Bioscience Methods 2024, Vol.15, No.3, 124-138 http://bioscipublisher.com/index.php/bm 131 (ML) and deep learning (DL) algorithms, such as convolutional neural networks (CNNs) and deep neural networks (DNNs), are employed to train the model (Ali et al., 2019; Sangha et al., 2021). For instance, an ensemble neural network can be trained using ECG data to predict pulmonary hypertension with high accuracy (Kwon et al., 2020; Kwon and Kim, 2020). Validation is the process of evaluating the model's performance on a separate dataset that was not used during training. This helps in assessing the model's generalizability and robustness. Techniques like cross-validation and the use of external validation datasets are commonly employed to ensure the model performs well on unseen data (Kwon et al., 2020; Kwon and Kim, 2020; Sangha et al., 2021). Additionally, sensitivity maps and Gradient-weighted Class Activation Mapping (Grad-CAM) are used to interpret the model's decision-making process, thereby enhancing its reliability (Sangha et al., 2021). 5.4 Integration with clinical workflows The final component involves integrating the AI-based diagnostic system into clinical workflows. This step is crucial for ensuring that the system is user-friendly and can be seamlessly adopted by healthcare professionals. Integration involves developing user interfaces, such as R Shiny apps, that allow clinicians to input patient data and receive diagnostic recommendations (Judge et al., 2023). The system should also be capable of providing real-time alerts and recommendations based on continuous monitoring data from wearable devices (Lee et al, 2022; Visco et al., 2023). Moreover, the AI system should be designed to complement existing clinical practices rather than replace them. For instance, AI can assist in triaging patients for further diagnostic tests like coronary angiography, thereby reducing procedural risks and improving patient outcomes (Alizadehsani et al., 2020). Rigorous evaluation and validation are essential to ensure the system's safety and effectiveness before it is integrated into routine clinical practice (Judge et al., 2023). In conclusion, the development of AI-based diagnostic systems for hypertensive heart disease involves a multi-faceted approach that includes data acquisition and preprocessing, feature extraction and selection, model training and validation, and integration with clinical workflows. Each of these components plays a vital role in ensuring the accuracy, reliability, and usability of the diagnostic system, ultimately leading to improved patient care and outcomes. 6 Case Study in place 6.1 Background and objectives of the case study Hypertensive heart disease (HHD) is a significant contributor to cardiovascular morbidity and mortality worldwide. The increasing prevalence of hypertension necessitates innovative diagnostic approaches to improve early detection and management. Artificial intelligence (AI) has emerged as a promising tool in this regard, offering the potential to enhance diagnostic accuracy and personalize treatment plans. This case study aims to explore the implementation of AI-based diagnostic systems for HHD in a clinical setting, focusing on their impact on patient care and outcomes. 6.2 Implementation of AI-based diagnostics in a clinical setting The implementation of AI-based diagnostic systems in clinical settings involves several steps, including data collection, algorithm development, and integration into clinical workflows. In this case study, we utilized a machine learning algorithm trained on electrocardiogram (ECG) data to detect hypertension and related cardiac conditions. The algorithm was developed using a large dataset of ECGs and clinical parameters from patients without cardiovascular disease (CVD) (Angelaki et al., 2022). The AI system was integrated into the hospital's electronic health record (EHR) system, allowing for real-time analysis of ECGs and immediate feedback to clinicians. The AI algorithm was designed to identify key features in the ECG that are indicative of hypertension, such as the S-wave, P-wave, and T-wave characteristics (Kwon et al., 2020; Kwon and Kim, 2020). Additionally, the system
RkJQdWJsaXNoZXIy MjQ4ODY0NQ==