Fall Detection for the Elderly Based on 3-Axis Accelerometer and Depth Sensor Fusion with Random Forest Classifier

The Applications of Artificial Intelligence for Assessing Fall Risk: Systematic Review

BACKGROUND

Falls and their consequences are a serious public health problem worldwide. Each year, 37.3 million falls requiring medical attention occur. Therefore, the analysis of fall risk is of great importance for prevention. Artificial intelligence (AI) represents an innovative tool for creating predictive statistical models of fall risk through data analysis.

OBJECTIVE

The aim of this review was to analyze the available evidence on the applications of AI in the analysis of data related to postural control and fall risk.

METHODS

A literature search was conducted in 6 databases with the following inclusion criteria: the articles had to be published within the last 5 years (from 2018 to 2024), they had to apply some method of AI, AI analyses had to be applied to data from samples consisting of humans, and the analyzed sample had to consist of individuals with independent walking with or without the assistance of external orthopedic devices.

RESULTS

We obtained a total of 3858 articles, of which 22 were finally selected. Data extraction for subsequent analysis varied in the different studies: 82% (18/22) of them extracted data through tests or functional assessments, and the remaining 18% (4/22) of them extracted through existing medical records. Different AI techniques were used throughout the articles. All the research included in the review obtained accuracy values of >70% in the predictive models obtained through AI.

CONCLUSIONS

The use of AI proves to be a valuable tool for creating predictive models of fall risk. The use of this tool could have a significant socioeconomic impact as it enables the development of low-cost predictive models with a high level of accuracy.

TRIAL REGISTRATION

PROSPERO CRD42023443277; https://tinyurl.com/4sb72ssv.

Towards Dynamic Multi-Modal Intent Sensing Using Probabilistic Sensor Networks

Intent sensing—the ability to sense what a user wants to happen—has many potential technological applications. Assistive medical devices, such as prosthetic limbs, could benefit from intent-based control systems, allowing for faster and more intuitive control. The accuracy of intent sensing could be improved by using multiple sensors sensing multiple environments. As users will typically pass through different sensing environments throughout the day, the system should be dynamic, with sensors dropping in and out as required. An intent-sensing algorithm that allows for this cannot rely on training from only a particular combination of sensors. It should allow any (dynamic) combination of sensors to be used. Therefore, the objective of this study is to develop and test a dynamic intent-sensing system under changing conditions. A method has been proposed that treats each sensor individually and combines them using Bayesian sensor fusion. This approach was tested on laboratory data obtained from subjects wearing Inertial Measurement Units and surface electromyography electrodes. The proposed algorithm was then used to classify functional reach activities and compare the performance to an established classifier (k-nearest-neighbours) in cases of simulated sensor dropouts. Results showed that the Bayesian sensor fusion algorithm was less affected as more sensors dropped out, supporting this intent-sensing approach as viable in dynamic real-world scenarios.

Triaxial Accelerometer-Based Falls and Activities of Daily Life Detection Using Machine Learning

The detection of activities of daily living (ADL) and the detection of falls is of utmost importance for addressing the issue of serious injuries and death as a consequence of elderly people falling. Wearable sensors can provide a viable solution for monitoring people in danger of falls with minimal external involvement from health or care home workers. In this work, we recorded accelerometer data from 35 healthy individuals performing various ADLs, as well as falls. Spatial and frequency domain features were extracted and used for the training of machine learning models with the aim of distinguishing between fall and no fall events, as well as between falls and other ADLs. Supervised classification experiments demonstrated the efficiency of the proposed approach, achieving an F1-score of 98.41% for distinguishing between fall and no fall events, and an F1-score of 88.11% for distinguishing between various ADLs, including falls. Furthermore, the created dataset, named "ShimFall&ADL" will be publicly released to facilitate further research on the field.