Transformation in Healthcare by Wearable Devices for Diagnostic and Guidance of Treatment

Authors: Aman Mahajan , William Kaiser, Gregory Pottie
Publication: ACM Transactions on Computing for Healthcare , March 2020
CCS concepts: Applied computing -> Life and medical sciences

 

01 Introduction

Wearable devices offering the potential for

1. Non-invasive, Constantly vigilant, Low-cost monitoring or individual condition
2. Fundamental advances in guiding healthcare

But it is accompanied by responsibilities like the assurance of accurate, precise, reliable, and resilient systems.

1. Inherent lack of accurate physical models.
2. Algorithm training are required to enable diagnostics, and it requires ground truth in the form of accurately curated data sources from carefully constructed subject trials with a properly distributed subject population of significant size.

 

02 Background and Motivation

When are wearable devices more effective than clinical instruments?

Wearable monitoring

• Perioperative care
• Detection of condition that appears only as ephemeral events
• High frequency of measurement interval(detailed detection)
• Personalize care
• Advance in clinical condition(residential or workplace environments) – remote healthcare and reduce the need to be transferred to the hospital

What are the physical limitations and uncertainties of wearable devices?

• Without expert guidance, it relies on automation.
• No expert supervision of users -> uncertainty associated with usage compliant.
• Variability in the coupling of transducers to subject tissue > uncertainty.
• Interference and noise > uncertainty.

 

03 Research in Computing Healthcare

Rely on Signal processing, Statistical Analytics, and Machine Learning

The application of machine learning requires sufficient training data
(very large subject number)

For a new device -> number of the training dataset is limited

Common occurring monitoring methods are based on the analysis of signal waveforms. -> classification

 

04 Research Path

1. Find research benefits.
2. Find Scientific foundations for the effectiveness of wearable monitoring.
3. Find methods for validation of system performance.
    (require the development of a comprehensive data acquisition system)
4. Design of clinical trials for both development and validation. (get data)
5. Design clinical workflow.
6. Design wearable transducers.
7. Development of analytical methods for condition indication.
8. Evaluation of system performance.
9. Ensuring compliance for continuous monitoring.
10. Vigilant detection and notification of failure.
11. Privacy and security check

 

05 Research Path Example

Deployment of a wearable device for cardiac disease diagnostics

1. Find research benefits.
   • effective and continuously vigilant cardiac condition monitoring. (AS detection)
2. Find Scientific foundations for the effectiveness of wearable monitoring.
   • PCG signal characterization → detection of heart sound murmurs in the cardiac systolic interval.
3. Find methods for validation of system performance.
   • The gold standard for AS assessment + blinded review of acoustic records by an expert.
   • independent reviews of subject condition → complete agreement regarding identification.
4. Design of clinical trials for both development and validation. (get data)
   • 96 patients with a broad range of cardiac conditions..
   • Additional institutional review board approval
5. Design clinical workflow.
   • straightforward
   • measurement episode is brief and less than 10 min.
6. Design wearable transducers.
   • passive characteristic → inherent safety
7. Development of analytical methods for condition indication.
   • noise subtraction and filtering signal waveform from PCG and ECG sensors.
   • SVM classifier
8. Evaluation of system performance.
   • standard t-test
   • ROC and AUC curve analysis
9. Ensuring compliance for continuous monitoring.
10. Vigilant detection and notification of failure.
11. Privacy and security check

 

06 Conclusion

This article describes the challenges and constraints for research in computing for Healthcare.

Unprecedented benefits in vigilant monitoring of individuals with wearable technology

→ bend the cost curve in healthcare.
→ rapid diagnostic and event detection.
→ deployment of new AI methods for the prediction of healthcare problems. (the new Computing for healthcare community will benefit from shared data.)