Deep Neural Networks for Human's Fall-risk Prediction using Force-Plate Time Series Signal.

Area of Science:

• Biomedical Engineering
• Gerontology
• Artificial Intelligence

Background:

• Falls are a significant risk for older adults, leading to injury and reduced quality of life.
• Early identification of fall risk is crucial for implementing preventive measures.
• Balance deficits are key indicators of fall risk, often assessed using force-plate data.

Purpose of the Study:

• To investigate the efficacy of deep neural networks in identifying human balance patterns for fall-risk prediction.
• To develop and evaluate a novel One-One-One Deep Neural Network algorithm for enhanced fall-risk assessment.
• To compare the performance of the proposed algorithm against existing deep learning models.

Main Methods:

• Utilized an open-source force-plate dataset from 163 participants (aged 18-86) under various standing conditions.
• Employed deep learning architectures including Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), and 1D Convolutional Neural Network (1D-CNN).
• Developed and implemented a novel One-One-One Deep Neural Network for fall-risk classification based on the Falls Efficacy Scale (FES).

Main Results:

• The proposed One-One-One Deep Neural Network significantly outperformed other algorithms and state-of-the-art models.
• Achieved exceptional performance metrics: 99.9% accuracy, 100% precision, and 100% sensitivity at the 12th epoch.
• Demonstrated the model's high efficiency in predicting fall-risk using force-plate time series signals.

Conclusions:

• The One-One-One Deep Neural Network presents a novel and highly efficient methodology for accurate human fall-risk prediction.
• This approach offers a promising tool for early detection and prevention of falls, particularly in vulnerable populations like older adults.
• The study highlights the potential of advanced deep learning techniques in clinical applications for health assessment.