The possibility of a healthy individual in their twenties experiencing a fall is so minimal that it’s rarely a concern. However, consider your older family members, particularly those over 65. But how exactly are relatives and close friends able to care for those at risk of falling? Constant care is required, which places strain on the caretaker. If a technology was created that could help alleviate the burden placed on the caretaker, the technology would greatly benefit the caretaker. Researchers at Universidad de Las Américas in Ecuador have created an artificial intelligence (AI) model that can detect human falls with 80% accuracy (Villegas-Ch et al.). This article will go over the need for the model, how the model was trained, and how the model works.

‍

“In the United States, it’s reported that each year, over 14 million adults aged 65 and above, which equates to one in four older adults, experience a fall” (CDC). This large number of falls is alarming and underscores the need for effective fall detection systems, a field where AI can be applied. “Falls are the leading cause of injury-related death among adults ages 65 and older, and the fall death rate is increasing” (CDC). Falls can result in more than just minor bruises or bumps. For older individuals, whose physical strength is often diminished, a simple fall can have fatal consequences. “Falls among adults ages 65 and older are very costly. Each year about $50 billion is spent on medical costs related to older adult falls—fatal injuries total $754 million, and the remainder is attributed to non-fatal fall injuries” (CDC). This substantial expenditure on treating the consequences of falls could be more effectively allocated to systems that aim to prevent falls in the first place, which would be a much better option.

‍

Researchers at Universidad de Las Américas set out to create an AI model to help solve this problem (Villegas-Ch et al.). To create a model, they used knowledge from various articles that researched and produced methods of monitoring an environment in real-time to overlay the skeletal structure onto a human (Villegas-Ch et al.). This would allow them to create a model that would use the position of the skeletal structure (angles) to predict the fall risk (Villegas-Ch et al.). They picked this method after reading research from a paper published by Nho et al. which went over the three methods of fall detection approaches: wearable-based, ambient-based, and vision-based (Nho et al.). This paper did not suggest that a vision-based approach was the most appropriate, but rather that it was an approach that could be used in certain scenarios (Nho et al.).

‍

To build the model, the researchers used C++ as their coding language (Villegas-Ch et al.) The model that was built relied on creating a map of body key points that would allow the skeletal structure to be rendered (Villegas-Ch et al.) It took real-time images using a Kinect camera and transformed them into guesses for fall detection using an AI model (Villegas-Ch et al.)  The results of the model were very good as false negatives were 2%, false positives were 18%, and falls detected were 80% of the time in a home environment (Villegas-Ch et al.)

‍

In conclusion, the research conducted by the team at Universidad de Las Américas in Ecuador has made significant strides in the field of AI-assisted fall detection (Villegas-Ch et al.) Their model, which overlays a skeletal structure onto a human figure in real-time, can predict fall risk with an accuracy of 80% (Villegas-Ch et al.). This technology has the potential to ease the burden on caregivers and could play an important role in preventing falls among older adults (Villegas-Ch et al.). The model’s low rates of false negatives (2%) and false positives (18%) further show its effectiveness. While there is still room for improvement, this research represents a promising step forward in using AI to enhance the quality of life for older adults and their caregivers. This study serves as a testament to the potential of AI in the healthcare industry.

‍

‍

Works Cited

Nho, Young-Hoon, et al. “Cluster-Analysis-Based User-Adaptive Fall Detection Using Fusion of Heart Rate Sensor and Accelerometer in a Wearable Device.” IEEE Access, vol. 8, Institute of Electrical and Electronics Engineers, Jan. 2020, pp. 40389–401, doi:10.1109/access.2020.2969453.

Older Adult Falls  | Fall Prevention | Injury Center | CDC. www.cdc.gov/falls/index.html.

Villegas-Ch, William, et al. “Model for the Detection of Falls With the Use of Artificial Intelligence as an Assistant for the Care of the Elderly.” Computation (Basel), vol. 10, no. 11, Multidisciplinary Digital Publishing Institute, Nov. 2022, p. 195, doi:10.3390/computation10110195.

‍