Optical fiber sensors (OFSs) have emerged as essential tools in the monitoring of physical, chemical, and bio-medical parameters in harsh situations due to their high sensitivity, electromagnetic interference (EMI) immunity, and long-term stability. However, the current literature contains scattered information in most reviews regarding individual sensing technologies or domains. This study provides a structured exploratory review in a novel inter-family analysis of both intrinsic and extrinsic configurations by analyzing more than 23,000 publications between 2019 and 2025 in five key domains: industry, medicine and biomedicine, environmental chemistry, civil/structural engineering, and aerospace. The analysis aims to critically discuss how functional principles/parameters and methods of interrogation affect the applicability of different OFS categories. The results reveal leading trends in the use of techniques like the use of fiber Bragg gratings (FBG) and distributed sensing in high-accuracy conditions or the rising role of extrinsic sensors in selective chemical situations and point out new approaches in areas like Artificial Intelligence (AI)- or Internet of Things (IoT)-integrated sensors. Further, this synthesis not only connects pieces of knowledge but also defines the technological barriers in terms of calibration cost and standardization: this provides strategic insight regarding future research and the scalability of industry deployment.

Cardiovascular diseases claim millions of lives each year, yet timely diagnosis remains a significant challenge due to the high number of patients and associated costs. Although various machine learning solutions have been proposed for this problem, most approaches rely on careful data preprocessing and feature engineering workflows that could benefit from more comprehensive documentation in research publications. To address this issue, this paper presents a machine learning framework for predicting heart attack risk online. Our systematic methodology integrates a unified pipeline featuring advanced data preprocessing, optimized feature selection, and an exhaustive hyperparameter search using cross-validated grid evaluation. We employ a metamodel ensemble strategy, testing and combining six traditional supervised models along with six stacking and voting ensemble models. The proposed system achieves accuracies ranging from 90.2% to 98.9% on three independent clinical datasets, outperforming current state-of-the-art methods. Additionally, it powers a deployable, lightweight web application for real-time decision support. By merging cutting-edge AI with clinical usability, this work offers a scalable solution for early intervention in cardiovascular care.

Structural health monitoring based on vibration signal analysis has demonstrated itself as a powerful technique for preventing structural damage, since only the dynamic response of structural elements is required. Moreover, in beam-like structures, damage alters the system’s vibration characteristics, providing valuable information for damage assessment. In this work, a light-intensity-modulated fibre-optic reflective vibration sensing system is combined with a multitask learning (MTL) framework for simultaneous classification and localization of damage in cantilever beams. The proposed framework adopts a unified neural network architecture that learns multiple tasks simultaneously and uses the spectral properties of the optical fibre vibration response as input features. Specifically, it performs 12-class structural-condition classification and mechanism-specific regression of crack- and wear-type damage locations. Under the evaluated acquisition-level protocol, the multitask framework achieved 96.4% classification accuracy and mean absolute localization errors of 1.16 mm for crack-type damage and 1.52 mm for wear-type damage. Performance was compared with multiple independently trained single-task baselines, including multilayer perceptron (MLP), support vector machine, k-nearest neighbours, and decision tree models, together with additional linear and probabilistic approaches. Across ten matched random seeds, comparison with the strongest overall single-task baseline, the MLP, showed a statistically significant but small classification advantage and larger reductions in localization error. The observed performance reflects generalization to new vibration acquisitions under the experimental conditions represented in the dataset, rather than to entirely unseen specimens, supporting the framework’s potential for simultaneous damage classification and localization in beam-like structures.