Deep learning and alternative learning strategies for retrospective real-world clinical data.
Ontology highlight
ABSTRACT:
SUBMITTER: Chen D
PROVIDER: S-EPMC6550223 | biostudies-literature | 2019
REPOSITORIES: biostudies-literature
Publications
Deep learning and alternative learning strategies for retrospective real-world clinical data.
Chen David D Liu Sijia S Kingsbury Paul P Sohn Sunghwan S Storlie Curtis B CB Habermann Elizabeth B EB Naessens James M JM Larson David W DW Liu Hongfang H
NPJ digital medicine 20190530
In recent years, there is increasing enthusiasm in the healthcare research community for artificial intelligence to provide big data analytics and augment decision making. One of the prime reasons for this is the enormous impact of deep learning for utilization of complex healthcare big data. Although deep learning is a powerful analytic tool for the complex data contained in electronic health records (EHRs), there are also limitations which can make the choice of deep learning inferior in some ...[more]
Similar Datasets
Project description:Drug repurposing is an effective strategy to identify new uses for existing drugs, providing the quickest possible transition from bench to bedside. Real-world data, such as electronic health records and insurance claims, provide information on large cohorts of users for many drugs. Here we present an efficient and easily customized framework for generating and testing multiple candidates for drug repurposing using a retrospective analysis of real-world data. Building upon well-established causal inference and deep learning methods, our framework emulates randomized clinical trials for drugs present in a large-scale medical claims database. We demonstrate our framework on a coronary artery disease cohort of millions of patients. We successfully identify drugs and drug combinations that substantially improve the coronary artery disease outcomes but haven't been indicated for treating coronary artery disease, paving the way for drug repurposing.
| S-EPMC9119409 | biostudies-literature
Project description:Artificial intelligence (AI) based on deep learning has shown excellent diagnostic performance in detecting various diseases with good-quality clinical images. Recently, AI diagnostic systems developed from ultra-widefield fundus (UWF) images have become popular standard-of-care tools in screening for ocular fundus diseases. However, in real-world settings, these systems must base their diagnoses on images with uncontrolled quality ("passive feeding"), leading to uncertainty about their performance. Here, using 40,562 UWF images, we develop a deep learning-based image filtering system (DLIFS) for detecting and filtering out poor-quality images in an automated fashion such that only good-quality images are transferred to the subsequent AI diagnostic system ("selective eating"). In three independent datasets from different clinical institutions, the DLIFS performed well with sensitivities of 96.9%, 95.6% and 96.6%, and specificities of 96.6%, 97.9% and 98.8%, respectively. Furthermore, we show that the application of our DLIFS significantly improves the performance of established AI diagnostic systems in real-world settings. Our work demonstrates that "selective eating" of real-world data is necessary and needs to be considered in the development of image-based AI systems.
| S-EPMC7603327 | biostudies-literature
Project description:ObjectivesTo determine the prevalence of atrial fibrillation (AF) and to assess how these patients are being cared for: what anticoagulants are being prescribed and are they being prescribed as recommended?DesignRetrospective longitudinal study.SettingThis study was conducted in the Regional Health Administration of Northern Portugal.ParticipantsThis study used a database that included 63526 patients with code K78 of the International Classification of Primary Care between January 2016 and December 2018.ResultsThe prevalence of AF among adults over 40 years in the northern region of Portugal was 2.3% in 2016, 2.8% in 2017 and 3% in 2018. From a total of 63 526 patients, 95.8% had an indication to receive anticoagulation therapy. Of these, 44 326 (72.9%) are being treated with anticoagulants: 17 936 (40.5%) were prescribed vitamin K antagonists (VKAs) and 26 390 (59.5%) were prescribed non-VKA anticoagulants. On the other hand, 2688 patients of the total (4.2%) had no indication to receive anticoagulation therapy. Of these 2688 patients, 1100 (40.9%) were receiving anticoagulants.ConclusionsThe prevalence of AF is 3%. Here, we report evidence of both undertreatment and overtreatment. Although having an indication, a considerable proportion of patients (27.1%) are not anticoagulated, and among patients with AF without an indication to receive anticoagulation therapy, a considerable proportion (40.9%) are receiving anticoagulants. The AF-React study brings extremely relevant conclusions to Portugal and follows real-world studies in patients with AF in Europe, presenting some data not yet studied.
| S-EPMC8009225 | biostudies-literature
Project description:BackgroundSuccessful integrations of machine learning into routine clinical care are exceedingly rare, and barriers to its adoption are poorly characterized in the literature.ObjectiveThis study aims to report a quality improvement effort to integrate a deep learning sepsis detection and management platform, Sepsis Watch, into routine clinical care.MethodsIn 2016, a multidisciplinary team consisting of statisticians, data scientists, data engineers, and clinicians was assembled by the leadership of an academic health system to radically improve the detection and treatment of sepsis. This report of the quality improvement effort follows the learning health system framework to describe the problem assessment, design, development, implementation, and evaluation plan of Sepsis Watch.ResultsSepsis Watch was successfully integrated into routine clinical care and reshaped how local machine learning projects are executed. Frontline clinical staff were highly engaged in the design and development of the workflow, machine learning model, and application. Novel machine learning methods were developed to detect sepsis early, and implementation of the model required robust infrastructure. Significant investment was required to align stakeholders, develop trusting relationships, define roles and responsibilities, and to train frontline staff, leading to the establishment of 3 partnerships with internal and external research groups to evaluate Sepsis Watch.ConclusionsMachine learning models are commonly developed to enhance clinical decision making, but successful integrations of machine learning into routine clinical care are rare. Although there is no playbook for integrating deep learning into clinical care, learnings from the Sepsis Watch integration can inform efforts to develop machine learning technologies at other health care delivery systems.
| S-EPMC7391165 | biostudies-literature
Project description:BackgroundCoronary heart disease (CHD) has become the leading cause of death and one of the most serious epidemic diseases worldwide. CHD is characterized by urgency, danger and severity, and dynamic treatment strategies for CHD patients are needed. We aimed to build and validate an AI model for dynamic treatment recommendations for CHD patients with the goal of improving patient outcomes and learning best practices from clinicians to help clinical decision support for treating CHD patients.MethodsWe formed the treatment strategy as a sequential decision problem, and applied an AI supervised reinforcement learning-long short-term memory (SRL-LSTM) framework that combined supervised learning (SL) and reinforcement learning (RL) with an LSTM network to track patients' states to learn a recommendation model that took a patient's diagnosis and evolving health status as input and provided a treatment recommendation in the form of whether to take specific drugs. The experiments were conducted by leveraging a real-world intensive care unit (ICU) database with 13,762 admitted patients diagnosed with CHD. We compared the performance of the applied SRL-LSTM model and several state-of-the-art SL and RL models in reducing the estimated in-hospital mortality and the Jaccard similarity with clinicians' decisions. We used a random forest algorithm to calculate the feature importance of both the clinician policy and the AI policy to illustrate the interpretability of the AI model.ResultsOur experimental study demonstrated that the AI model could help reduce the estimated in-hospital mortality through its RL function and learn the best practice from clinicians through its SL function. The similarity between the clinician policy and the AI policy regarding the surviving patients was high, while for the expired patients, it was much lower. The dynamic treatment strategies made by the AI model were clinically interpretable and relied on sensible clinical features extracted according to monitoring indexes and risk factors for CHD patients.ConclusionsWe proposed a pipeline for constructing an AI model to learn dynamic treatment strategies for CHD patients that could improve patient outcomes and mimic the best practices of clinicians. And a lot of further studies and efforts are needed to make it practical.
| S-EPMC8845235 | biostudies-literature
Project description:With the rapid development of science, technology, and engineering, large amounts of data have been generated in many fields in the past 20 years. In the process of medical research, data are constantly generated, and large amounts of real-world data form a "data disaster." Effective data analysis and mining are based on data availability and high data quality. The premise of high data quality is the need to clean the data. Data cleaning is the process of detecting and correcting "dirty data," which is the basis of data analysis and management. Moreover, data cleaning is a common technology for improving data quality. However, the current literature on real-world research provides little guidance on how to efficiently and ethically set up and perform data cleaning. To address this issue, we proposed a data cleaning framework for real-world research, focusing on the 3 most common types of dirty data (duplicate, missing, and outlier data), and a normal workflow for data cleaning to serve as a reference for the application of such technologies in future studies. We also provided relevant suggestions for common problems in data cleaning.
| S-EPMC10557005 | biostudies-literature
Project description:This study tests the generalisability of three Brain Tumor Segmentation (BraTS) challenge models using a multi-center dataset of varying image quality and incomplete MRI datasets. In this retrospective study, DeepMedic, no-new-Unet (nn-Unet), and NVIDIA-net (nv-Net) were trained and tested using manual segmentations from preoperative MRI of glioblastoma (GBM) and low-grade gliomas (LGG) from the BraTS 2021 dataset (1251 in total), in addition to 275 GBM and 205 LGG acquired clinically across 12 hospitals worldwide. Data was split into 80% training, 5% validation, and 15% internal test data. An additional external test-set of 158 GBM and 69 LGG was used to assess generalisability to other hospitals' data. All models' median Dice similarity coefficient (DSC) for both test sets were within, or higher than, previously reported human inter-rater agreement (range of 0.74-0.85). For both test sets, nn-Unet achieved the highest DSC (internal = 0.86, external = 0.93) and the lowest Hausdorff distances (10.07, 13.87 mm, respectively) for all tumor classes (p < 0.001). By applying Sparsified training, missing MRI sequences did not statistically affect the performance. nn-Unet achieves accurate segmentations in clinical settings even in the presence of incomplete MRI datasets. This facilitates future clinical adoption of automated glioma segmentation, which could help inform treatment planning and glioma monitoring.
| S-EPMC10622563 | biostudies-literature
Project description:Overly restrictive eligibility criteria for clinical trials may limit the generalizability of the trial results to their target real-world patient populations. We developed a novel machine learning approach using large collections of real-world data (RWD) to better inform clinical trial eligibility criteria design. We extracted patients' clinical events from electronic health records (EHRs), which include demographics, diagnoses, and drugs, and assumed certain compositions of these clinical events within an individual's EHRs can determine the subphenotypes-homogeneous clusters of patients, where patients within each subgroup share similar clinical characteristics. We introduced an outcome-guided probabilistic model to identify those subphenotypes, such that the patients within the same subgroup not only share similar clinical characteristics but also at similar risk levels of encountering severe adverse events (SAEs). We evaluated our algorithm on two previously conducted clinical trials with EHRs from the OneFlorida+ Clinical Research Consortium. Our model can clearly identify the patient subgroups who are more likely to suffer or not suffer from SAEs as subphenotypes in a transparent and interpretable way. Our approach identified a set of clinical topics and derived novel patient representations based on them. Each clinical topic represents a certain clinical event composition pattern learned from the patient EHRs. Tested on both trials, patient subgroup (#SAE=0) and patient subgroup (#SAE>0) can be well-separated by k-means clustering using the inferred topics. The inferred topics characterized as likely to align with the patient subgroup (#SAE>0) revealed meaningful combinations of clinical features and can provide data-driven recommendations for refining the exclusion criteria of clinical trials. The proposed supervised topic modeling approach can infer the clinical topics from the subphenotypes with or without SAEs. The potential rules for describing the patient subgroups with SAEs can be further derived to inform the design of clinical trial eligibility criteria.
| S-EPMC9837131 | biostudies-literature
Project description:BackgroundSevere lung cancer is a novel concept that describes a patient with poor performance status (PS; 2-4) but with a high probability of receiving survival benefit and improvement in the PS score. However, there is currently no relevant research or real-world data on those with severe lung cancer, such as incidence, cause, clinical features, and risk factors.MethodsThe data from patients with advanced lung cancer attending multiple centers from January 1, 2022, to June 30, 2022, were collected for a cross-sectional study. In addition, data from fatal cases from January 1, 2019, to June 30, 2022, were retrospectively collected as another cohort. And we developed a questionnaire to assess clinicians' mastery of severe lung cancer.ResultsThree participating institutes enrolled the data set of 1,725 patients, and the dataset of 269 fatal cases were included in another cohort; the incidence of severe lung cancer was 13.10% and 37.55%, respectively. Severe lung cancer patients were mainly stage IV elderly male patients without gene mutation and a history of resection. And the proportion of smoking and comorbidities in severe lung cancer patients is more than in non-severe lung cancer patients (50.4% vs. 40.8%, P=0.006; 46.9% vs. 36.4%, P=0.002). Treatment-related adverse events (AEs) (46.0%) accounted for the largest proportion of the primary causes of severe lung cancer in the cross-sectional study, while cancer-related symptoms (54.5%) accounted for the largest proportion of the primary causes of sever lung cancer in the 101 fatal cases. For the fatal cases, the overall survival of severe lung cancer patients caused by cancer-related symptoms was longer than that caused by treatment-related AEs (8 vs. 3 months; P=0.019). A total of 616 clinicians completed the questionnaire; 90.26% of clinicians agreed with the concept of severe lung cancer.ConclusionsThe incidence of severe lung cancer cannot be ignored based on real-world data. Treatment-related AEs are gradually account for more of the causes of severe lung cancer, surpassing cancer-related symptoms and comorbidities. Furthermore, the prognosis of patients with advanced lung cancer who develop severe lung cancer due to treatment-related AEs is worse than cancer-related symptoms.
| S-EPMC10088003 | biostudies-literature
Project description:Objective: Studies have demonstrated an association between somatic POLE exonuclease domain mutations (EDMs) and the prognosis of colorectal cancer (CRC). However, the prognostic value of POLE non-EDMs remains unclear. This retrospective study aimed to explore the possible relationships between POLE mutation subtypes and CRC prognosis. Methods: The 272 CRC patients from the First Affiliated Hospital of Zhengzhou University (ZZ cohort) and 499 CRC patients from The Cancer Genome Atlas database (TCGA cohort) were retrospectively collected. The cases were divided into subgroups based on POLE mutation sites and microsatellite instability (MSI) status. The continuous variables were compared among three subgroups with Kruskal-Wallis tests. Pairwise comparisons between three groups were performed by Bonferroni correction method, and adjusted p < 0.05 was considered statistically significant. The categorical variables were compared with Chi-square test and Fisher’s exact test. The Kaplan—Meier curves and Cox regression models were conducted to evaluate prognostic values of POLE mutations. Results: In the ZZ cohort, POLE EDMs (2.6%) were significantly associated with younger age (p = 0.018) and localized in the left colon (p = 0.001). POLE non-EDMs were significantly associated with MSI-high status (p < 0.001) and localization in the right colon (p = 0.001). In the TCGA cohort, the tumor mutation burden (TMB) of both POLE EDM tumors (p < 0.001) and POLE non-EDM tumors (p < 0.001) was significantly higher than that of POLE wild-type (WT) tumors. A similar trend was observed in the ZZ cohort, although there were no significant differences. In the ZZ cohort, the POLE EDM group had higher progression-free survival (PFS) (p = 0.002) and overall survival (OS) (p = 0.042) than the POLE non-EDM group and POLE WT group. We also report one CRC patient harboring a germline POLE mutation who received camrelizumab and exhibited long-term stable disease. Conclusion: Both POLE-EDMs and POLE non-EDMs were associated with significantly increased TMB in CRC and may be biomarkers for CRC treatment and prognosis. Current evidence does not support an effect of POLE non-EDMs on PFS and OS. A significant association between POLE EDMs and improved PFS and OS may exist, but future studies with larger sample sizes are needed. Entire coding region of the POLE gene should be screened.
| S-EPMC9719917 | biostudies-literature