Citations

Photoplethysmography (PPG) is a non-invasive optical technique commonly used to measure heart rate and oxygen saturation, but its waveform contains additional physiological information that can be analyzed using artificial intelligence (AI). This narrative review summarizes the emerging applications of AI-based PPG in cardiovascular, respiratory, sleep, hemodynamic, pregnancy-related, and portal-hypertension assessment, with the aim of evaluating its potential beyond conventional monitoring and identifying barriers to clinical translation. The literature search was conducted using PubMed, Google Scholar, IEEE Xplore, ScienceDirect, and SpringerLink. Additional relevant studies were identified through screening the reference lists of included articles. Studies published between 2002 and 2026 were identified to capture the development of PPG from conventional monitoring to newer AI-based applications. Human studies were prioritized, while relevant computational, simulated, synthetic, ex vivo, and technical studies were also included. Studies unrelated to PPG, duplicates, and studies with limited relevance were excluded. A total of 96 references were included, covering AI approaches such as convolutional and deep neural networks, ensemble methods, transfer learning, U-Net, generative adversarial networks, and Transformer-based models. Overall, the reviewed evidence suggests that AI-based PPG may support blood pressure estimation, atrial fibrillation detection, sleep and respiratory monitoring, vascular aging assessment, pulmonary hypertension screening, preeclampsia assessment, volume-status and compensatory-reserve assessment, and exploratory assessment related to portal hypertension. However, clinical translation remains limited by motion artifacts, sensor and measurement-site variability, skin-pigmentation-related bias, physiological and environmental influences, heterogeneous methods, limited external validation, and inconsistent clinical and regulatory standards.

Artificial intelligence (AI) is rapidly transforming healthcare through applications in clinical decision support, diagnostic imaging, population health management, and workflow optimization. Despite these advances, real-world deployment continues to expose critical challenges related to safety, bias, transparency, and integration into clinical workflows. Algorithmic bias can exacerbate health disparities, limited explainability may undermine clinician trust, and insufficient validation and post-deployment monitoring can compromise patient safety. Although the World Health Organization (WHO) has established six ethical principles for AI in health, including autonomy, well-being and safety, transparency, accountability, equity, and sustainability, translating these high-level principles into practical and enforceable governance mechanisms remains a persistent challenge. This narrative review synthesizes insights from bioethics, health policy, computer science, and clinical medicine to identify gaps in current AI governance approaches and proposes a lifecycle-aligned governance-by-design framework that operationalizes WHO ethical principles across key stages of the healthcare AI lifecycle, including data collection, model development, validation, deployment, and post-deployment monitoring. The framework integrates concrete governance mechanisms such as consent governance, fairness evaluation, external validation, explainability, clinician oversight, and continuous performance monitoring. Overall, this work advances a practical, lifecycle-integrated approach to AI governance and provides a structured foundation for developing safe, equitable, and trustworthy AI systems in healthcare.

Abstract Pain assessment plays a critical role in clinical decision-making, yet conventional methods rely heavily on subjective self-reports such as the Visual Analog Scale (VAS) or the Numeric Rating Scale (NRS). These approaches fail in scenarios involving infants, sedated patients, unconscious individuals, or those with cognitive or communication impairments. To address this gap, this research proposes PainNet, an AI-driven, real-time facial pain recognition system that integrates deep learning-based object detection with Action Unit (AU)–based pain intensity scoring. The system uses YOLOv8 for rapid facial expression detection and incorporates Py-Feat to extract pain-relevant AUs, enabling computation of the Prkachin and Solomon Pain Intensity (PSPI) score. A balanced dataset comprising 500 images sourced from Delaware, BioVid, and RaFD was curated, annotated, and preprocessed for training. The YOLOv8-based binary pain vs. no-pain classifier achieved a mean Average Precision (mAP) of 0.89 and an F1-score of 0.80 on the test dataset, demonstrating strong performance in diverse environments and facial variations. The design emphasizes computational efficiency by activating AU analysis only when YOLOv8 detects probable pain, making PainNet suitable for continuous monitoring in ICUs, elder-care facilities, and telemedicine systems. This study demonstrates that the fusion of real-time object detection and clinically validated AU scoring can significantly advance automated, objective, and scalable pain assessment.

Abstract Interstitial lung disease (ILD) is a chronic, irreversible condition that is challenging to diagnose due to subtle symptoms and limitations of conventional tools. Noninvasive lung sound analysis using phonopulmography offers a promising alternative. This study evaluated the feasibility of a deep learning model for distinguishing normal and pathological lung sounds in ILD. Lung sound recordings were collected from 12 subjects (6 healthy, 6 ILD) and segmented beat-to-beat using simultaneously acquired ECG and phonopulmography, yielding 2,280 segments. Mel-frequency cepstral coefficients (MFCCs; 13 per segment) were extracted, and a one-dimensional CNN-based You Only Listen Once (YOLO) model was trained for segment classification. Evaluation used Leave-One-Out Cross-Validation, with patient-level classification by aggregating segment predictions via majority vote. On the test set (190 segments per class), the model achieved an average segment-level accuracy of 0.738 ± 0.087 and mean AUC of 0.850 ± 0.062. Precision, recall, and F1-score were 0.775 ± 0.148, 0.755 ± 0.183, and 0.738 ± 0.089, respectively. These preliminary results suggest that beat-level MFCCs with a 1D CNN-based YOLO model enable accurate segment- and patient-level classification. This feasibility study highlights the potential of AI-assisted auscultation for noninvasive monitoring and clinical decision support in ILD management, pending further validation in larger cohorts.

Abstract Accurate assessment of left-ventricular ejection fraction (EF) is a critical metric in diagnosing and managing heart failure; however, methods of assessment remain resource intensive, not widely available in low-acuity or remote settings and are intermittent in nature. This study explores the potential of a novel framework, cardiac electromechanical fingerprinting (cEMF) using synchronized electrocardiography (ECG) and seismocardiography (SCG) collected from a wearable sensing patch as a low-cost, non-invasive, continuous estimate of EF that can be used for rapid point of care screening. A machine learning pipeline was developed to process raw ECG and SCG data to derive EF. Signals were cleaned using band-pass filtering, NeuroKit2-based ECG cleaning and fiducial detection of critical electromechanical events. We extracted an ECG-SCG feature space consisting of heart-rate variability metrics, ECG morphological characterization, SCG timing intervals, spectral and entropy-based features. The top thirty most important features for EF were used to train and evaluate several regressor models. The best performing model was the Gaussian Process Regressor with an MAE of 5.43. Our findings demonstrate the promise of producing clinically relevant cardiac function measures, such as EF, from wearable electromechanical signals. Future work using larger datasets and complementary physiological signals such as phonocardiogram (PCG) may significantly strengthen EF% estimation and other cardiac functions for clinical applicability.

Background/Objectives: Precocious puberty (PP), defined as the onset of secondary sexual characteristics before 8 years in girls and 9 years in boys, is associated with psychosocial distress, compromised adult height, and long-term metabolic risk. Early identification remains challenging, as current diagnostic approaches are largely reactive and rely on invasive or resource-intensive testing. This narrative review examines how artificial intelligence (AI) can support earlier risk prediction and detection of PP through integration of clinical, hormonal, imaging, lifestyle, and environmental data. Methods: A narrative literature review was conducted using PubMed, Scopus, Embase, Web of Science, and Google Scholar to identify relevant studies published between 2005 and 2025. Eligible studies included original research and high-quality reviews that examined AI-based approaches, such as machine learning and deep learning, in pediatric endocrinology, particularly for the prediction or diagnosis of central or peripheral precocious puberty. Studies incorporating clinical, hormonal, radiological, lifestyle, environmental, or multi-omics data relevant to AI modeling were included. Results: AI models, including XGBoost, random forest, convolutional neural networks, and regression-based approaches, have demonstrated potential utility in predicting central precocious puberty using hormonal, imaging, and growth data. Reported applications include automated bone age assessment, lifestyle and dietary risk stratification, and exploratory use of wearable-derived behavioral data. However, progress is limited by small pediatric datasets, population bias, limited interpretability, and unresolved ethical challenges related to privacy, consent, and equity. Conclusions: Artificial intelligence represents a promising decision-support approach for earlier, non-invasive, and individualized risk assessment in precocious puberty. Future progress will depend on the integration of longitudinal, multimodal data, the development of ethical models, and interdisciplinary collaboration among pediatric endocrinologists, data scientists, and public health stakeholders.

Hypertension has been traditionally known to be highlighted by mean blood pressure; however, emerging evidence exhibits that blood pressure variability (BPV), including short-term, day-to-day, and visit-to-visit fluctuations can have an implication across multiple body systems. Elevated BPV reflects repetitive hemodynamic stress, affecting the physiologic hemostasis contributing to vascular injury and end organ damage. This narrative review is a compilation of recent evidence on the prognostic value of BPV, explained by pathophysiology, various devices with its measurement approaches, and, essentially, the clinical implication of BPV and the use of such devices utilizing artificial intelligence. A comprehensive literature search across PubMed, Cochrane Library, Scopus, and Web of Science were conducted, focusing on observational studies, cohorts, randomized trials, and meta-analyses. Higher BPV has been associated with an increased risk of cardiovascular mortality, stroke, coronary events, and heart failure, the progression of chronic kidney disease, cognitive decline, and preeclampsia, among other end organ damage, despite mean blood pressure. The various pathophysiologic mechanisms include autonomic dysregulation, arterial stiffness, endothelial dysfunction, circadian rhythm alteration, and systemic inflammation, which result in vascular remodeling and multisystem damage. Antihypertensive medications such as calcium channel blockers and renin–angiotensin–aldosterone system inhibitors seem to reduce BPV; randomized trials have not specifically investigated their BPV-reducing effects. The aim of this review is to highlight that BPV is a dynamic marker of multisystem risk, and question how various AI-based devices can aid continuous BPV monitoring and patient specific risk stratification.

Background: Colorectal cancer (CRC) is the second leading cause of cancer-related mortality worldwide, accounting for approximately 10% of all cancer cases. Despite the proven effectiveness of conventional screening modalities such as colonoscopy and fecal immunochemical testing (FIT), their invasive nature, high cost, and limited patient compliance hinder widespread adoption. Recent advancements in artificial intelligence (AI) and bowel sound-based signal processing have enabled non-invasive approaches for gastrointestinal diagnostics. Among these, bowel sound analysis—historically considered subjective—has reemerged as a promising biomarker using digital auscultation and machine learning. Objective: This review explores the potential of AI-powered bowel sound analytics for early detection, screening, and characterization of colorectal cancer. It aims to assess current methodologies, summarize reported performance metrics, and highlight translational opportunities and challenges in clinical implementation. Methods: A narrative review was conducted across PubMed, Scopus, Embase, and Cochrane databases using the terms colorectal cancer, bowel sounds, phonoenterography, artificial intelligence, and non-invasive diagnosis. Eligible studies involving human bowel sound-based recordings, AI-based sound analysis, or machine learning applications in gastrointestinal pathology were reviewed for study design, signal acquisition methods, AI model architecture, and diagnostic accuracy. Results: Across studies using convolutional neural networks (CNNs), gradient boosting, and transformer-based models, reported diagnostic accuracies ranged from 88% to 96%. Area under the curve (AUC) values were ≥0.83, with F1 scores between 0.71 and 0.85 for bowel sound classification. In CRC-specific frameworks such as BowelRCNN, AI models successfully differentiate abnormal bowel sound intervals and spectral patterns associated with tumor-related motility disturbances and partial obstruction. Distinct bowel sound-based signatures—such as prolonged sound-to-sound intervals and high-pitched “tinkling” proximal to lesions—demonstrate the physiological basis for CRC detection through bowel sound-based biomarkers. Conclusions: AI-driven bowel sound analysis represents an emerging, exploratory research direction rather than a validated colorectal cancer screening modality. While early studies demonstrate physiological plausibility and technical feasibility, no large-scale, CRC-specific validation studies currently establish sensitivity, specificity, PPV, or NPV for cancer detection. Accordingly, bowel sound analytics should be viewed as hypothesis-generating and potentially complementary to established screening tools, rather than a near-term alternative to validated modalities such as FIT, multitarget stool DNA testing, or colonoscopy.

This study proposes a novel two-dimensional Empirical Mode Decomposition (2D EMD)-based deep learning framework to enhance model performance in multi-class image classification tasks and potential early detection of diseases in healthcare using medical imaging. To validate this approach, we apply it to gastrointestinal (GI) endoscopic image classification using the publicly available Kvasir dataset, which contains eight GI image classes with 1000 images each. The proposed 2D EMD-based design procedure decomposes images into a full set of intrinsic mode functions (IMFs) to enhance image features beneficial for AI model development. Integrating 2D EMD into a deep learning pipeline, we evaluate its impact on four popular models (ResNet152, VGG19bn, MobileNetV3L, and SwinTransformerV2S). The results demonstrate that subtracting IMFs from the original image consistently improves accuracy, F1-score, and AUC for all models. The study reveals a notable enhancement in model performance, with an approximately 9% increase in accuracy compared to counterparts without EMD integration for ResNet152. Similarly, there is an increase of around 18% for VGG19L, 3% for MobileNetV3L, and 8% for SwinTransformerV2. Additionally, explainable AI (XAI) techniques, such as Grad-CAM, illustrate that the model focuses on GI regions for predictions. This study highlights the efficacy of 2D EMD in enhancing deep learning model performance for GI image classification, with potential applications in other domains.

Phonoenterography (PEG) offers a non-invasive and radiation-free technique to assess gastrointestinal activity through acoustic signal analysis. In this feasibility study, 110 high-resolution PEG recordings (44.1 kHz, 16-bit) were acquired from eight healthy individuals, yielding 6314 prominent bowel sound (PBS) segments through automated segmentation. Each event was characterized using a 279-feature acoustic profile comprising Mel-frequency cepstral coefficients (MFCCs), their first-order derivatives (Δ-MFCCs), and six global spectral parameters. After normalization and dimensionality reduction with PCA and UMAP (cosine distance, 35 neighbors, minimum distance = 0.01), five clustering strategies were evaluated. K-Means (k = 5) achieved the most favorable balance between cluster quality (silhouette = 0.60; Calinski–Harabasz = 19,165; Davies–Bouldin = 0.68) and interpretability, consistently identifying five acoustic patterns: single-burst, multiple-burst, harmonic, random-continuous, and multi-modal. Temporal modeling of clustered events further revealed distinct sequential dynamics, with Single-Burst events showing the longest dwell times, random continuous the shortest, and strong diagonal elements in the transition matrix confirming measurable state persistence. Frequent transitions between random continuous and multi-modal states suggested dynamic exchanges between transient and overlapping motility patterns. Together, these findings demonstrate that unsupervised PEG-based analysis can capture both acoustic variability and temporal organization of bowel sounds. This annotation-free approach provides a scalable framework for real-time gastrointestinal monitoring and holds potential for clinical translation in conditions such as postoperative ileus, bowel obstruction, irritable bowel syndrome, and inflammatory bowel disease.