Bioscience Methods 2024, Vol.15, No.3, 124-138 http://bioscipublisher.com/index.php/bm 129 Supervised learning algorithms are trained on labeled data, where the input-output pairs are known. In cardiovascular diagnostics, supervised learning can be used to predict the onset of hypertension by analyzing patient data such as blood pressure readings, ECG signals, and demographic information (Li et al., 2022; Visco et al., 2023). For example, an AI algorithm developed for predicting pulmonary hypertension (PH) using ECG data demonstrated high accuracy, with an area under the receiver operating characteristic curve (AUC) of 0.859 and 0.902 for internal and external validation, respectively (Kwon et al., 2020). Unsupervised learning algorithms, on the other hand, do not require labeled data and are used to identify hidden patterns within the data. These algorithms can help in phenotyping complex cardiovascular diseases by clustering patients with similar characteristics, which can lead to more personalized treatment plans (Shameer et al., 2018). Reinforcement learning algorithms learn by interacting with the environment and receiving feedback in the form of rewards or penalties. In cardiovascular medicine, these algorithms can be used to optimize treatment strategies by continuously learning from patient outcomes and adjusting the treatment plans accordingly (Shameer et al., 2018). 4.3 Deep learning approaches in hhd detection Deep learning (DL), a subset of machine learning, involves the use of neural networks with multiple layers to model complex relationships within the data. DL has shown significant promise in the detection and diagnosis of hypertensive heart disease (HHD) due to its ability to process and analyze large volumes of medical data with high accuracy. One notable application of DL in HHD detection is the use of convolutional neural networks (CNNs) to analyze MRI native T1 maps for differentiating between hypertrophic cardiomyopathy (HCM) and HHD. A study demonstrated that a DL model based on T1 mapping outperformed traditional methods, achieving an AUC of 0.830 compared to 0.545 for native T1 and 0.800 for radiomics (Wang et al., 2023). This highlights the potential of DL in improving diagnostic accuracy and reducing the need for invasive procedures. DL is also being used to develop digital biomarkers for hypertension and cardiovascular risk stratification. For instance, a deep learning model (HTN-AI) was trained to detect hypertension and stratify the risk of hypertension-associated cardiovascular diseases using 12-lead ECGs. The model demonstrated high discriminatory power, with an AUC of 0.791 and 0.762 in different test samples, and was significantly associated with the risk of incident cardiovascular events such as heart failure, myocardial infarction, and stroke (Al-Alusi et al., 2023). Furthermore, DL models have been applied to wearable device data for continuous monitoring and early detection of cardiovascular conditions. A systematic review and meta-analysis found that deep neural networks showed superior performance in detecting atrial fibrillation from wearable device data, with a meta-analyzed AUC of 0.981 compared to 0.961 for conventional ML algorithms (Lee et al., 2022). In conclusion, AI-based diagnostic systems, particularly those leveraging ML and DL algorithms, hold great potential in enhancing the detection and management of hypertensive heart disease. These systems can analyze diverse data sources, provide accurate predictions, and support personalized treatment plans, ultimately improving patient outcomes. However, challenges such as data privacy, algorithm transparency, and the need for further methodological development must be addressed to fully realize the benefits of AI in cardiovascular diagnostics. 5 Key Components of AI-Based Diagnostic Systems for HHD 5.1 Data acquisition and preprocessing Data acquisition is the foundational step in developing AI-based diagnostic systems for hypertensive heart disease (HHD). This involves collecting high-quality, relevant data from various sources such as electronic health records (EHRs), wearable devices, and imaging technologies. For instance, wearable devices can continuously monitor blood pressure (BP) and other vital signs, providing a rich dataset for AI algorithms to analyze (Figure 3) (Lee et al, 2022; Visco et al., 2023). The data collected often includes photoplethysmograph (PPG) signals,
RkJQdWJsaXNoZXIy MjQ4ODY0NQ==