Machine learning‐assisted point‐of‐care diagnostics for cardiovascular healthcare

Figures

• 

  FIGURE 1. Point‐of‐care (POC) devices for cardiovascular healthcare. (a) Overview of present and emerging wearable technologies. Reproduced with permission. 45 Copyright 2021, Springer Nature. (b) Nanoengineered multichannel immunosensors with unsupervised clustering to predict acute thrombosis. Reproduced with permission. 52 Copyright 2024, American Association for the Advancement of Science (AAAS). (c) Magnetoelastic generator (MEG)‐based e‐textiles have been designed for advanced biomechanical sensing applications. Reproduced with permission. 15 Copyright 2021, American Chemical Society (ACS). (d) Nanosensor technology, where silicon nanowires, arranged on microchips and coated with specific molecules, detect analytes through changes in electrical conductance, allowing for instant detection and reuse upon dissociation of the analytes. Reproduced with permission. 20 Copyright 2011, Springer Nature. (e) Depicts the synthesis process for various metal–organic frameworks (MAFs) and the employed digital sensing strategy. Reproduced with permission. 55 Copyright 2022, AAAS.

• 

  FIGURE 2. Overview of machine learning classifications and algorithms. (a) Key machine learning algorithms. Reproduced with permission. 65 Copyright 2020, American Chemical Society (ACS). (b) Comparative analysis of supervised, unsupervised, and semi‐supervised learning paradigms. (c) Visualization of input data: Hidden layer 1, hidden layer 2, and output. Reproduced with permission. 74 Copyright 2019, Springer Nature. (d) An example of a large‐scale neural network that takes various data types as input, learning effective feature representations in its lower‐level towers for each data type. Reproduced with permission. 74 Copyright 2019, Springer Nature. (e) Illustration of the architecture of ResNet‐18 model. (f) CNNs in medical applications. Reproduced with permission. 74 Copyright 2019, Springer Nature.

• 

  FIGURE 3. Unsupervised learning in cardiovascular healthcare. (a) scRNA‐seq of AmbCardiac neurons identifies a genetic signature of brainstem parasympathetic neurons. Reproduced with permission. 91 Copyright 2022, Springer Nature. (b) Using the unsupervised uniform manifold approximation and projection (UMAP) technique to cluster 220,752 nuclei and 49,723 cells. Reproduced with permission. 93 Copyright 2022, Springer Nature. (c) CyTOF can be employed to examine the diverse cellular categories and prevalence of immune cells among patients, utilizing impartial methods like Louvain clustering. Reproduced with permission. 95 Copyright 2022, Springer Nature. (d) A 10‐μm cryo‐section was made to extract nuclei from the tissue adjacent to it, followed by fluorescence‐activated nuclei sorting (FANS) for single‐nucleus RNA sequencing (snRNA‐seq) and single‐nucleus ATAC sequencing (snATAC‐seq). Reproduced with permission. 97 Copyright 2022, Springer Nature.

• 

  FIGURE 4. Machine learning‐assisted point‐of‐care (POC) devices for cardiovascular healthcare. (a) Design of a textile cardiovascular monitoring system, featuring a self‐powered triboelectric sensor that wirelessly transmits cardiovascular signals to a cellphone. Reproduced with permission. 114 Copyright 2021, Wiley‐VCH. (b) The integration of artificial intelligence (AI) in clinical practice is being facilitated by modern electrocardiogram (ECG) technologies, which enable the storage, transfer, processing, and analysis of large volumes of digital data to meet diverse patient needs. Reproduced with permission. 117 Copyright 2021, Springer Nature. (c) Unsupervised machine learning is used for dimensionality reduction and k‐means clustering to identify two subgroups of cardiac resynchronization therapy (CRT) patients based on ECG QRS complex waveforms. Reproduced with permission. 118 Copyright 2020, American Heart Association (AHA). (d) Emerging Mobile Health (M‐Health) systems, leveraging advanced technologies such as deep learning, artificial intelligence, cloud computing, and big models. Reproduced with permission. 124 Copyright 2020, Elsevier. (e) Machine learning‐powered multimodal…

• 

  FIGURE 5. Outlook of POC devices with machine learning for cardiovascular healthcare. (a) The figure outlines a progression from early, to intermediate, to late fusion stages. This shows the flow of data from a centralized information repository, through the computational framework, and culminating in the analytical results. Reproduced with permission. 139 Copyright 2022, Springer Nature. (b) Health data analysis enables real‐time tracking of individual health and provide preemptive medical advice, while multidisciplinary collaboration on interoperability, standards, and regulations will help bridge the gap between laboratory research and clinical practice, advancing translational research. Reproduced with permission. 15 Copyright 2021, American Chemical Society (ACS). (c) Various types of data and their potential roles in multimodal biomedical AI. 140 Copyright 2022, Springer Nature. (d) Schematic illustration showcasing the connections and interplay among data, algorithms, participants, and methods within the sphere of secure and confidential AI. Reproduced with permission. 163 Copyright 2020, Springer Nature.