Machine and deep learning methods for clinical outcome prediction based on physiological data of COVID-19 patients: a scoping review

INTRODUCTION

Since the beginning of the COVID-19 pandemic, numerous machine and deep learning (MDL) methods have been proposed in the literature to analyze patient physiological data. The objective of this review is to summarize various aspects of these methods and assess their practical utility for predicting various clinical outcomes.

METHODS

We searched PubMed, Scopus, and Cochrane Library, screened and selected the studies matching the inclusion criteria. The clinical analysis focused on the characteristics of the patient cohorts in the studies included in this review, the specific tasks in the context of the COVID-19 pandemic that machine and deep learning methods were used for, and their practical limitations. The technical analysis focused on the details of specific MDL methods and their performance.

RESULTS

Analysis of the 48 selected studies revealed that the majority (∼54 %) of them examined the application of MDL methods for the prediction of survival/mortality-related patient outcomes, while a smaller fraction (∼13 %) of studies also examined applications to the prediction of patients' physiological outcomes and hospital resource utilization. 21 % of the studies examined the application of MDL methods to multiple clinical tasks. Machine and deep learning methods have been shown to be effective at predicting several outcomes of COVID-19 patients, such as disease severity, complications, intensive care unit (ICU) transfer, and mortality. MDL methods also achieved high accuracy in predicting the required number of ICU beds and ventilators.

CONCLUSION

Machine and deep learning methods have been shown to be valuable tools for predicting disease severity, organ dysfunction and failure, patient outcomes, and hospital resource utilization during the COVID-19 pandemic. The discovered knowledge and our conclusions and recommendations can also be useful to healthcare professionals and artificial intelligence researchers in managing future pandemics.

EHR-QC: A streamlined pipeline for automated electronic health records standardisation and preprocessing to predict clinical outcomes

The adoption of electronic health records (EHRs) has created opportunities to analyse historical data for predicting clinical outcomes and improving patient care. However, non-standardised data representations and anomalies pose major challenges to the use of EHRs in digital health research. To address these challenges, we have developed EHR-QC, a tool comprising two modules: the data standardisation module and the preprocessing module. The data standardisation module migrates source EHR data to a standard format using advanced concept mapping techniques, surpassing expert curation in benchmarking analysis. The preprocessing module includes several functions designed specifically to handle healthcare data subtleties. We provide automated detection of data anomalies and solutions to handle those anomalies. We believe that the development and adoption of tools like EHR-QC is critical for advancing digital health. Our ultimate goal is to accelerate clinical research by enabling rapid experimentation with data-driven observational research to generate robust, generalisable biomedical knowledge.

From Data to Wisdom: Biomedical Knowledge Graphs for Real-World Data Insights

Graph data models are an emerging approach to structure clinical and biomedical information. These models offer intriguing opportunities for novel approaches in healthcare, such as disease phenotyping, risk prediction, and personalized precision care. The combination of data and information in a graph model to create knowledge graphs has rapidly expanded in biomedical research, but the integration of real-world data from the electronic health record has been limited. To broadly apply knowledge graphs to EHR and other real-world data, a deeper understanding of how to represent these data in a standardized graph model is needed. We provide an overview of the state-of-the-art research for clinical and biomedical data integration and summarize the potential to accelerate healthcare and precision medicine research through insight generation from integrated knowledge graphs.

Applied machine learning in cancer research: A systematic review for patient diagnosis, classification and prognosis

Artificial Intelligence (AI) has recently altered the landscape of cancer research and medical oncology using traditional Machine Learning (ML) algorithms and cutting-edge Deep Learning (DL) architectures. In this review article we focus on the ML aspect of AI applications in cancer research and present the most indicative studies with respect to the ML algorithms and data used. The PubMed and dblp databases were considered to obtain the most relevant research works of the last five years. Based on a comparison of the proposed studies and their research clinical outcomes concerning the medical ML application in cancer research, three main clinical scenarios were identified. We give an overview of the well-known DL and Reinforcement Learning (RL) methodologies, as well as their application in clinical practice, and we briefly discuss Systems Biology in cancer research. We also provide a thorough examination of the clinical scenarios with respect to disease diagnosis, patient classification and cancer prognosis and survival. The most relevant studies identified in the preceding year are presented along with their primary findings. Furthermore, we examine the effective implementation and the main points that need to be addressed in the direction of robustness, explainability and transparency of predictive models. Finally, we summarize the most recent advances in the field of AI/ML applications in cancer research and medical oncology, as well as some of the challenges and open issues that need to be addressed before data-driven models can be implemented in healthcare systems to assist physicians in their daily practice.

Knowledge-aware multi-center clinical dataset adaptation: Problem, method, and application

Adaptable utilization of clinical data collected from multiple centers, prompted by the need to overcome the shifts between the dataset distributions, and exploit these different datasets for potential clinical applications, has received significant attention in recent years. In this study, we propose a novel approach to this task by infusing an external knowledge graph (KG) into multi-center clinical data mining. Specifically, we propose an adversarial learning model to capture shared patient feature representations from multi-center heterogeneous clinical datasets, and employ an external KG to enrich the semantics of the patient sample by providing both clinical center-specific and center-general knowledge features, which are trained with a graph convolutional autoencoder. We evaluate the proposed model on a real clinical dataset extracted from the general cardiology wards of a Chinese hospital and a well-known public clinical dataset (MIMIC III, pertaining to ICU clinical settings) for the task of predicting acute kidney injury in patients with heart failure. The achieved experimental results demonstrate the efficacy of our proposed model.

Penumbra-based radiomics signature as prognostic biomarkers for thrombolysis of acute ischemic stroke patients: a multicenter cohort study

BACKGROUND AND PURPOSE

This study aimed at developing a radiomics signature (R score) as prognostic biomarkers based on penumbra quantification and to validate the radiomics nomogram to predict the clinical outcomes for thrombolysis for acute ischemic stroke (AIS) patients.

METHODS

In total, 168 patients collected from seven centers were retrospectively included. A score of mismatch was defined as MIS. Based on a short-term clinical label, 456 radiomics features were evaluated with feature selection methods. R score was constructed with the selected features. To compare the predictive capabilities of the clinical factors, MIS, and R score, three nomograms were developed and evaluated, according to the short-term clinical assessment on day 7. Finally, the radiomics nomogram was validated by predicting the 3-month clinical outcomes of AIS patients, in an external cohort.

RESULTS

R scores were found to be significantly higher in patients with favorable clinical outcomes in both training and validation datasets. The predictive value of the radiomics nomogram estimating favorable clinical outcomes was modest, with a concordance index (C-index) of 0.695 [95% confidence interval (CI) 0.667-0.723) in an external validation dataset. In addition, the area under curve (AUC) of the radiomics nomogram predicting favorable clinical outcome reached 0.886 (95% CI 0.809-0.963) on day 7 and 0.777 (95% CI 0.666-0.888) at 3 months.

CONCLUSIONS

The radiomics signature is an independent biomarker for estimating the clinical outcomes in AIS patients. By improving the individualized prediction of the clinical outcome for AIS patients 3 months after onset, the radiomics nomogram adds more value to the current clinical decision-making process.

Predicting Clinical Outcome of Stroke Patients with Tractographic Feature

The volume of stroke lesion is the gold standard for predicting the clinical outcome of stroke patients. However, the presence of stroke lesion may cause neural disruptions to other brain regions, and these potentially damaged regions may affect the clinical outcome of stroke patients. In this paper, we introduce the tractographic feature to capture these potentially damaged regions and predict the modified Rankin Scale (mRS), which is a widely used outcome measure in stroke clinical trials. The tractographic feature is built from the stroke lesion and average connectome information from a group of normal subjects. The tractographic feature takes into account different functional regions that may be affected by the stroke, thus complementing the commonly used stroke volume features. The proposed tractographic feature is tested on a public stroke benchmark Ischemic Stroke Lesion Segmentation 2017 and achieves higher accuracy than the stroke volume and the state-of-the-art feature on predicting the mRS grades of stroke patients. Also, the tractographic feature yields a lower average absolute error than the commonly used stroke volume feature.

A deep neural network approach to predicting clinical outcomes of neuroblastoma patients

Background

The availability of high-throughput omics datasets from large patient cohorts has allowed the development of methods that aim at predicting patient clinical outcomes, such as survival and disease recurrence. Such methods are also important to better understand the biological mechanisms underlying disease etiology and development, as well as treatment responses. Recently, different predictive models, relying on distinct algorithms (including Support Vector Machines and Random Forests) have been investigated. In this context, deep learning strategies are of special interest due to their demonstrated superior performance over a wide range of problems and datasets. One of the main challenges of such strategies is the “small n large p” problem. Indeed, omics datasets typically consist of small numbers of samples and large numbers of features relative to typical deep learning datasets. Neural networks usually tackle this problem through feature selection or by including additional constraints during the learning process.

Methods

We propose to tackle this problem with a novel strategy that relies on a graph-based method for feature extraction, coupled with a deep neural network for clinical outcome prediction. The omics data are first represented as graphs whose nodes represent patients, and edges represent correlations between the patients’ omics profiles. Topological features, such as centralities, are then extracted from these graphs for every node. Lastly, these features are used as input to train and test various classifiers.

Results

We apply this strategy to four neuroblastoma datasets and observe that models based on neural networks are more accurate than state of the art models (DNN: 85%-87%, SVM/RF: 75%-82%). We explore how different parameters and configurations are selected in order to overcome the effects of the small data problem as well as the curse of dimensionality.

Conclusions

Our results indicate that the deep neural networks capture complex features in the data that help predicting patient clinical outcomes.

Early warning and clinical outcome prediction of acute-on-chronic hepatitis B liver failure

Hepatitis B virus (HBV) associated acute-on-chronic liver failure (ACLF) is an increasingly recognized fatal liver disease encompassing a severe acute exacerbation of liver function in patients with chronic hepatitis B (CHB). Despite the introduction of an artificial liver support system and antiviral therapy, the short-term prognosis of HBV-ACLF is still extremely poor unless emergency liver transplantation is performed. In such a situation, stopping or slowing the progression of CHB to ACLF at an early stage is the most effective way of reducing the morbidity and mortality of HBV-ACLF. It is well-known that the occurrence and progression of HBV-ACLF is associated with many factors, and the outcomes of HBV-ACLF patients can be significantly improved if timely and appropriate interventions are provided. In this review, we highlight recent developments in early warning and clinical outcome prediction in patients with HBV-ACLF and provide an outlook for future research in this field.

High Serum MiR-130a Levels Are Associated with Severe Perihematomal Edema and Predict Adverse Outcome in Acute ICH

The development and/or progression of perihematomal edema (PHE) in patients with acute spontaneous intracerebral hemorrhage (ICH) vary substantially with different individuals. Although hematoma volume is a useful indicator for predicting PHE, its predictive power was not good at the early stage of ICH. Better predictors are urgently needed. In this study, we found that miR-130a was elevated in the serum of ICH patients and was an independent indicator positively associated with PHE volume within the first 3 days after onset. The R (2) was further evaluated when it is used in combination with hematoma mass. Serum miR-130a levels were associated with clinical outcome (National Institute of Health Stroke Scale (NIHSS) scores at day 14 and modified Rankin Scale (mRS) scores at day 90) only in patients with deep hematoma. Moreover, miR-130a was significantly increased in rat serum and perihematomal tissues and was in line with the change in brain edema. MiR-130a inhibitors reduced brain edema, blood-brain barrier (BBB) permeability, and increased neurological deficit scores, and miR-130a mimics increased monolayer permeability. Thrombin-stimulated brain microvascular endothelial cells (BMECs) were a main source of miR-130a under ICH. In the experimental model, the elevated miR-130a level was accompanied by the decreased caveolin-1 and increased matrix metalloproleinase (MMP)-2/9. Meanwhile, caveolin-1 (cav-1) was reduced by miR-130a mimics, accompanied by an increase in MMP-2/9 expression. The upregulated MMP-2/9 was then downregulated by cavtratin, a cav-1 scaffolding domain peptide. This regulation mechanism was authenticated in a thrombin-induced cellular ICH model. Our results suggest that serum miR-130a may serve as a useful early biomarker for monitoring post-ICH PHE and predicting prognosis and may be helpful in the decision-making of individualized therapy.

Knowledge-driven genomic interactions: an application in ovarian cancer

Background

Effective cancer clinical outcome prediction for understanding of the mechanism of various types of cancer has been pursued using molecular-based data such as gene expression profiles, an approach that has promise for providing better diagnostics and supporting further therapies. However, clinical outcome prediction based on gene expression profiles varies between independent data sets. Further, single-gene expression outcome prediction is limited for cancer evaluation since genes do not act in isolation, but rather interact with other genes in complex signaling or regulatory networks. In addition, since pathways are more likely to co-operate together, it would be desirable to incorporate expert knowledge to combine pathways in a useful and informative manner.

Methods

Thus, we propose a novel approach for identifying knowledge-driven genomic interactions and applying it to discover models associated with cancer clinical phenotypes using grammatical evolution neural networks (GENN). In order to demonstrate the utility of the proposed approach, an ovarian cancer data from the Cancer Genome Atlas (TCGA) was used for predicting clinical stage as a pilot project.

Results

We identified knowledge-driven genomic interactions associated with cancer stage from single knowledge bases such as sources of pathway-pathway interaction, but also knowledge-driven genomic interactions across different sets of knowledge bases such as pathway-protein family interactions by integrating different types of information. Notably, an integration model from different sources of biological knowledge achieved 78.82% balanced accuracy and outperformed the top models with gene expression or single knowledge-based data types alone. Furthermore, the results from the models are more interpretable because they are framed in the context of specific biological pathways or other expert knowledge.

Conclusions

The success of the pilot study we have presented herein will allow us to pursue further identification of models predictive of clinical cancer survival and recurrence. Understanding the underlying tumorigenesis and progression in ovarian cancer through the global view of interactions within/between different biological knowledge sources has the potential for providing more effective screening strategies and therapeutic targets for many types of cancer.

Incorporating inter-relationships between different levels of genomic data into cancer clinical outcome prediction

In order to improve our understanding of cancer and develop multi-layered theoretical models for the underlying mechanism, it is essential to have enhanced understanding of the interactions between multiple levels of genomic data that contribute to tumor formation and progression. Although there exist recent approaches such as a graph-based framework that integrates multi-omics data including copy number alteration, methylation, gene expression, and miRNA data for cancer clinical outcome prediction, most of previous methods treat each genomic data as independent and the possible interplay between them is not explicitly incorporated to the model. However, cancer is dysregulated by multiple levels in the biological system through genomic, epigenomic, transcriptomic, and proteomic level. Thus, genomic features are likely to interact with other genomic features in the different genomic levels. In order to deepen our knowledge, it would be desirable to incorporate such inter-relationship information when integrating multi-omics data for cancer clinical outcome prediction. In this study, we propose a new graph-based framework that integrates not only multi-omics data but inter-relationship between them for better elucidating cancer clinical outcomes. In order to highlight the validity of the proposed framework, serous cystadenocarcinoma data from TCGA was adopted as a pilot task. The proposed model incorporating inter-relationship between different genomic features showed significantly improved performance compared to the model that does not consider inter-relationship when integrating multi-omics data. For the pair between miRNA and gene expression data, the model integrating miRNA, for example, gene expression, and inter-relationship between them with an AUC of 0.8476 (REI) outperformed the model combining miRNA and gene expression data with an AUC of 0.8404. Similar results were also obtained for other pairs between different levels of genomic data. Integration of different levels of data and inter-relationship between them can aid in extracting new biological knowledge by drawing an integrative conclusion from many pieces of information collected from diverse types of genomic data, eventually leading to more effective screening strategies and alternative therapies that may improve outcomes.

Bayesian evolutionary hypergraph learning for predicting cancer clinical outcomes

Predicting the clinical outcomes of cancer patients is a challenging task in biomedicine. A personalized and refined therapy based on predicting prognostic outcomes of cancer patients has been actively sought in the past decade. Accurate prognostic prediction requires higher-order representations of complex dependencies among genetic factors. However, identifying the co-regulatory roles and functional effects of genetic interactions on cancer prognosis is hindered by the complexity of the interactions. Here we propose a prognostic prediction model based on evolutionary learning that identifies higher-order prognostic biomarkers of cancer clinical outcomes. The proposed model represents the interactions of prognostic genes as a combinatorial space. It adopts a flexible hypergraph structure composed of a large population of hyperedges that encode higher-order relationships among many genetic factors. The hyperedge population is optimized by an evolutionary learning method based on sequential Bayesian sampling. The proposed learning approach effectively balances performance and parsimony of the model using information-theoretic dependency and complexity-theoretic regularization priors. Using MAQC-II project data, we demonstrate that our model can handle high-dimensional data more effectively than state-of-the-art classification models. We also identify potential gene interactions characterizing prognosis and recurrence risk in cancer.

Prediction of Clinical Outcome in Acute Hemorrhagic Stroke from a Single CT Scan on Admission

Background:

From a single CT scan in primary intracerebral hemorrhage (ICH), clinical outcome can be assessed on admission by using the CT scan parameters.

Aims:

The study aims to find out how hematoma volume, location of stroke, midline shift, intraventricular extension of bleed and ventricle compression influence the clinical outcome in patients with acute ICH.

Materials and Methods:

Non-contrast CT scan was done on admission in hospital for every patient with acute hemorrhagic stroke and was analyzed accordingly. Clinical assessments were done in National Institute of Health Stroke Scale (NIHSS). Chi-square test and multiple logistic regression analysis were used for statistical analysis.

Results:

Mean hematoma volume associated with death before 30 days is 33.16 cm³ (P < 0.0001), with survived after 30 days is 15.45 cm³ (P < 0.0001), with NIHSS score ≥16 is 29.03 cm³ (P < 0.0001) and with NIHSS score <16 is 13.69 cm³ (P < 0.0001). Independent poor prognostic factors were hematoma volume > 30 cm³ (OR = 27.857), brain stem hemorrhage (OR = 6.000), intraventricular extension of bleed from other location (OR = 7.846), presence of ventricular compression alone (OR = 2.700) and in combination with midline shift of ≥ 5 mm (OR = 2.124).

Conclusions:

From a single CT scan during hospital admission, mortality and morbidity in next 30 days can be predicted. A hematoma volume >30 cm³, brain stem hematoma, intraventricular extension of bleed and ventricular compression along and with midline shift are associated with early mortality in ICH.