Remote Patient Monitoring Using Radio Frequency Identification (RFID) Technology and Machine Learning for Early Detection of Suicidal Behaviour in Mental Health Facilities

Xiaohui Tao, Thanveer Basha Shaik, Niall Higgins, Raj Gururajan, Xujuan Zhou

TL;DR

This paper explores using RFID technology and machine learning for non-invasive remote patient monitoring in mental health facilities to detect early signs of suicidal behavior. They found that the optimal placement of RFID readers is crucial for signal strength and that Decision Tree models performed best for predicting this signal strength.

What Problem Does This Paper Address?

Mental health facilities face a significant challenge in continuously monitoring patients for suicidal ideation and self-harm, especially during periods of reduced staff supervision like evenings and nights. Traditional monitoring methods can be intrusive or impractical for patients who are not bedridden. This research aims to develop a non-invasive Remote Patient Monitoring (RPM) system using Radio Frequency Identification (RFID) technology and machine learning to identify early indicators of suicidal behavior, thereby enhancing patient safety and care efficiency.

Methods

The study proposes an RPM system that leverages RFID technology to monitor patient vital signs and subtle movements. The core of their methodology involves two stages:

  1. Determining Optimal Reader-Antenna Placement: They simulated a hospital ward and experimented with various fixed positions for RFID reader-antennas. The goal was to find configurations that maximize the Received Signal Strength Indicator (RSSI) from passive RFID tags, which is crucial for accurate data collection.
  2. Machine Learning for Data Analysis: Once optimal positions were identified, they used machine learning models to analyze the collected RFID data. Specifically, they investigated:
  3. Ordinary Least Squares (OLS) Linear Regression: To establish a baseline understanding of the relationship between reader-antenna positions and RSSI.
  4. Machine Learning Models: Decision Tree, Random Forest, and XGBoost were trained to predict signal strength based on reader-antenna configurations.
  5. Ensemble Learning: A weighted average ensemble model was developed to combine the strengths of the individual machine learning models for potentially improved performance.

Datasets

The study utilized a custom dataset collected from a simulated hospital ward. This dataset consisted of:

  • Reader-Antenna Positions: 16 arbitrarily chosen positions for two UHF RFID reader-antennas.
  • RFID Tag Data: Received Signal Strength Indicator (RSSI) values, distances between the tag and antennas, and frequency.

This dataset was generated specifically for this research to evaluate the impact of reader-antenna placement on signal strength and to train machine learning models.

Key Insights and Ideas

  • Reader-Antenna Placement is Paramount: The study strongly emphasizes that the physical placement of RFID readers is a critical factor in achieving reliable signal strength and, consequently, accurate data for monitoring. Suboptimal placement can significantly degrade performance.
  • Decision Tree Dominance: Among the tested machine learning models, the Decision Tree algorithm demonstrated superior performance in predicting RSSI, suggesting its suitability for this specific task of analyzing RFID signal strength based on spatial configurations.
  • Ensemble Learning's Limited Impact (in this context): While ensemble learning is often beneficial, this study found that their weighted average ensemble model did not outperform the best individual model (Decision Tree). This might be due to the relatively simple nature of the problem and the limited number of variables.
  • Non-Invasive Monitoring Potential: The research highlights the feasibility of using RFID for non-invasive monitoring of vital signs and movement, which is particularly advantageous in psychiatric settings where patient comfort and freedom of movement are important.

Key Results

  • Optimal Placement Identified: The study identified specific reader-antenna positions that yielded the highest RSSI values, indicating better tag readability. For instance, position 'j' was found to have the highest RSSI.
  • Decision Tree Performance: The Decision Tree model achieved the best performance metrics, with a Mean Absolute Error (MAE) of 0.01 and Mean Squared Error (MSE) of 0.003.
  • Other Models: Random Forest and XGBoost also showed promising results, though with slightly higher error rates compared to Decision Tree.
  • OLS Regression Insights: Linear regression revealed that distances from the antennas to the tag (Distance_1 and Distance_2) were significant predictors of RSSI, with shorter distances generally correlating with higher RSSI.

Limitations and Future Work

  • Static Tag Limitation: The current study used a static RFID tag, which does not fully replicate the dynamic movement of patients.
  • Limited Range: The circular polarization technique, while offering orientation independence, resulted in a shorter read range (around 1.5m), which could be a limitation in larger ward settings.
  • Ensemble Learning Underperformance: The weighted average ensemble model did not improve performance, possibly due to the simplicity of the dataset and the limited number of variables.
  • Future Work: The authors suggest future research directions including:
  • Using dynamic RFID tags to capture patient motion.
  • Integrating Near-field Coherent Sensing (NCS) technology to monitor vital signs and subtle movements more comprehensively.
  • Developing generic machine learning models for broader application in remote patient monitoring systems.
  • Replicating the study in a real clinical environment.

Summary updated

Abstract

Remote Patient Monitoring (RPM) has gained great popularity with an aim to measure vital signs and gain patient related information in clinics. RPM can be achieved with noninvasive digital technology without hindering a patient’s daily activities and can enhance the efficiency of healthcare delivery in acute clinical settings. In this study, an RPM system was built using radio frequency identification (RFID) technology for early detection of suicidal behaviour in a hospital-based mental health facility. A range of machine learning models such as Linear Regression, Decision Tree, Random Forest, and XGBoost were investigated to help determine the optimum fixed positions of RFID reader–antennas in a simulated hospital ward. Empirical experiments showed that Decision Tree had the best performance compared to Random Forest and XGBoost models. An Ensemble Learning model was also developed, took advantage of these machine learning models based on their individual performance. The research set a path to analyse dynamic moving RFID tags and builds an RPM system to help retrieve patient vital signs such as heart rate, pulse rate, respiration rate and subtle motions to make this research state-of-the-art in terms of managing acute suicidal and self-harm behaviour in a mental health ward.
BibTeX
@inproceedings{tao2021remote,
  author = {Xiaohui Tao and Thanveer Basha Shaik and Niall Higgins and Raj Gururajan and Xujuan Zhou},
  title = {{Remote Patient Monitoring Using Radio Frequency Identification (RFID) Technology and Machine Learning for Early Detection of Suicidal Behaviour in Mental Health Facilities}},
  booktitle = {Sensors},
  year = {2021},
  url = {https://openreview.net/forum?id=McxP5IKjOh},
}
Figure