FHBF: Federated hybrid boosted forests with dropout rates for supervised learning tasks across highly imbalanced clinical datasets.

Although several studies have deployed gradient boosting trees (GBT) as a robust classifier for federated learning tasks (federated GBT [FGBT]), even with dropout rates (federated gradient boosting trees with dropout rate [FDART]), none of them have investigated the overfitting effects of FGBT across heterogeneous and highly imbalanced datasets within federated environments nor the effect of dropouts in the loss function. In this work, we present the federated hybrid boosted forests (FHBF) algorithm, which incorporates a hybrid weight update approach to overcome ill-posed problems that arise from overfitting effects during the training across highly imbalanced datasets in the cloud. Eight case studies were conducted to stress the performance of FHBF against existing algorithms toward the development of robust AI models for lymphoma development across 18 European federated databases. Our results highlight the robustness of FHBF, yielding an average loss of 0.527 compared with FGBT (0.611) and FDART (0.584) with increased classification performance (0.938 sensitivity, 0.732 specificity).

Addressing the clinical unmet needs in primary Sjögren's Syndrome through the sharing, harmonization and federated analysis of 21 European cohorts.

For many decades, the clinical unmet needs of primary Sjögren's Syndrome (pSS) have been left unresolved due to the rareness of the disease and the complexity of the underlying pathogenic mechanisms, including the pSS-associated lymphomagenesis process. Here, we present the HarmonicSS cloud-computing exemplar which offers beyond the state-of-the-art data analytics services to address the pSS clinical unmet needs, including the development of lymphoma classification models and the identification of biomarkers for lymphomagenesis. The users of the platform have been able to successfully interlink, curate, and harmonize 21 regional, national, and international European cohorts of 7,551 pSS patients with respect to the ethical and legal issues for data sharing. Federated AI algorithms were trained across the harmonized databases, with reduced execution time complexity, yielding robust lymphoma classification models with 85% accuracy, 81.25% sensitivity, 85.4% specificity along with 5 biomarkers for lymphoma development. To our knowledge, this is the first GDPR compliant platform that provides federated AI services to address the pSS clinical unmet needs.

A federated AI strategy for the classification of patients with Mucosa Associated Lymphoma Tissue (MALT) lymphoma across multiple harmonized cohorts.

Mucosa Associated Lymphoma Tissue (MALT) type is an extremely rare type of lymphoma which occurs in less than 3% of patients with primary Sjögren's Syndrome (pSS). No reported studies so far have been able to investigate risk factors for MALT development across multiple cohort databases with sufficient statistical power. Here, we present a generalized, federated AI (artificial intelligence) strategy which enables the training of AI algorithms across multiple harmonized databases. A case study is conducted towards the development of MALT classification models across 17 databases on pSS. Advanced AI algorithms were developed, including federated Multinomial Naïve Bayes (FMNB), federated gradient boosting trees (FGBT), FGBT with dropouts (FDART), and the federated Multilayer Perceptron (FMLP). The FDART with dropout rate 0.3 achieved the best performance with sensitivity 0.812, and specificity 0.829, yielding 8 biomarkers as prominent for MALT development.

Overcoming the Barriers That Obscure the Interlinking and Analysis of Clinical Data Through Harmonization and Incremental Learning.

Goal: To present a framework for data sharing, curation, harmonization and federated data analytics to solve open issues in healthcare, such as, the development of robust disease prediction models. Methods: Data curation is applied to remove data inconsistencies. Lexical and semantic matching methods are used to align the structure of the heterogeneous, curated cohort data along with incremental learning algorithms including class imbalance handling and hyperparameter optimization to enable the development of disease prediction models. Results: The applicability of the framework is demonstrated in a case study of primary Sjögren's Syndrome, yielding harmonized data with increased quality and more than 85% agreement, along with lymphoma prediction models with more than 80% sensitivity and specificity. Conclusions: The framework provides data quality, harmonization and analytics workflows that can enhance the statistical power of heterogeneous clinical data and enables the development of robust models for disease prediction.

Utility of Multimodality Intravascular Imaging and the Local Hemodynamic Forces to Predict Atherosclerotic Disease Progression.

OBJECTIVES: This study sought to examine the utility of multimodality intravascular imaging and of the endothelial shear stress (ESS) distribution to predict atherosclerotic evolution. BACKGROUND: There is robust evidence that intravascular ultrasound (IVUS)-derived plaque characteristics and ESS distribution can predict, with however limited accuracy, atherosclerotic evolution; nevertheless, it is yet unclear whether multimodality imaging and ESS mapping enable more accurate prediction of coronary plaque progression. METHODS: A total of 44 patients admitted with a myocardial infarction that had successful revascularization and 3-vessel IVUS and optical coherence tomography (OCT) imaging at baseline and 13-month follow-up were included in the study. The IVUS data acquired at baseline in the nonculprit vessels were fused with x-ray angiography to reconstruct coronary anatomy and in the obtained models blood flow simulation was performed and the ESS was estimated. The baseline plaque characteristics and ESS distribution were used to identify predictors of disease progression: defined as a lumen reduction and an increase in plaque burden at follow-up. RESULTS: Seventy-three vessels were included in the final analysis. Baseline ESS and the IVUS-derived but not the OCT-derived plaque characteristics were independently associated with a decrease in lumen area and an increase in plaque burden. Low ESS (odds ratio: 0.45; 95% confidence interval: 0.28 to 0.71; p < 0.001) and plaque burden (odds ratio: 0.73; 95% confidence interval: 0.54 to 0.97; p = 0.030) were the only independent predictors of disease progression at follow-up. The accuracy of the IVUS-derived plaque characteristics in predicting disease progression did not improve when ESS (AUC: 0.824 vs. 0.847; p = 0.127) or when OCT variables and ESS (AUC: 0.842; p = 0.611) were added into the model. CONCLUSIONS: ESS and OCT-derived variables did not improve the efficacy of IVUS in predicting disease progression. Further research is required to investigate whether multimodality imaging combined with ESS mapping will allow more reliable vulnerable plaque detection. (Comparison of Biomatrix Versus Gazelle in ST-Elevation Myocardial Infarction [STEMI] [COMFORTABLE]; NCT00962416).

Enhancing medical data quality through data curation: a case study in primary Sjögren's syndrome.

OBJECTIVES: To address the need for automatically assessing the quality of clinical data in terms of accuracy, relevance, conformity, and completeness, through the concise development and application of an automated method which is able to automatically detect problematic fields and match clinical terms under a specific domain. METHODS: The proposed methodology involves the automated construction of three diagnostic reports that summarise valuable information regarding the types and ranges of each term in the dataset, along with the detected outliers, inconsistencies, and missing values, followed by a set of clinically relevant terms based on a reference model which serves as a set of terms which describes the domain knowledge of a disease of interest. RESULTS: A case study was conducted using anonymised data from 250 patients who were diagnosed with primary Sjögren's syndrome (pSS), yielding reliable outcomes that were highlighted for clinical evaluation. Our method was able to successfully identify 28 features with detected outliers, and unknown data types, as well as, identify outliers, missing values, similar terms, and inconsistencies within the dataset. The data standardisation method was able to match 76 out of 85 (89.41%) pSS-related terms according to a standard pSS reference model which has been introduced by the clinicians. CONCLUSIONS: Our results confirm the clinical value of the data curation method towards the improvement of the dataset quality through the precise identification of outliers, missing values, inconsistencies, and similar terms, as well as, through the automated detection of pSS-related relevant terms towards data standardisation.

HEARTEN KMS - A knowledge management system targeting the management of patients with heart failure.

The aim of this work is to present the HEARTEN Knowledge Management System, one of the core modules of the HEARTEN platform. The HEARTEN platform is an mHealth collaborative environment enabling the Heart Failure patients to self-manage the disease and remain adherent, while allowing the other ecosystem actors (healthcare professionals, caregivers, nutritionists, physical activity experts, psychologists) to monitor the patient's health progress and offer personalized, predictive and preventive disease management. The HEARTEN Knowledge Management System is a tool which provides multiple functionalities to the ecosystem actors for the assessment of the patient's condition, the estimation of the patient's adherence, the prediction of potential adverse events, the calculation of Heart Failure related scores, the extraction of statistics, the association of patient clinical and non-clinical data and the provision of alerts and suggestions. The innovation of this tool lays in the analysis of multi-parametric personal data coming from different sources, including for the first time breath and saliva biomarkers, and the use of machine learning techniques. The HEARTEN Knowledge Management System consists of nine modules. The accuracy of the KMS modules ranges from 78% to 95% depending on the module/functionality.

Medical data quality assessment: On the development of an automated framework for medical data curation.

Data quality assessment has gained attention in the recent years since more and more companies and medical centers are highlighting the importance of an automated framework to effectively manage the quality of their big data. Data cleaning, also known as data curation, lies in the heart of the data quality assessment and is a key aspect prior to the development of any data analytics services. In this work, we present the objectives, functionalities and methodological advances of an automated framework for data curation from a medical perspective. The steps towards the development of a system for data quality assessment are first described along with multidisciplinary data quality measures. A three-layer architecture which realizes these steps is then presented. Emphasis is given on the detection and tracking of inconsistencies, missing values, outliers, and similarities, as well as, on data standardization to finally enable data harmonization. A case study is conducted in order to demonstrate the applicability and reliability of the proposed framework on two well-established cohorts with clinical data related to the primary Sjögren's Syndrome (pSS). Our results confirm the validity of the proposed framework towards the automated and fast identification of outliers, inconsistencies, and highly-correlated and duplicated terms, as well as, the successful matching of more than 85% of the pSS-related medical terms in both cohorts, yielding more accurate, relevant, and consistent clinical data.

The Evolution of mHealth Solutions for Heart Failure Management.

In the last decade, the uptake of information and communication technologies and the advent of mobile internet resulted in improved connectivity and penetrated different fields of application. In particular, the adoption of the mobile devices is expected to reform the provision and delivery of healthcare, overcoming geographical, temporal, and other organizational limitations. mHealth solutions are able to provide meaningful clinical information allowing effective and efficient management of chronic diseases, such as heart failure. A variety of data can be collected, such as lifestyle, sensor/biosensor, and health-related information. The analysis of these data empowers patients and the involved ecosystem actors, improves the healthcare delivery, and facilitates the transformation of existing health services. The aim of this study is to provide an overview of (i) the current practice in the management of heart failure, (ii) the available mHealth solutions, either in the form of the commercial applications, research projects, or related studies, and (iii) the several challenges related to the patient and healthcare professionals' acceptance, the payer and provider perspective, and the regulatory constraints.

A computational approach for the estimation of heart failure patients status using saliva biomarkers.

The aim of this work is to present a computational approach for the estimation of the severity of heart failure (HF) in terms of New York Heart Association (NYHA) class and the characterization of the status of the HF patients, during hospitalization, as acute, progressive or stable. The proposed method employs feature selection and classification techniques. However, it is differentiated from the methods reported in the literature since it exploits information that biomarkers fetch. The method is evaluated on a dataset of 29 patients, through a 10-fold-cross-validation approach. The accuracy is 94 and 77% for the estimation of HF severity and the status of HF patients during hospitalization, respectively.

Predicting adherence of patients with HF through machine learning techniques.

Heart failure (HF) is a chronic disease characterised by poor quality of life, recurrent hospitalisation and high mortality. Adherence of patient to treatment suggested by the experts has been proven a significant deterrent of the above-mentioned serious consequences. However, the non-adherence rates are significantly high; a fact that highlights the importance of predicting the adherence of the patient and enabling experts to adjust accordingly patient monitoring and management. The aim of this work is to predict the adherence of patients with HF, through the application of machine learning techniques. Specifically, it aims to classify a patient not only as medication adherent or not, but also as adherent or not in terms of medication, nutrition and physical activity (global adherent). Two classification problems are addressed: (i) if the patient is global adherent or not and (ii) if the patient is medication adherent or not. About 11 classification algorithms are employed and combined with feature selection and resampling techniques. The classifiers are evaluated on a dataset of 90 patients. The patients are characterised as medication and global adherent, based on clinician estimation. The highest detection accuracy is 82 and 91% for the first and the second classification problem, respectively.

Impact of local endothelial shear stress on neointima and plaque following stent implantation in patients with ST-elevation myocardial infarction: A subgroup-analysis of the COMFORTABLE AMI-IBIS 4 trial.

BACKGROUND: Numerous studies have demonstrated an association between endothelial shear stress (ESS) and neointimal formation after stent implantation. However, the role of ESS on the composition of neointima and underlying plaque remains unclear. METHODS: Patients recruited in the Comfortable AMI-IBIS 4 study implanted with bare metal stents (BMS) or biolimus eluting stents (BES) that had biplane coronary angiography at 13 month follow-up were included in the analysis. The intravascular ultrasound virtual-histology (IVUS-VH) and the angiographic data were used to reconstruct the luminal surface, and the stent in the stented segments. Blood flow simulation was performed in the stent surface, which was assumed to represent the luminal surface at baseline, to assess the association between ESS and neointima thickness. The predominant ESS was estimated in 3-mm segments and was correlated with the amount of neointima, neointimal tissue composition, and with the changes in the underlying plaque burden and composition. RESULTS: Forty three patients (18 implanted with BMS and 25 with BES) were studied. In both stent groups negative correlations were noted between ESS and neointima thickness in BMS (P < 0.001) and BES (P = 0.002). In BMS there was a negative correlation between predominant ESS and the percentage of the neointimal necrotic core component (P = 0.015). In BES group, the limited neointima formation did not allow evaluation of the effect of ESS on its tissue characteristics. ESS did not affect vessel wall remodeling and the plaque burden and composition behind BMS (P > 0.10) and BES (P > 0.45). CONCLUSIONS: ESS determines neointimal formation in both BMS and BES and affects the composition of the neointima in BMS. Conversely, ESS does not impact the plaque behind struts irrespective of stent type throughout 13 months of follow-up.

Effect of the endothelial shear stress patterns on neointimal proliferation following drug-eluting bioresorbable vascular scaffold implantation: an optical coherence tomography study.

OBJECTIVES: This study sought to investigate the effect of endothelial shear stress (ESS) on neointimal formation following an Absorb bioresorbable vascular scaffold (BVS) (Abbott Vascular, Santa Clara, California) implantation. BACKGROUND: Cumulative evidence, derived from intravascular ultrasound-based studies, has demonstrated a strong association between local ESS patterns and neointimal formation in bare-metal stents, whereas in drug-eluting stents, there are contradictory data about the effect of ESS on the vessel wall healing process. The effect of ESS on neointimal development following a bioresorbable scaffold implantation remains unclear. METHODS: Twelve patients with an obstructive lesion in a relatively straight arterial segment, who were treated with an Absorb BVS and had serial optical coherence tomographic examination at baseline and 1-year follow-up, were included in the current analysis. The optical coherence tomographic data acquired at follow-up were used to reconstruct the scaffolded segment. Blood flow simulation was performed on the luminal surface at baseline defined by the Absorb BVS struts, and the computed ESS was related to the neointima thickness measured at 1-year follow-up. RESULTS: At baseline, the scaffolded segments were exposed to a predominantly low ESS environment (61% of the measured ESS was <1 Pa). At follow-up, the mean neointima thickness was 113 ± 45 µm, whereas the percentage scaffold volume obstruction was 13.1 ± 6.6%. A statistically significant inverse correlation was noted between baseline logarithmic transformed ESS and neointima thickness at 1-year follow-up in all studied segments (correlation coefficient range -0.140 to -0.662). Mixed linear regression analysis between baseline logarithmic transformed ESS and neointima thickness at follow-up yielded a slope of -31 µm/ln(Pa) and a y-intercept of 99 µm. CONCLUSIONS: The hemodynamic microenvironment appears to regulate neointimal response following an Absorb BVS implantation. These findings underline the role of the ESS patterns on vessel wall healing and should be taken into consideration in the design of bioresorbable devices.

Fusion of optical coherence tomographic and angiographic data for more accurate evaluation of the endothelial shear stress patterns and neointimal distribution after bioresorbable scaffold implantation: comparison with intravascular ultrasound-derived reconstructions.

Intravascular ultrasound (IVUS)-based reconstructions have been traditionally used to examine the effect of endothelial shear stress (ESS) on neointimal formation. The aim of this analysis is to compare the association between ESS and neointimal thickness (NT) in models obtained by the fusion of optical coherence tomography (OCT) and coronary angiography and in the reconstructions derived by the integration of IVUS and coronary angiography. We analyzed data from six patients implanted with an Absorb bioresorbable vascular scaffold that had biplane angiography, IVUS and OCT investigation at baseline and 6 or 12 months follow-up. The IVUS and OCT follow-up data were fused separately with the angiographic data to reconstruct the luminal morphology at baseline and follow-up. Blood flow simulation was performed on the baseline reconstructions and the ESS was related to NT. In the OCT-based reconstructions the ESS were lower compared to the IVUS-based models (1.29 ± 0.66 vs. 1.87 ± 0.66 Pa, P = 0.030). An inverse correlation was noted between the logarithmic transformed ESS and the measured NT in all the OCT-based models which was higher than the correlation reported in five of the six IVUS-derived models (-0.52 ± 0.19 Pa vs. -0.10 ± 0.04, P = 0.028). Fusion of OCT and coronary angiography appears superior to IVUS-based reconstructions; therefore it should be the method of choice for the study of the effect of the ESS on neointimal proliferation.

Segmentation of microarray images using pixel classification-comparison with clustering-based methods.

OBJECTIVE: DNA microarray technology yields expression profiles for thousands of genes, in a single hybridization experiment. The quantification of the expression level is performed using image analysis. In this paper we introduce a supervised method for the segmentation of microarray images using classification techniques. The method is able to characterize the pixels of the image as signal, background and artefact. METHODS AND MATERIAL: The proposed method includes five steps: (a) an automated gridding method which provides a cell of the image for each spot. (b) Three multichannel vector filters are employed to preprocess the raw image. (c) Features are extracted from each pixel of the image. (d) The dimension of the feature set is reduced. (e) Support vector machines are used for the classification of pixels as signal, background, artefacts. The proposed method is evaluated using both real images from the Stanford microarray database and simulated images generated by a microarray data simulator. The signal and the background pixels, which are responsible for the quantification of the expression levels, are efficiently detected. RESULTS: A quality measure (qindex) and the pixel-by-pixel accuracy are used for the evaluation of the proposed method. The obtained qindex varies from 0.742 to 0.836. The obtained accuracy for the real images is about 98%, while the accuracies for the good, normal and bad quality simulated images are 96, 93 and 71%, respectively. The proposed classification method is compared to clustering-based techniques, which have been proposed for microarray image segmentation. This comparison shows that the classification-based method reports better results, improving the performance by up to 20%. CONCLUSIONS: The proposed method can be used for segmentation of microarray images with high accuracy, indicating that segmentation can be improved using classification instead of clustering. The proposed method is supervised and it can only be used when training data are available.

A new methodology for accurate 3-dimensional coronary artery reconstruction using routine intravascular ultrasound and angiographic data: implications for widespread assessment of endothelial shear stress in humans.

AIMS: To develop and validate a new methodology that allows accurate 3-dimensional (3-D) coronary artery reconstruction using standard, simple angiographic and intravascular ultrasound (IVUS) data acquired during routine catheterisation enabling reliable assessment of the endothelial shear stress (ESS) distribution. METHODS AND RESULTS: Twenty-two patients (22 arteries: 7 LAD; 7 LCx; 8 RCA) who underwent angiography and IVUS examination were included. The acquired data were used for 3-D reconstruction using a conventional method and a new methodology that utilised the luminal 3-D centreline to place the detected IVUS borders and anatomical landmarks to estimate their orientation. The local ESS distribution was assessed by computational fluid dynamics. In corresponding consecutive 3 mm segments, lumen, plaque and ESS measurements in the 3-D models derived by the centreline approach were highly correlated to those derived from the conventional method (r>0.98 for all). The centreline methodology had a 99.5% diagnostic accuracy for identifying segments exposed to low ESS and provided similar estimations to the conventional method for the association between the change in plaque burden and ESS (centreline method: slope= -1.65%/Pa, p=0.078; conventional method: slope= -1.64%/Pa, p=0.084; p =0.69 for difference between the two methodologies). CONCLUSIONS: The centreline methodology provides geometrically correct models and permits reliable ESS computation. The ability to utilise data acquired during routine coronary angiography and IVUS examination will facilitate clinical investigation of the role of local ESS patterns in the natural history of coronary atherosclerosis.

Spot addressing for microarray images structured in hexagonal grids.

In this work, an efficient method for spot addressing in images, which are generated by the scanning of hexagonal structured microarrays, is proposed. Initially, the blocks of the image are separated using the projections of the image. Next, all the blocks of the image are processed separately for the detection of each spot. The spot addressing procedure begins with the detection of the high intensity objects, which are probably the spots of the image. Next, the Growing Concentric Hexagon algorithm, which uses the properties of the hexagonal grid, is introduced for the detection of the non-hybridized spots. Finally, the Voronoi diagram is applied to the centers of the detected spots for the gridding of the image. The method is evaluated using spots generated from the scanning of the Beadchip of Illumina, which is used for the detection of Single Nucleotide Polymorphisms in the human genome, and uses hexagonal structure for the location of the spots. For the evaluation, the detected centers for each of the spot in the image are compared to the centers of the annotation, obtaining up to 98% accuracy for the spot addressing procedure.

Polymerase chain reaction (PCR) and sequence specific oligonucleotide probes (SSOP) genotyping assay for detection of genes associated with rheumatoid arthritis and multiple sclerosis.

In this paper an assay for the detection of genes associated with rheumatoid arthritis (RA) and multiple sclerosis, using polymerase chain reaction (PCR) and sequence specific oligonucleotide probes (SSOP) is presented, in order to be further applied in a portable Lab-On-Chip (LOC) device. A substantial part of these reagents were based on the literature (11th International Histocompatibility Workshop, IHW), bearing the advantage of proven successful implementation in genotyping, while others were designed for this study. More precisely, our methodology discriminates HLA-DRB1 as DRB1*01, *04 and *10, which include shared epitope (SE) alleles associated with RA and additionally DRB1*15 allele, including DRB1*1501 associated with MS (broad genotyping method). To further present the basic elements of the assay for high resolution genotyping of SE DRB1 alleles, we provide as an example the case of HLA-DRB1*10 alleles (HLADRB1* 100101, *100102, *100103, *1002 and *1003). Regarding the methodology for developing a detection assay, for SNPs associated with RA or MS the basic steps are presented. DNA sequence data are obtained from IMGT/HLA and SNP database. Online software tools are used to define hybridization specificity of primers and probes towards human DNA, leading to hybridization patterns that uniquely designate a target allele and evaluate parameters influencing PCR efficiency. Respecting current technological limitations of autonomous molecular-based LOC systems the approach of broad genotyping of HLA-DRB1*01/*04/*10/*15 genes, is intended to be initially used, leaving, high resolution genotyping of SE alleles for future implementations. This method is easy to be updated and extended to detect additional associated loci with RA or MS.

Developing a genomic-based point-of-care diagnostic system for rheumatoid arthritis and multiple sclerosis.

In this paper the methodology of designing a genomic-based point-of-care diagnostic system composed of a microfluidic Lab-On-Chip, algorithms for microarray image information extraction and knowledge modeling of clinico-genomic patient data is presented. The data are processed by genome wide association studies for two complex diseases: rheumatoid arthritis and multiple sclerosis. Respecting current technological limitations of autonomous molecular-based Lab-On-Chip systems the approach proposed in this work aims to enhance the diagnostic accuracy of the miniaturized LOC system. By providing a decision support system based on the data mining technologies, a robust portable integrated point-of-care diagnostic assay will be implemented. Initially, the gene discovery process is described followed by the detection of the most informative SNPs associated with the diseases. The clinical data and the selected associated SNPs are modeled using data mining techniques to allow the knowledge modeling framework to provide the diagnosis for new patients performing the point-of-care examination. The microfluidic LOC device supplies the diagnostic component of the platform with a set of SNPs associated with the diseases and the ruled-based decision support system combines this genomic information with the clinical data of the patient to outcome the final diagnostic result.