Deep learning for clustering of multivariate clinical patient trajectories with missing values

Progression subtypes in Parkinson's disease identified by a data-driven multi cohort analysis

The progression of Parkinson's disease (PD) is heterogeneous across patients, affecting counseling and inflating the number of patients needed to test potential neuroprotective treatments. Moreover, disease subtypes might require different therapies. This work uses a data-driven approach to investigate how observed heterogeneity in PD can be explained by the existence of distinct PD progression subtypes. To derive stable PD progression subtypes in an unbiased manner, we analyzed multimodal longitudinal data from three large PD cohorts and performed extensive cross-cohort validation. A latent time joint mixed-effects model (LTJMM) was used to align patients on a common disease timescale. Progression subtypes were identified by variational deep embedding with recurrence (VaDER). In each cohort, we identified a fast-progressing and a slow-progressing subtype, reflected by different patterns of motor and non-motor symptoms progression, survival rates, treatment response, features extracted from DaTSCAN imaging and digital gait assessments, education, and Alzheimer's disease pathology. Progression subtypes could be predicted with ROC-AUC up to 0.79 for individual patients when a one-year observation period was used for model training. Simulations demonstrated that enriching clinical trials with fast-progressing patients based on these predictions can reduce the required cohort size by 43%. Our results show that heterogeneity in PD can be explained by two distinct subtypes of PD progression that are stable across cohorts. These subtypes align with the brain-first vs. body-first concept, which potentially provides a biological explanation for subtype differences. Our predictive models will enable clinical trials with significantly lower sample sizes by enriching fast-progressing patients.

Identify adolescents' help-seeking intention on suicide through self- and caregiver's assessments of psychobehavioral problems: deep clustering of the Tokyo TEEN Cohort study

Background

Psychopathological and behavioral problems in adolescence are highly comorbid, making their developmental trajectories complex and unclear partly due to technical limitations. We aimed to classify these trajectories using deep learning and identify predictors of cluster membership.

Methods

We conducted a population-based cohort study on 3171 adolescents from three Tokyo municipalities, with 2344 pairs of adolescents and caregivers participating at all four timepoints (ages 10, 12, 14, and 16) from 2012 to 2021. Adolescent psychopathological and behavioral problems were assessed by using self-report questionnaires. Both adolescents and caregivers assessed depression/anxiety and psychotic-like experiences. Caregivers assessed obsession/compulsion, dissociation, sociality problem, hyperactivity/inattention, conduct problem, somatic symptom, and withdrawal. Adolescents assessed desire for slimness, self-harm, and suicidal ideation. These trajectories were clustered with variational deep embedding with recurrence, and predictors were explored using multinomial logistic regression.

Findings

Five clusters were identified: unaffected (60.5%), minimal problems; internalizing (16.2%), persistent or worsening internalizing problems; discrepant (9.9%), subjective problems overlooked by caregivers; externalizing (9.6%), persistent externalizing problems; and severe (3.9%), chronic severe problems across symptoms. Stronger autistic traits and experience of bullying victimization commonly predicted the four "affected" clusters. The discrepant cluster, showing the highest risks for self-harm and suicidal ideation, was predicted by avoiding help-seeking for depression. The severe cluster predictors included maternal smoking during pregnancy, not bullying others, caregiver's psychological distress, and adolescent's dissatisfaction with family.

Interpretation

Approximately 40% of adolescents were classified as "affected" clusters. Proactive societal attention is warranted toward adolescents in the discrepant cluster whose suicidality is overlooked and who have difficulty seeking help.

Funding

Japan Ministry of Health, Labor and Welfare, Japan Agency for Medical Research and Development, and Japan Science and Technology Agency.

Artificial intelligence-based clustering and characterization of Parkinson's disease trajectories

Parkinson's disease (PD) is a highly heterogeneous disease both with respect to arising symptoms and its progression over time. This hampers the design of disease modifying trials for PD as treatments which would potentially show efficacy in specific patient subgroups could be considered ineffective in a heterogeneous trial cohort. Establishing clusters of PD patients based on their progression patterns could help to disentangle the exhibited heterogeneity, highlight clinical differences among patient subgroups, and identify the biological pathways and molecular players which underlie the evident differences. Further, stratification of patients into clusters with distinct progression patterns could help to recruit more homogeneous trial cohorts. In the present work, we applied an artificial intelligence-based algorithm to model and cluster longitudinal PD progression trajectories from the Parkinson's Progression Markers Initiative. Using a combination of six clinical outcome scores covering both motor and non-motor symptoms, we were able to identify specific clusters of PD that showed significantly different patterns of PD progression. The inclusion of genetic variants and biomarker data allowed us to associate the established progression clusters with distinct biological mechanisms, such as perturbations in vesicle transport or neuroprotection. Furthermore, we found that patients of identified progression clusters showed significant differences in their responsiveness to symptomatic treatment. Taken together, our work contributes to a better understanding of the heterogeneity encountered when examining and treating patients with PD, and points towards potential biological pathways and genes that could underlie those differences.

Machine learning within the Parkinson's progression markers initiative: Review of the current state of affairs

The Parkinson's Progression Markers Initiative (PPMI) has collected more than a decade's worth of longitudinal and multi-modal data from patients, healthy controls, and at-risk individuals, including imaging, clinical, cognitive, and 'omics' biospecimens. Such a rich dataset presents unprecedented opportunities for biomarker discovery, patient subtyping, and prognostic prediction, but it also poses challenges that may require the development of novel methodological approaches to solve. In this review, we provide an overview of the application of machine learning methods to analyzing data from the PPMI cohort. We find that there is significant variability in the types of data, models, and validation procedures used across studies, and that much of what makes the PPMI data set unique (multi-modal and longitudinal observations) remains underutilized in most machine learning studies. We review each of these dimensions in detail and provide recommendations for future machine learning work using data from the PPMI cohort.

ImputeGAN: Generative Adversarial Network for Multivariate Time Series Imputation

Since missing values in multivariate time series data are inevitable, many researchers have come up with methods to deal with the missing data. These include case deletion methods, statistics-based imputation methods, and machine learning-based imputation methods. However, these methods cannot handle temporal information, or the complementation results are unstable. We propose a model based on generative adversarial networks (GANs) and an iterative strategy based on the gradient of the complementary results to solve these problems. This ensures the generalizability of the model and the reasonableness of the complementation results. We conducted experiments on three large-scale datasets and compare them with traditional complementation methods. The experimental results show that imputeGAN outperforms traditional complementation methods in terms of accuracy of complementation.

Clustering of trauma patients based on longitudinal data and the application of machine learning to predict recovery

Predicting recovery after trauma is important to provide patients a perspective on their estimated future health, to engage in shared decision making and target interventions to relevant patient groups. In the present study, several unsupervised techniques are employed to cluster patients based on longitudinal recovery profiles. Subsequently, these data-driven clusters were assessed on clinical validity by experts and used as targets in supervised machine learning models. We present a formalised analysis of the obtained clusters that incorporates evaluation of (i) statistical and machine learning metrics, (ii) clusters clinical validity with descriptive statistics and medical expertise. Clusters quality assessment revealed that clusters obtained through a Bayesian method (High Dimensional Supervised Classification and Clustering) and a Deep Gaussian Mixture model, in combination with oversampling and a Random Forest for supervised learning of the cluster assignments provided among the most clinically sensible partitioning of patients. Other methods that obtained higher classification accuracy suffered from cluster solutions with large majority classes or clinically less sensible classes. Models that used just physical or a mix of physical and psychological outcomes proved to be among the most sensible, suggesting that clustering on psychological outcomes alone yields recovery profiles that do not conform to known risk factors.

Generation of realistic synthetic data using Multimodal Neural Ordinary Differential Equations

Individual organizations, such as hospitals, pharmaceutical companies, and health insurance providers, are currently limited in their ability to collect data that are fully representative of a disease population. This can, in turn, negatively impact the generalization ability of statistical models and scientific insights. However, sharing data across different organizations is highly restricted by legal regulations. While federated data access concepts exist, they are technically and organizationally difficult to realize. An alternative approach would be to exchange synthetic patient data instead. In this work, we introduce the Multimodal Neural Ordinary Differential Equations (MultiNODEs), a hybrid, multimodal AI approach, which allows for generating highly realistic synthetic patient trajectories on a continuous time scale, hence enabling smooth interpolation and extrapolation of clinical studies. Our proposed method can integrate both static and longitudinal data, and implicitly handles missing values. We demonstrate the capabilities of MultiNODEs by applying them to real patient-level data from two independent clinical studies and simulated epidemiological data of an infectious disease.

Identification of robust deep neural network models of longitudinal clinical measurements

Deep learning (DL) from electronic health records holds promise for disease prediction, but systematic methods for learning from simulated longitudinal clinical measurements have yet to be reported. We compared nine DL frameworks using simulated body mass index (BMI), glucose, and systolic blood pressure trajectories, independently isolated shape and magnitude changes, and evaluated model performance across various parameters (e.g., irregularity, missingness). Overall, discrimination based on variation in shape was more challenging than magnitude. Time-series forest-convolutional neural networks (TSF-CNN) and Gramian angular field(GAF)-CNN outperformed other approaches (P < 0.05) with overall area-under-the-curve (AUCs) of 0.93 for both models, and 0.92 and 0.89 for variation in magnitude and shape with up to 50% missing data. Furthermore, in a real-world assessment, the TSF-CNN model predicted T2D with AUCs reaching 0.72 using only BMI trajectories. In conclusion, we performed an extensive evaluation of DL approaches and identified robust modeling frameworks for disease prediction based on longitudinal clinical measurements.

Leveraging the Potential of Digital Technology for Better Individualized Treatment of Parkinson's Disease

Recent years have witnessed a strongly increasing interest in digital technology within medicine (sensor devices, specific smartphone apps) and specifically also neurology. Quantitative measures derived from digital technology could provide Digital Biomarkers (DMs) enabling a quantitative and continuous monitoring of disease symptoms, also outside clinics. This includes the possibility to continuously and sensitively monitor the response to treatment, hence opening the opportunity to adapt medication pathways quickly. In addition, DMs may in the future allow early diagnosis, stratification of patient subgroups and prediction of clinical outcomes. Thus, DMs could complement or in certain cases even replace classical examiner-based outcome measures and molecular biomarkers measured in cerebral spinal fluid, blood, urine, saliva, or other body liquids. Altogether, DMs could play a prominent role in the emerging field of precision medicine. However, realizing this vision requires dedicated research. First, advanced data analytical methods need to be developed and applied, which extract candidate DMs from raw signals. Second, these candidate DMs need to be validated by (a) showing their correlation to established clinical outcome measures, and (b) demonstrating their diagnostic and/or prognostic value compared to established biomarkers. These points again require the use of advanced data analytical methods, including machine learning. In addition, the arising ethical, legal and social questions associated with the collection and processing of sensitive patient data and the use of machine learning methods to analyze these data for better individualized treatment of the disease, must be considered thoroughly. Using Parkinson's Disease (PD) as a prime example of a complex multifactorial disorder, the purpose of this article is to critically review the current state of research regarding the use of DMs, discuss open challenges and highlight emerging new directions.

Unraveling the heterogeneity in Alzheimer's disease progression across multiple cohorts and the implications for data-driven disease modeling

INTRODUCTION

Given study-specific inclusion and exclusion criteria, Alzheimer's disease (AD) cohort studies effectively sample from different statistical distributions. This heterogeneity can propagate into cohort-specific signals and subsequently bias data-driven investigations of disease progression patterns.

METHODS

We built multi-state models for six independent AD cohort datasets to statistically compare disease progression patterns across them. Additionally, we propose a novel method for clustering cohorts with regard to their progression signals.

RESULTS

We identified significant differences in progression patterns across cohorts. Models trained on cohort data learned cohort-specific effects that bias their estimations. We demonstrated how six cohorts relate to each other regarding their disease progression.

DISCUSSION

Heterogeneity in cohort datasets impedes the reproducibility of data-driven results and validation of progression models generated on single cohorts. To ensure robust scientific insights, it is advisable to externally validate results in independent cohort datasets. The proposed clustering assesses the comparability of cohorts in an unbiased, data-driven manner.

A deep learning approach for predicting severity of COVID-19 patients using a parsimonious set of laboratory markers

The SARS-CoV-2 virus has caused tremendous healthcare burden worldwide. Our focus was to develop a practical and easy-to-deploy system to predict the severe manifestation of disease in patients with COVID-19 with an aim to assist clinicians in triage and treatment decisions. Our proposed predictive algorithm is a trained artificial intelligence-based network using 8,427 COVID-19 patient records from four healthcare systems. The model provides a severity risk score along with likelihoods of various clinical outcomes, namely ventilator use and mortality. The trained model using patient age and nine laboratory markers has the prediction accuracy with an area under the curve (AUC) of 0.78, 95% CI: 0.77–0.82, and the negative predictive value NPV of 0.86, 95% CI: 0.84–0.88 for the need to use a ventilator and has an accuracy with AUC of 0.85, 95% CI: 0.84–0.86, and the NPV of 0.94, 95% CI: 0.92–0.96 for predicting in-hospital 30-day mortality.

•

Algorithm using 9 laboratory markers & age may predict severity in patients with COVID-19

•

Model was trained and tested on a multicenter sample of 10,937 patients

•

Algorithm can predict ventilator use (NPV, 0.86) and mortality (NPV, 0.94)

•

High NPV suggests utility as an adjunct to aid in triaging of patients with COVID-19

Deep Time-Series Clustering: A Review

We present a comprehensive, detailed review of time-series data analysis, with emphasis on deep time-series clustering (DTSC), and a case study in the context of movement behavior clustering utilizing the deep clustering method. Specifically, we modified the DCAE architectures to suit time-series data at the time of our prior deep clustering work. Lately, several works have been carried out on deep clustering of time-series data. We also review these works and identify state-of-the-art, as well as present an outlook on this important field of DTSC from five important perspectives.

Disease Modelling of Cognitive Outcomes and Biomarkers in the European Prevention of Alzheimer's Dementia Longitudinal Cohort

A key challenge for the secondary prevention of Alzheimer’s dementia is the need to identify individuals early on in the disease process through sensitive cognitive tests and biomarkers. The European Prevention of Alzheimer’s Dementia (EPAD) consortium recruited participants into a longitudinal cohort study with the aim of building a readiness cohort for a proof-of-concept clinical trial and also to generate a rich longitudinal data-set for disease modelling. Data have been collected on a wide range of measurements including cognitive outcomes, neuroimaging, cerebrospinal fluid biomarkers, genetics and other clinical and environmental risk factors, and are available for 1,828 eligible participants at baseline, 1,567 at 6 months, 1,188 at one-year follow-up, 383 at 2 years, and 89 participants at three-year follow-up visit. We novelly apply state-of-the-art longitudinal modelling and risk stratification approaches to these data in order to characterise disease progression and biological heterogeneity within the cohort. Specifically, we use longitudinal class-specific mixed effects models to characterise the different clinical disease trajectories and a semi-supervised Bayesian clustering approach to explore whether participants can be stratified into homogeneous subgroups that have different patterns of cognitive functioning evolution, while also having subgroup-specific profiles in terms of baseline biomarkers and longitudinal rate of change in biomarkers.

Deep learning for brain disorders: from data processing to disease treatment

In order to reach precision medicine and improve patients' quality of life, machine learning is increasingly used in medicine. Brain disorders are often complex and heterogeneous, and several modalities such as demographic, clinical, imaging, genetics and environmental data have been studied to improve their understanding. Deep learning, a subpart of machine learning, provides complex algorithms that can learn from such various data. It has become state of the art in numerous fields, including computer vision and natural language processing, and is also growingly applied in medicine. In this article, we review the use of deep learning for brain disorders. More specifically, we identify the main applications, the concerned disorders and the types of architectures and data used. Finally, we provide guidelines to bridge the gap between research studies and clinical routine.

Functional data analysis and prediction tools for continuous glucose-monitoring studies

Introduction

To identify phenotypes of type 1 diabetes based on glucose curves from continuous glucose-monitoring (CGM) using functional data (FD) analysis to account for longitudinal glucose patterns. We present a reliable prediction model that can accurately predict glycemic levels based on past data collected from the CGM sensor and real-time risk of hypo-/hyperglycemic for individuals with type 1 diabetes.

Methods

A longitudinal cohort study of 443 type 1 diabetes patients with CGM data from a completed trial. The FD analysis approach, sparse functional principal components (FPCs) analysis was used to identify phenotypes of type 1 diabetes glycemic variation. We employed a nonstationary stochastic linear mixed-effects model (LME) that accommodates between-patient and within-patient heterogeneity to predict glycemic levels and real-time risk of hypo-/hyperglycemic by creating specific target functions for these excursions.

Results

The majority of the variation (73%) in glucose trajectories was explained by the first two FPCs. Higher order variation in the CGM profiles occurred during weeknights, although variation was higher on weekends. The model has low prediction errors and yields accurate predictions for both glucose levels and real-time risk of glycemic excursions.

Conclusions

By identifying these distinct longitudinal patterns as phenotypes, interventions can be targeted to optimize type 1 diabetes management for subgroups at the highest risk for compromised long-term outcomes such as cardiac disease or stroke. Further, the estimated change/variability in an individual's glucose trajectory can be used to establish clinically meaningful and patient-specific thresholds that, when coupled with probabilistic predictive inference, provide a useful medical-monitoring tool.

Differences in cohort study data affect external validation of artificial intelligence models for predictive diagnostics of dementia - lessons for translation into clinical practice

Artificial intelligence (AI) approaches pose a great opportunity for individualized, pre-symptomatic disease diagnosis which plays a key role in the context of personalized, predictive, and finally preventive medicine (PPPM). However, to translate PPPM into clinical practice, it is of utmost importance that AI-based models are carefully validated. The validation process comprises several steps, one of which is testing the model on patient-level data from an independent clinical cohort study. However, recruitment criteria can bias statistical analysis of cohort study data and impede model application beyond the training data. To evaluate whether and how data from independent clinical cohort studies differ from each other, this study systematically compares the datasets collected from two major dementia cohorts, namely, the Alzheimer's Disease Neuroimaging Initiative (ADNI) and AddNeuroMed. The presented comparison was conducted on individual feature level and revealed significant differences among both cohorts. Such systematic deviations can potentially hamper the generalizability of results which were based on a single cohort dataset. Despite identified differences, validation of a previously published, ADNI trained model for prediction of personalized dementia risk scores on 244 AddNeuroMed subjects was successful: External validation resulted in a high prediction performance of above 80% area under receiver operator characteristic curve up to 6 years before dementia diagnosis. Propensity score matching identified a subset of patients from AddNeuroMed, which showed significantly smaller demographic differences to ADNI. For these patients, an even higher prediction performance was achieved, which demonstrates the influence systematic differences between cohorts can have on validation results. In conclusion, this study exposes challenges in external validation of AI models on cohort study data and is one of the rare cases in the neurology field in which such external validation was performed. The presented model represents a proof of concept that reliable models for personalized predictive diagnostics are feasible, which, in turn, could lead to adequate disease prevention and hereby enable the PPPM paradigm in the dementia field.

Data science in neurodegenerative disease: its capabilities, limitations, and perspectives

PURPOSE OF REVIEW

With the advancement of computational approaches and abundance of biomedical data, a broad range of neurodegenerative disease models have been developed. In this review, we argue that computational models can be both relevant and useful in neurodegenerative disease research and although the current established models have limitations in clinical practice, artificial intelligence has the potential to overcome deficiencies encountered by these models, which in turn can improve our understanding of disease.

RECENT FINDINGS

In recent years, diverse computational approaches have been used to shed light on different aspects of neurodegenerative disease models. For example, linear and nonlinear mixed models, self-modeling regression, differential equation models, and event-based models have been applied to provide a better understanding of disease progression patterns and biomarker trajectories. Additionally, the Cox-regression technique, Bayesian network models, and deep-learning-based approaches have been used to predict the probability of future incidence of disease, whereas nonnegative matrix factorization, nonhierarchical cluster analysis, hierarchical agglomerative clustering, and deep-learning-based approaches have been employed to stratify patients based on their disease subtypes. Furthermore, the interpretation of neurodegenerative disease data is possible through knowledge-based models which use prior knowledge to complement data-driven analyses. These knowledge-based models can include pathway-centric approaches to establish pathways perturbed in a given condition, as well as disease-specific knowledge maps, which elucidate the mechanisms involved in a given disease. Collectively, these established models have revealed high granular details and insights into neurodegenerative disease models.

SUMMARY

In conjunction with increasingly advanced computational approaches, a wide spectrum of neurodegenerative disease models, which can be broadly categorized into data-driven and knowledge-driven, have been developed. We review the state of the art data and knowledge-driven models and discuss the necessary steps which are vital to bring them into clinical application.

Challenges of Integrative Disease Modeling in Alzheimer's Disease

Dementia-related diseases like Alzheimer's Disease (AD) have a tremendous social and economic cost. A deeper understanding of its underlying pathophysiologies may provide an opportunity for earlier detection and therapeutic intervention. Previous approaches for characterizing AD were targeted at single aspects of the disease. Yet, due to the complex nature of AD, the success of these approaches was limited. However, in recent years, advancements in integrative disease modeling, built on a wide range of AD biomarkers, have taken a global view on the disease, facilitating more comprehensive analysis and interpretation. Integrative AD models can be sorted in two primary types, namely hypothetical models and data-driven models. The latter group split into two subgroups: (i) Models that use traditional statistical methods such as linear models, (ii) Models that take advantage of more advanced artificial intelligence approaches such as machine learning. While many integrative AD models have been published over the last decade, their impact on clinical practice is limited. There exist major challenges in the course of integrative AD modeling, namely data missingness and censoring, imprecise human-involved priori knowledge, model reproducibility, dataset interoperability, dataset integration, and model interpretability. In this review, we highlight recent advancements and future possibilities of integrative modeling in the field of AD research, showcase and discuss the limitations and challenges involved, and finally, propose avenues to address several of these challenges.