Empowering Patient Similarity Networks through Innovative Data-Quality-Aware Federated Profiling.

Continuous monitoring of patients involves collecting and analyzing sensory data from a multitude of sources. To overcome communication overhead, ensure data privacy and security, reduce data loss, and maintain efficient resource usage, the processing and analytics are moved close to where the data are located (e.g., the edge). However, data quality (DQ) can be degraded because of imprecise or malfunctioning sensors, dynamic changes in the environment, transmission failures, or delays. Therefore, it is crucial to keep an eye on data quality and spot problems as quickly as possible, so that they do not mislead clinical judgments and lead to the wrong course of action. In this article, a novel approach called federated data quality profiling (FDQP) is proposed to assess the quality of the data at the edge. FDQP is inspired by federated learning (FL) and serves as a condensed document or a guide for node data quality assurance. The FDQP formal model is developed to capture the quality dimensions specified in the data quality profile (DQP). The proposed approach uses federated feature selection to improve classifier precision and rank features based on criteria such as feature value, outlier percentage, and missing data percentage. Extensive experimentation using a fetal dataset split into different edge nodes and a set of scenarios were carefully chosen to evaluate the proposed FDQP model. The results of the experiments demonstrated that the proposed FDQP approach positively improved the DQ, and thus, impacted the accuracy of the federated patient similarity network (FPSN)-based machine learning models. The proposed data-quality-aware federated PSN architecture leveraging FDQP model with data collected from edge nodes can effectively improve the data quality and accuracy of the federated patient similarity network (FPSN)-based machine learning models. Our profiling algorithm used lightweight profile exchange instead of full data processing at the edge, which resulted in optimal data quality achievement, thus improving efficiency. Overall, FDQP is an effective method for assessing data quality in the edge computing environment, and we believe that the proposed approach can be applied to other scenarios beyond patient monitoring.

Network Approaches for Precision Oncology.

The growth of multi-omic tumour profile datasets along with knowledge of genome regulatory networks has created an unprecedented opportunity to advance precision oncology. Achieving this goal requires computational methods that can make sense of and combine heterogeneous data sources. Interpretability and integration of prior knowledge is of particular relevance for genomic models to minimize ungeneralizable models, promote rational treatment design, and make use of sparse genetic mutation data. While networks have long been used to capture genomic interactions at the levels of genes, proteins, and pathways, the use of networks in precision oncology is relatively new. In this chapter, I provide an introduction to network-based approaches used to integrate multi-modal data sources for patient stratification and patient classification. There is a particular emphasis on methods using patient similarity networks (PSNs) as part of the design. I separately discuss strategies for inferring driver mutations from individual patient mutation data. Finally, I discuss challenges and opportunities the field will need to overcome to achieve its full potential, with an outlook towards a clinic of the future.

Network analytics and machine learning for predicting length of stay in elderly patients with chronic diseases at point of admission.

BACKGROUND: An aging population with a burden of chronic diseases puts increasing pressure on health care systems. Early prediction of the hospital length of stay (LOS) can be useful in optimizing the allocation of medical resources, and improving healthcare quality. However, the data available at the point of admission (PoA) are limited, making it difficult to forecast the LOS accurately. METHODS: In this study, we proposed a novel approach combining network analytics and machine learning to predict the LOS in elderly patients with chronic diseases at the PoA. Two networks, including multimorbidity network (MN) and patient similarity network (PSN), were constructed and novel network features were created. Five machine learning models (eXtreme Gradient Boosting, Gradient Boosting Decision Tree, Random Forest, Linear Support Vector Machine, and Deep Neural Network) with different input feature sets were developed to compare their performance. RESULTS: The experimental results indicated that the network features can bring significant improvements to the performances of the prediction models, suggesting that the MN and PSN are useful for LOS predictions. CONCLUSION: Our predictive framework which integrates network science with data mining can forecast the LOS effectively at the PoA and provide decision support for hospital managers, which highlights the potential value of network-based machine learning in healthcare field.

Gender Differences in Contribution of Smoking, Low Physical Activity, and High BMI to Increased Risk of Early Reoperation After TKA.

BACKGROUND: The reliable preoperative identification of patients at a high risk of early reoperations (<2 years after primary surgery) after total knee arthroplasty (TKA) could lead to adjustments of the surgical procedure and counseling, thus lowering the percentage of revision surgeries. METHODS: The unselected cohort consisted of 1885 patients (695 men and 1190 women) who underwent TKA implantation between September 2010 and April 2017 at a single tertiary orthopedic center. Multivariate patient similarity networks were applied to identify patient groups at a high risk of early reoperations based on 25 preoperative parameters. RESULTS: Early reoperations (109 cases, 5.8%) were less frequent in women (4.4%; median time to reoperation, 2.0 months) than in men (8.2%; 7.5 months), reaching the highest incidence in younger men (10.9%; <66 years). Of the tested preoperative parameters, the risk of reoperation in men was more likely associated with smoking or obesity (body mass index [BMI] > 30). In women, low physical activity and high BMI were the most likely risk factors for early reoperations. Other factors did not affect the risk of early reoperations, including the primary diagnosis, comorbidities, and surgeon-implanting TKA. CONCLUSION: This study demonstrates the effect of smoking, physical activity, and BMI on the risk of early reoperation after TKA, with the different contribution in men/women. Identification of patient subgroups with a higher risk of early revision after TKA is needed for clinical implementation of precision medicine in orthopedics.

MMGCN: Multi-modal multi-view graph convolutional networks for cancer prognosis prediction.

BACKGROUND AND OBJECTIVE: Accurate prognosis prediction for cancer patients plays a significant role in the formulation of treatment strategies, considerably impacting personalized medicine. Recent advancements in this field indicate that integrating information from various modalities, such as genetic and clinical data, and developing multi-modal deep learning models can enhance prediction accuracy. However, most existing multi-modal deep learning methods either overlook patient similarities that benefit prognosis prediction or fail to effectively capture diverse information due to measuring patient similarities from a single perspective. To address these issues, a novel framework called multi-modal multi-view graph convolutional networks (MMGCN) is proposed for cancer prognosis prediction. METHODS: Initially, we utilize the similarity network fusion (SNF) algorithm to merge patient similarity networks (PSNs), individually constructed using gene expression, copy number alteration, and clinical data, into a fused PSN for integrating multi-modal information. To capture diverse perspectives of patient similarities, we treat the fused PSN as a multi-view graph by considering each single-edge-type subgraph as a view graph, and propose multi-view graph convolutional networks (GCNs) with a view-level attention mechanism. Moreover, an edge homophily prediction module is designed to alleviate the adverse effects of heterophilic edges on the representation power of GCNs. Finally, comprehensive representations of patient nodes are obtained to predict cancer prognosis. RESULTS: Experimental results demonstrate that MMGCN outperforms state-of-the-art baselines on four public datasets, including METABRIC, TCGA-BRCA, TCGA-LGG, and TCGA-LUSC, with the area under the receiver operating characteristic curve achieving 0.827 ± 0.005, 0.805 ± 0.014, 0.925 ± 0.007, and 0.746 ± 0.013, respectively. CONCLUSIONS: Our study reveals the effectiveness of the proposed MMGCN, which deeply explores patient similarities related to different modalities from a broad perspective, in enhancing the performance of multi-modal cancer prognosis prediction. The source code is publicly available at https://github.com/ping-y/MMGCN.

Analysis of Chinese Medical Syndrome Features of Ischemic Stroke Based on Similarity of Symptoms Subgroup.

OBJECTIVE: To derive the Chinese medicine (CM) syndrome classification and subgroup syndrome characteristics of ischemic stroke patients. METHODS: By extracting the CM clinical electronic medical records (EMRs) of 7,170 hospitalized patients with ischemic stroke from 2016 to 2018 at Weifang Hospital of Traditional Chinese Medicine, Shandong Province, China, a patient similarity network (PSN) was constructed based on the symptomatic phenotype of the patients. Thereafter the efficient community detection method BGLL was used to identify subgroups of patients. Finally, subgroups with a large number of cases were selected to analyze the specific manifestations of clinical symptoms and CM syndromes in each subgroup. RESULTS: Seven main subgroups of patients with specific symptom characteristics were identified, including M3, M2, M1, M5, M0, M29 and M4. M3 and M0 subgroups had prominent posterior circulatory symptoms, while M3 was associated with autonomic disorders, and M4 manifested as anxiety; M2 and M4 had motor and motor coordination disorders; M1 had sensory disorders; M5 had more obvious lung infections; M29 had a disorder of consciousness. The specificity of CM syndromes of each subgroup was as follows. M3, M2, M1, M0, M29 and M4 all had the same syndrome as wind phlegm pattern; M3 and M0 both showed hyperactivity of Gan (Liver) yang pattern; M2 and M29 had similar syndromes, which corresponded to intertwined phlegm and blood stasis pattern and phlegm-stasis obstructing meridians pattern, respectively. The manifestations of CM syndromes often appeared in a combination of 2 or more syndrome elements. The most common combination of these 7 subgroups was wind-phlegm. The 7 subgroups of CM syndrome elements were specifically manifested as pathogenic wind, pathogenic phlegm, and deficiency pathogens. CONCLUSIONS: There were 7 main symptom similarity-based subgroups in ischemic stroke patients, and their specific characteristics were obvious. The main syndromes were wind phlegm pattern and hyperactivity of Gan yang pattern.

Predicting the risk of mortality in ICU patients based on dynamic graph attention network of patient similarity.

Predicting the risk of mortality of hospitalized patients in the ICU is essential for timely identification of high-risk patients and formulate and adjustment of treatment strategies when patients are hospitalized. Traditional machine learning methods usually ignore the similarity between patients and make it difficult to uncover the hidden relationships between patients, resulting in poor accuracy of prediction models. In this paper, we propose a new model named PS-DGAT to solve the above problem. First, we construct a patient-weighted similarity network by calculating the similarity of patient clinical data to represent the similarity relationship between patients; second, we fill in the missing features and reconstruct the patient similarity network based on the data of neighboring patients in the network; finally, from the reconstructed patient similarity network after feature completion, we use the dynamic attention mechanism to extract and learn the structural features of the nodes to obtain a vector representation of each patient node in the low-dimensional embedding The vector representation of each patient node in the low-dimensional embedding space is used to achieve patient mortality risk prediction. The experimental results show that the accuracy is improved by about 1.8% compared with the basic GAT and about 8% compared with the traditional machine learning methods.

GNN-surv: Discrete-Time Survival Prediction Using Graph Neural Networks.

Survival prediction models play a key role in patient prognosis and personalized treatment. However, their accuracy can be improved by incorporating patient similarity networks, which uncover complex data patterns. Our study uses Graph Neural Networks (GNNs) to enhance discrete-time survival predictions (GNN-surv) by leveraging relationships in these networks. We build these networks using cancer patients' genomic and clinical data and train various GNN models on them, integrating Logistic Hazard and PMF survival models. GNN-surv models exhibit superior performance in survival prediction across two urologic cancer datasets, outperforming traditional MLP models. They maintain robustness and effectiveness under varying graph construction hyperparameter µ values, with performance boosts of up to 14.6% and 7.9% in the time-dependent concordance index and reductions in the integrated brier score of 26.7% and 24.1% in the BLCA and KIRC datasets, respectively. Notably, these models also maintain their effectiveness across three different types of GNN models, suggesting potential adaptability to other cancer datasets. The superior performance of our GNN-surv models underscores their wide applicability in the fields of oncology and personalized medicine, providing clinicians with a more accurate tool for patient prognosis and personalized treatment planning. Future studies can further optimize these models by incorporating other survival models or additional data modalities.

Network medicine for patients' stratification: From single-layer to multi-omics.

Precision medicine research increasingly relies on the integrated analysis of multiple types of omics. In the era of big data, the large availability of different health-related information represents a great, but at the same time untapped, chance with a potentially fundamental role in the prevention, diagnosis and prognosis of diseases. Computational methods are needed to combine this data to create a comprehensive view of a given disease. Network science can model biomedical data in terms of relationships among molecular players of different nature and has been successfully proposed as a new paradigm for studying human diseases. Patient stratification is an open challenge aimed at identifying subtypes with different disease manifestations, severity, and expected survival time. Several stratification approaches based on high-throughput gene expression measurements have been successfully applied. However, few attempts have been proposed to exploit the integration of various genotypic and phenotypic data to discover novel sub-types or improve the detection of known groupings. This article is categorized under: Cancer > Biomedical Engineering Cancer > Computational Models Cancer > Genetics/Genomics/Epigenetics.

A classification method of gastric cancer subtype based on residual graph convolution network.

Background: Clinical diagnosis and treatment of tumors are greatly complicated by their heterogeneity, and the subtype classification of cancer frequently plays a significant role in the subsequent treatment of tumors. Presently, the majority of studies rely far too heavily on gene expression data, omitting the enormous power of multi-omics fusion data and the potential for patient similarities. Method: In this study, we created a gastric cancer subtype classification model called RRGCN based on residual graph convolutional network (GCN) using multi-omics fusion data and patient similarity network. Given the multi-omics data's high dimensionality, we built an artificial neural network Autoencoder (AE) to reduce the dimensionality of the data and extract hidden layer features. The model is then built using the feature data. In addition, we computed the correlation between patients using the Pearson correlation coefficient, and this relationship between patients forms the edge of the graph structure. Four graph convolutional network layers and two residual networks with skip connections make up RRGCN, which reduces the amount of information lost during transmission between layers and prevents model degradation. Results: The results show that RRGCN significantly outperforms other classification methods with an accuracy as high as 0.87 when compared to four other traditional machine learning methods and deep learning models. Conclusion: In terms of subtype classification, RRGCN excels in all areas and has the potential to offer fresh perspectives on disease mechanisms and disease progression. It has the potential to be used for a broader range of disorders and to aid in clinical diagnosis.

Are risk factors necessary for pretest probability assessment of coronary artery disease? A patient similarity network analysis of the PROMISE trial.

BACKGROUND: Pretest probability (PTP) calculators utilize epidemiological-level findings to provide patient-level risk assessment of obstructive coronary artery disease (CAD). However, their limited accuracies question whether dissimilarities in risk factors necessarily result in differences in CAD. Using patient similarity network (PSN) analyses, we wished to assess the accuracy of risk factors and imaging markers to identify ≥50% luminal narrowing on coronary CT angiography (CCTA) in stable chest-pain patients. METHODS: We created four PSNs representing: patient characteristics, risk factors, non-coronary imaging markers and calcium score. We used spectral clustering to group individuals with similar risk profiles. We compared PSNs to a contemporary PTP score incorporating calcium score and risk factors to identify ≥50% luminal narrowing on CCTA in the CT-arm of the PROMISE trial. We also conducted subanalyses in different age and sex groups. RESULTS: In 3556 individuals, the calcium score PSN significantly outperformed patient characteristic, risk factor, and non-coronary imaging marker PSNs (AUC: 0.81 vs. 0.57, 0.55, 0.54; respectively, p â€‹< â€‹0.001 for all). The calcium score PSN significantly outperformed the contemporary PTP score (AUC: 0.81 vs. 0.78, p â€‹< â€‹0.001), and using 0, 1-100 and â€‹> â€‹100 cut-offs provided comparable results (AUC: 0.81 vs. 0.81, p â€‹= â€‹0.06). Similar results were found in all subanalyses. CONCLUSION: Calcium score on its own provides better individualized obstructive CAD prediction than contemporary PTP scores incorporating calcium score and risk factors. Risk factors may not be able to improve the diagnostic accuracy of calcium score to predict ≥50% luminal narrowing on CCTA.

A Novel Patient Similarity Network (PSN) Framework Based on Multi-Model Deep Learning for Precision Medicine.

Precision medicine can be defined as the comparison of a new patient with existing patients that have similar characteristics and can be referred to as patient similarity. Several deep learning models have been used to build and apply patient similarity networks (PSNs). However, the challenges related to data heterogeneity and dimensionality make it difficult to use a single model to reduce data dimensionality and capture the features of diverse data types. In this paper, we propose a multi-model PSN that considers heterogeneous static and dynamic data. The combination of deep learning models and PSN allows ample clinical evidence and information extraction against which similar patients can be compared. We use the bidirectional encoder representations from transformers (BERT) to analyze the contextual data and generate word embedding, where semantic features are captured using a convolutional neural network (CNN). Dynamic data are analyzed using a long-short-term-memory (LSTM)-based autoencoder, which reduces data dimensionality and preserves the temporal features of the data. We propose a data fusion approach combining temporal and clinical narrative data to estimate patient similarity. The experiments we conducted proved that our model provides a higher classification accuracy in determining various patient health outcomes when compared with other traditional classification algorithms.

Network Analysis for Uncovering the Relationship between Host Response and Clinical Factors to Virus Pathogen: Lessons from SARS-CoV-2.

Analysing complex datasets while maintaining the interpretability and explainability of outcomes for clinicians and patients is challenging, not only in viral infections. These datasets often include a variety of heterogeneous clinical, demographic, laboratory, and personal data, and it is not a single factor but a combination of multiple factors that contribute to patient characterisation and host response. Therefore, multivariate approaches are needed to analyse these complex patient datasets, which are impossible to analyse with univariate comparisons (e.g., one immune cell subset versus one clinical factor). Using a SARS-CoV-2 infection as an example, we employed a patient similarity network (PSN) approach to assess the relationship between host immune factors and the clinical course of infection and performed visualisation and data interpretation. A PSN analysis of ~85 immunological (cellular and humoral) and ~70 clinical factors in 250 recruited patients with coronavirus disease (COVID-19) who were sampled four to eight weeks after a PCR-confirmed SARS-CoV-2 infection identified a minimal immune signature, as well as clinical and laboratory factors strongly associated with disease severity. Our study demonstrates the benefits of implementing multivariate network approaches to identify relevant factors and visualise their relationships in a SARS-CoV-2 infection, but the model is generally applicable to any complex dataset.