Abstract
Palliative care prioritizes patients’ quality of life with symptom management, personal care, and minimal technology use to respect privacy. However, assessing symptoms, particularly for patients near the end of life who often are unable to communicate, mostly depends on observation by healthcare providers, which may lack completeness and objectivity. Non-invasive radar-based monitoring of heart rate (HR) and respiration rate (RR) offers potential for detecting symptoms without compromising palliative care principles. We used a multi-radar system placed under patients’ mattresses to continuously monitor HRs and RRs in a palliative care ward by tracking body surface movement. Data on symptom burden, including timing and severity, were collected via patient self-reports (MIDOS 2) and staff observations (HOPE). The relationship between radar-derived vital signs and distress-related symptom burden changes over hours, as well as medication effects following their peak impact, was analyzed using various statistical methods. Our findings show that HR and RR increase over the course of hours during symptom exacerbation (HR: +2.8 bpm, RR: +0.7 brpm) and decrease with amelioration (HR: –1.8 bpm, RR: –0.2 brpm), with a significant relationship observed between HR and symptom exacerbation (p=0.002). Additionally, HR trends suggested variability potentially related to medication effects, warranting further study. Radar-based monitoring holds promise for burden-free vital sign monitoring in palliative care and assisting healthcare professionals in symptom detection.
Introduction
In palliative care, the primary goal is maintaining quality of life, which requires precise symptom management—such as controlling pain or shortness of breath. However, many patients near the end of life are unable to communicate their distress effectively. Currently, clinicians rely on periodic observations or self-reports, both of which are prone to subjectivity, bias, and "social expectancy" (where patients under-report symptoms to avoid bothering staff). While intensive care units use continuous monitoring (ECG, pulse oximetry), these technologies are often seen as too intrusive for the sensitive, home-like environment of a palliative ward. This study investigates a "burden-free" monitoring solution that provides objective data without wires, skin contact, or privacy-invading cameras.
Methods
The researchers deployed a multi-radar remote sensing system consisting of four continuous-wave (CW) radar modules operating in the 24GHz band. These were placed discreetly under the patients' mattresses at chest level to track body surface movement.
Technical Innovations
To extract reliable vital signs from the radar signals, several advanced techniques were employed:
- Vital Sign Extraction: An ellipse reconstruction technique was used to mitigate signal offsets.
- Heart Sound Detection: Heart rate (HR) was estimated by detecting individual heart sounds (S1 and S2) using a Hidden Semi-Markov Model (HSMM), allowing for accurate instantaneous HR detection.
- Outlier Detection: A novel two-stage iterative outlier detection process combined global median filtering with a localized sliding-window Z-score method to preserve natural HR variability while stripping out motion artifacts.
The study involved 77 palliative care patients. Ground truth symptom data were collected via the MIDOS² (patient self-reports) and HOPE (staff observations) questionnaires, alongside retrospective analysis of Clinical Health Records (CHR).
Results and Discussion
The core hypothesis is that symptom distress triggers sympathetic nervous system activation, which the radar can detect as a "vegetative proxy."
Symptom Exacerbation and Amelioration
During periods of worsening symptoms, the radar detected a mean increase of +2.8 bpm in HR and +0.7 brpm in respiration rate (RR). The relationship between HR and exacerbation was statistically significant (p=0.002). Conversely, when symptoms improved, HR decreased by -1.8 bpm and RR by -0.2 brpm. The system was notably better at detecting symptom exacerbation than amelioration, suggesting that the physiological "alarm" of distress is a stronger signal than the relaxation following treatment.
Medication Effectiveness
In short-term analysis, the system tracked the physiological response to medication. When medication was effective, a clear downward trend in HR was observed as the patient reached the expected peak effect of the drug.
Conversely, ineffective medication (where the patient still reported high distress) was associated with a continued sympathetic response: HR increased by +2.0 bpm and RR by +1.0 brpm.
Conclusion
Radar-based monitoring holds significant promise for burden-free vital sign monitoring in palliative care, assisting healthcare professionals in objective symptom detection. While physical activity and environmental factors can affect measurements, the technology successfully preserved valid data even during patient distress. Future research will focus on Heart Rate Variability (HRV) as a more sensitive biomarker and refining the technology to better distinguish between physical activity and symptomatic distress.
Summary updated