Multimodal biomedical AI

Acosta, Julián; Falcone, Guido J; Rajpurkar, Pranav; Topol, Eric J.

doi:10.1038/s41591-022-01981-2

Cited by 458 publications

(207 citation statements)

References 151 publications

Supporting

Mentioning

205

Contrasting

Unclassified

Order By: Relevance

“…Monitoring : wearable devices offer a way to continuously monitor health parameters such as heart rate and heart rhythm, etc., in a user-friendly, non-invasive way. The continuous monitoring of physiological parameters then offers a potential solution to more timely access to CVD-based healthcare [ 9 , 10 , 14 , 47 , 50 , 52 , 53 , 59 , 66 ].…”

Section: Discussionmentioning

confidence: 99%

“…Other issues that need to be considered include how to separate actionable data from noise to improve data collecting accuracy; the interpretability of the AI algorithms may also raise challenges for physicians who are expected to diagnose the results of the data from wearables, as well as the challenges of integrating data from wearable devices with clinical data. Finally, the use of AI and wearable devices in the detection diagnosis and prediction of CVD will require wider adoption by clinicians and patients if they are to become more mainstream in the field [ 8 , 9 , 18 , 65 , 109 , 110 , 116 ].…”

Section: Discussionmentioning

confidence: 99%

“…The early identification and prediction of CVD is critical for improving healthcare outcomes, but this remains challenging. The development of a variety of sensors, communication networks, and wearable and portable devices has enabled the acquisition of multiple forms of data that have the potential to improve the diagnosis of CVD [ 7 , 8 , 9 , 10 , 11 , 12 ]. The manual analysis of often complex time-series data is difficult, requiring the development of algorithms that analyse such data to potentially enhance diagnosis.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Applying Artificial Intelligence to Wearable Sensor Data to Diagnose and Predict Cardiovascular Disease: A Review

Huang

Wang

Ramsey

et al. 2022

Sensors

View full text Add to dashboard Cite

Cardiovascular disease (CVD) is the world’s leading cause of mortality. There is significant interest in using Artificial Intelligence (AI) to analyse data from novel sensors such as wearables to provide an earlier and more accurate prediction and diagnosis of heart disease. Digital health technologies that fuse AI and sensing devices may help disease prevention and reduce the substantial morbidity and mortality caused by CVD worldwide. In this review, we identify and describe recent developments in the application of digital health for CVD, focusing on AI approaches for CVD detection, diagnosis, and prediction through AI models driven by data collected from wearables. We summarise the literature on the use of wearables and AI in cardiovascular disease diagnosis, followed by a detailed description of the dominant AI approaches applied for modelling and prediction using data acquired from sensors such as wearables. We discuss the AI algorithms and models and clinical applications and find that AI and machine-learning-based approaches are superior to traditional or conventional statistical methods for predicting cardiovascular events. However, further studies evaluating the applicability of such algorithms in the real world are needed. In addition, improvements in wearable device data accuracy and better management of their application are required. Lastly, we discuss the challenges that the introduction of such technologies into routine healthcare may face.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Applying Artificial Intelligence to Wearable Sensor Data to Diagnose and Predict Cardiovascular Disease: A Review

Huang

Wang

Ramsey

et al. 2022

Sensors

View full text Add to dashboard Cite

show abstract

“…Regression analysis is also part of data-driven methods, but these are only appropriate for linear predictions ( Hemphill et al, 2011 ; Volovici et al, 2022 ), whereas time series of different modalities include multiple features (dimensions) and interactions. For these complex interactions, model-based methods are more appropriate ( Hemphill et al, 2011 ; Acosta et al, 2022 ). We included an explorative study using hierarchical clustering ( Rajagopalan et al, 2022 ).…”

Section: Discussionmentioning

confidence: 99%

Cerebral multimodality monitoring in adult neurocritical care patients with acute brain injury: A narrative review

et al. 2022

View full text Add to dashboard Cite

Cerebral multimodality monitoring (MMM) is, even with a general lack of Class I evidence, increasingly recognized as a tool to support clinical decision-making in the neuroscience intensive care unit (NICU). However, literature and guidelines have focused on unimodal signals in a specific form of acute brain injury. Integrating unimodal signals in multiple signal monitoring is the next step for clinical studies and patient care. As such, we aimed to investigate the recent application of MMM in studies of adult patients with traumatic brain injury (TBI), subarachnoid hemorrhage (SAH), intracerebral hemorrhage (ICH), acute ischemic stroke (AIS), and hypoxic ischemic brain injury following cardiac arrest (HIBI). We identified continuous or daily updated monitoring modalities and summarized the monitoring setting, study setting, and clinical characteristics. In addition, we discussed clinical outcome in intervention studies. We identified 112 MMM studies, including 11 modalities, over the last 7 years (2015–2022). Fifty-eight studies (52%) applied only two modalities. Most frequently combined were ICP monitoring (92 studies (82%)) together with PbtO2 (63 studies (56%). Most studies included patients with TBI (59 studies) or SAH (53 studies). The enrollment period of 34 studies (30%) took more than 5 years, whereas the median sample size was only 36 patients (q1- q3, 20–74). We classified studies as either observational (68 studies) or interventional (44 studies). The interventions were subclassified as systemic (24 studies), cerebral (10 studies), and interventions guided by MMM (11 studies). We identified 20 different systemic or cerebral interventions. Nine (9/11, 82%) of the MMM-guided studies included clinical outcome as an endpoint. In 78% (7/9) of these MMM-guided intervention studies, a significant improvement in outcome was demonstrated in favor of interventions guided by MMM. Clinical outcome may be improved with interventions guided by MMM. This strengthens the belief in this application, but further interdisciplinary collaborations are needed to overcome the heterogeneity, as illustrated in the present review. Future research should focus on increasing sample sizes, improved data collection, refining definitions of secondary injuries, and standardized interventions. Only then can we proceed with complex outcome studies with MMM-guided treatment.

show abstract

“…An additional benefit of the DL methods are their inherent capabilities to integrate multi-modal data. In the future, personalized medicine will likely be realized by combining genomics, proteomics, imaging, textual descriptions, biochemical measurements and various other factors that describe or contribute to a person's medical state and trajectory 37,48 . Therefore, methods that allow for such multimodal integration are expected to become pervasive in personalized medicine.…”

Section: Discussionmentioning

confidence: 99%

Improved prediction of blood biomarkers using deep learning

Sigurdsson

Ravn

Winther

et al. 2022

Preprint

View full text Add to dashboard Cite

Blood and urine biomarkers are an essential part of modern medicine, not only for diagnosis, but also for their direct influence on disease. Many biomarkers have a genetic component, and they have been studied extensively with genome-wide association studies (GWAS) and methods that compute polygenic scores (PGSs). However, these methods generally assume both an additive allelic model and an additive genetic architecture for the target outcome, and thereby risk not capturing non-linear allelic effects nor epistatic interactions. Here, we trained and evaluated deep-learning (DL) models for PGS prediction of 34 blood and urine biomarkers in the UK Biobank cohort, and compared them to linear methods. For lipid traits, the DL models greatly outperformed the linear methods, which we found to be consistent across diverse populations. Furthermore, the DL models captured non-linear effects in covariates, non-additive genotype (allelic) effects, and epistatic interactions between SNPs. Finally, when using only genome-wide significant SNPs from GWAS, the DL models performed equally well or better for all 34 traits tested. Our findings suggest that DL can serve as a valuable addition to existing methods for genotype-phenotype modelling in the era of increasing data availability.

show abstract

Multimodal biomedical AI

Cited by 458 publications

References 151 publications

Applying Artificial Intelligence to Wearable Sensor Data to Diagnose and Predict Cardiovascular Disease: A Review

Applying Artificial Intelligence to Wearable Sensor Data to Diagnose and Predict Cardiovascular Disease: A Review

Cerebral multimodality monitoring in adult neurocritical care patients with acute brain injury: A narrative review

Improved prediction of blood biomarkers using deep learning

Contact Info

Product

Resources

About