment decisions in the face of the escalating global incidence of this debilitating disease. 

Biomarkers of early pathogenesis of type 1 diabetes (T1D) are crucial to enable effective prevention measures in at-risk populations before significant damage occurs to their insulin producing beta-cell mass. We recently introduced the concept of integrated parallel multi-omics and employed a novel data augmentation approach which identified promising candidate biomarkers from a small cohort of high-risk T1D subjects. We now validate selected biomarkers to generate a potential composite signature of T1D risk. 

We present a novel algorithm that is able to generate deep synthetic COVID-19 pneumonia CT scan slices using a very small sample of positive training images in tandem with a larger number of normal images. This generative algorithm produces images of sufficient accuracy to enable a DNN classifier to achieve high classification accuracy using as few as 10 positive training slices (from 10 positive cases), which to the best of our knowledge is one order of magnitude fewer than the next closest published work at the time of writing. Deep learning with extremely small positive training volumes is a very difficult problem and has been an important topic during the COVID-19 pandemic, because for quite some time it was difficult to obtain large volumes of COVID-19-positive images for training. Algorithms that can learn to screen for diseases using few examples are an important area of research. Furthermore, algorithms to produce deep synthetic images with smaller data volumes have the added benefit of reducing the barriers of data sharing between healthcare institutions. We present the cycle-consistent segmentation-generative adversarial network (CCS-GAN). CCS-GAN combines style transfer with pulmonary segmentation and relevant transfer learning from negative images in order to create a larger volume of synthetic positive images for the purposes of improving diagnostic classification performance. The performance of a VGG-19 classifier plus CCS-GAN was trained using a small sample of positive image slices ranging from at most 50 down to as few as 10 COVID-19-positive CT scan images. CCS-GAN achieves high accuracy with few positive images and thereby greatly reduces the barrier of acquiring large training volumes in order to train a diagnostic classifier for COVID-19. 

We introduce an active, semisupervised algorithm that utilizes Bayesian experimental design to address the shortage of annotated images required to train and validate Artificial Intelligence (AI) models for lung cancer screening with computed tomography (CT) scans. Our approach incorporates active learning with semisupervised expectation maximization to emulate the human in the loop for additional ground truth labels to train, evaluate, and update the neural network models. Bayesian experimental design is used to intelligently identify which unlabeled samples need ground truth labels to enhance the model’s performance. We evaluate the proposed Active Semi-supervised Expectation Maximization for Computer aided diagnosis (CAD) tasks (ASEM-CAD) using three public CT scans datasets: the National Lung Screening Trial (NLST), the Lung Image Database Consortium (LIDC), and Kaggle Data Science Bowl 2017 for lung cancer classification using CT scans. ASEM-CAD can accurately classify suspicious lung nodules and lung cancer cases with an area under the curve (AUC) of 0.94 (Kaggle), 0.95 (NLST), and 0.88 (LIDC) with significantly fewer labeled images compared to a fully supervised model. This study addresses one of the significant challenges in early lung cancer screenings using low-dose computed tomography (LDCT) scans and is a valuable contribution towards the development and validation of deep learning algorithms for lung cancer screening and other diagnostic radiology examinations. 

Venous thromboembolism (VTE) is a preventable complication of hospitalization. VTE risk-assessment models (RAMs) including the Caprini and Padua RAMs quantify VTE risk based on demographic and clinical characteristics. Both RAMs have performed well in selected high-risk cohorts with relatively small sample sizes but few studies have evaluated the RAMs in large, unselected cohorts. We assessed the ability of both RAMs to predict VTE in a large, nationwide, diverse cohort of surgical and nonsurgical patients. 

Background: Type 1 diabetes (T1D) is a devastating disease with serious health complications. Early T1D biomarkers that could enable timely detection and prevention before the onset of clinical symptoms are paramount but currently unavailable. Despite their promise, omics approaches have so far failed to deliver such biomarkers, likely due to the fragmented nature of information obtained through the single omics approach. We recently demonstrated the utility of parallel multi-omics for the identification of T1D biomarker signatures. Our studies also identified challenges. Methods: Here, we evaluated a novel computational approach of data imputation and amplification as one way to overcome challenges associated with the relatively small number of subjects in these studies. Results: Using proprietary algorithms, we amplified our quadra-omics (proteomics, metabolomics, lipidomics, and transcriptomics) dataset from nine subjects a thousand-fold and analyzed the data using Ingenuity Pathway Analysis (IPA) software to assess the change in its analytical capabilities and biomarker prediction power in the amplified datasets compared to the original. These studies showed the ability to identify an increased number of T1D-relevant pathways and biomarkers in such computationally amplified datasets, especially, at imputation ratios close to the “golden ratio” of 38.2%:61.8%. Specifically, the Canonical Pathway and Diseases and Functions modules identified higher numbers of inflammatory pathways and functions relevant to autoimmune T1D, including novel ones not identified in the original data. The Biomarker Prediction module also predicted in the amplified data several unique biomarker candidates with direct links to T1D pathogenesis. Conclusions: These preliminary findings indicate that such large-scale data imputation and amplification approaches are useful in facilitating the discovery of candidate integrated biomarker signatures of T1D or other diseases by increasing the predictive range of existing data mining tools, especially when the size of the input data is inherently limited. 

In this work, we formulate and solve a new type of approximate nearest neighbor search (ANNS) problems called ANNS after linear transformation (ALT). In ANNS-ALT, we search for the vector (in a dataset) that, after being linearly transformed by a user-specified query matrix, is closest to a query vector. It is a very general mother problem in the sense that a wide range of baby ANNS problems that have important applications in databases and machine learning can be reduced to and solved as ANNS-ALT, or its dual that we call ANNS-ALTD. We propose a novel and computationally efficient solution, called ONe Index for All Kernels (ONIAK), to ANNS-ALT and all its baby problems when the data dimension d is not too large (say d ≤ 200). In ONIAK, a universal index is built, once and for all, for answering all future ANNS-ALT queries that can have distinct query matrices. We show by experiments that, when d is not too large, ONIAK has better query performance than linear scan on the mother problem (of ANNS-ALT), and has query performances comparable to those of the state-of-the-art solutions on the baby problems. However, the algorithmic technique behind this universal index approach suffers from a so-called dimension blowup problem that can make the indexing time prohibitively long for a large dataset. We propose a novel algorithmic technique, called fast GOE quadratic form (FGoeQF), that completely solves the (prohibitively long indexing time) fallout of the dimension blowup problem. We also propose a Johnson-Lindenstrauss transform (JLT) based ANNS-ALT (and ANNS-ALTD) solution that significantly outperforms any competitor when d is large. 

Multi-site clinical trial systems face security challenges when streamlining information sharing while protecting patient privacy. In addition, patient enrollment, transparency, traceability, data integrity, and reporting in clinical trial systems are all critical aspects of maintaining data compliance. A Blockchain-based clinical trial framework has been proposed by lots of researchers and industrial companies recently, but its limitations of lack of data governance, limited confidentiality, and high communication overhead made data-sharing systems insecure and not efficient. We propose 𝖲𝗈𝗍𝖾𝗋𝗂𝖺, a privacy-preserving smart contracts framework, to manage, share and analyze clinical trial data on fabric private chaincode (FPC). Compared to public Blockchain, fabric has fewer participants with an efficient consensus protocol. 𝖲𝗈𝗍𝖾𝗋𝗂𝖺 consists of several modules: patient consent and clinical trial approval management chaincode, secure execution for confidential data sharing, API Gateway, and decentralized data governance with adaptive threshold signature (ATS). We implemented two versions of 𝖲𝗈𝗍𝖾𝗋𝗂𝖺 with non-SGX deploys on AWS blockchain and SGX-based on a local data center. We evaluated the response time for all of the access endpoints on AWS Managed Blockchain, and demonstrated the utilization of SGX-based smart contracts for data sharing and analysis.