Fog Data: Enhancing Telehealth Big Data Through Fog Computing

Harishchandra Dubey; Jing Yang; Nick Constant; Amir Mohammad Amiri; Qing Yang; Kunal Makodiya

Source-linked AI summary

Fog Data: Enhancing Telehealth Big Data Through Fog Computing

Harishchandra Dubey, Jing Yang, Nick Constant, Amir Mohammad Amiri, Qing Yang, Kunal Makodiya

arXiv:1605.09437v2May 30, 2016cs.CY

TL;DR

Telehealth sensors produce rapidly growing multimodal data that requires reduction and edge analytics. The paper proposes and evaluates Fog Data, a low-power service-oriented architecture that mines timestamped sensor data locally and sends unique patterns and clinically relevant information to the cloud. Case studies on speech disorders and cardiovascular monitoring report orders-of-magnitude data reduction and substantial system-efficiency improvements, while embedded processors impose resource constraints on compression methods.

Problem
Growing multimodal, heterogeneous telehealth sensor data requires intelligent reduction, mining, and analytics at edge devices to limit bandwidth and transmission power.
Method
Fog Data uses a low-power embedded Fog computer to analyze wearable-sensor time series, identify similar patterns, and extract clinically relevant information before cloud transmission.
Results
Case studies on speech motor disorders and cardiovascular problems showed orders-of-magnitude data reduction and substantial improvement in system efficiency.
Takeaways & Limitations
Fog Data can reduce telehealth cloud storage, data transmission, and transmission power by performing healthcare-data processing locally.
Takeaways & Limitations
Embedded Fog processors constrain compression choices because higher-ratio programs require CPU power and RAM unsuitable for these devices.

Abstract

from arXiv · show

The size of multi-modal, heterogeneous data collected through various sensors is growing exponentially. It demands intelligent data reduction, data mining and analytics at edge devices. Data compression can reduce the network bandwidth and transmission power consumed by edge devices. This paper proposes, validates and evaluates Fog Data, a service-oriented architecture for Fog computing. The center piece of the proposed architecture is a low power embedded computer that carries out data mining and data analytics on raw data collected from various wearable sensors used for telehealth applications. The embedded computer collects the sensed data as time series, analyzes it, and finds similar patterns present. Patterns are stored, and unique patterns are transmited. Also, the embedded computer extracts clinically relevant information that is sent to the cloud. A working prototype of the proposed architecture was built and used to carry out case studies on telehealth big data applications. Specifically, our case studies used the data from the sensors worn by patients with either speech motor disorders or cardiovascular problems. We implemented and evaluated both generic and application specific data mining techniques to show orders of magnitude data reduction and hence transmission power savings. Quantitative evaluations were conducted for comparing various data mining techniques and standard data compression techniques. The obtained results showed substantial improvement in system efficiency using the Fog Data architecture.

1. INTRODUCTION

Telehealth wearable sensors generate high-volume, heterogeneous, multimodal data that requires efficient processing. Fog Data addresses these challenges through service-oriented Fog computing with onsite analytics and filtering before cloud transmission.

1. INTRODUCTION: Wearable and ambient sensors enable continuous, non-invasive telehealth monitoring but generate large, heterogeneous, multimodal datasets requiring real-time processing.The paper identifies collection, storage, and analysis of distributed medical big data as a central deployment challenge.
1. INTRODUCTION: Figure 1 presents Fog Data as a service-oriented architecture intended to reduce storage requirements and increase telehealth big-data efficiency.The figure summarizes the architecture’s stated system-level objectives.
1. INTRODUCTION: Fog Data is a service-oriented architecture that interfaces with in-home telehealth devices for person-centered health monitoring.Its distinctive feature is onsite analytics that reduces data stored and transmitted to the cloud.
1. INTRODUCTION: The architecture uses wearable and ambient devices to collect raw time-series signals, then an energy-efficient embedded Fog computer preliminarily analyzes and filters them.The Fog computer transmits necessary data to the cloud after local processing.

2. RELATED WORKS

Telehealth extends healthcare monitoring into homes through wearable sensing, but its sensor-rich infrastructure produces medical big data. Fog computing addresses this setting by providing local connectivity, computation, storage, and security before cloud analysis.

2.1. TeleHealth and Medical Big Data: Telehealth uses wearable sensors and body sensor networks to monitor patients remotely in homes and other non-clinical settings.Examples include continuous vital-sign, COPD, activity, and sleep-quality monitoring systems.
2.1. TeleHealth and Medical Big Data: Sensor-rich telehealth infrastructure generates medical big data for remote diagnosis and clinical interventions.The data must support health monitoring and disease management outside hospitals.
2.2. Fog Computing as a Smart Gateway: Fog computing deploys computing services on remote devices to provide low latency, high efficiency, and high reliability for end-user applications.The paper presents it as a smart gateway between wearable sensing and cloud services.
2.2. Fog Computing as a Smart Gateway: As a smart gateway, Fog computing acquires sensor data, processes it into medical-grade logs, maintains an onsite database, and supports data security.Processing can range from simple filtering to complex wavelet analysis, while local records can be queried internally and externally.
2.2. Fog Computing as a Smart Gateway: The paper implements Fog computing infrastructure and demonstrates two case studies on wearable telehealth data.The stated case studies address speech disorders and ECG-related healthcare problems.

3. THE ARCHITECTURE OF FOG DATA

Fog Data connects wearable sensing, local Fog processing, and cloud storage and analysis. Its information flow converts timestamped time-series health data into features and patterns while retaining clinically relevant outputs for cloud use.

3. THE ARCHITECTURE OF FOG DATA: Fog Data is organized as three subsystems: a body sensor network for acquisition, a Fog gateway computer for onsite processing, and a cloud server for storage and back-end analysis.The Fog gateway connects patients and wearable sensors with physicians.
3. THE ARCHITECTURE OF FOG DATA: The Fog computer reduces cloud storage and Internet bandwidth needs by processing health data locally and sending clinically relevant information.It supports on-demand computation of relevant features from accumulated health data.
3. THE ARCHITECTURE OF FOG DATA: Fog Data supports diverse wearable sensors, including smartwatches, wearable ECG systems, and pulse glasses, while accommodating their limited memory and computing resources.The architecture provides accumulation for continuously monitored patient data.
3. THE ARCHITECTURE OF FOG DATA: Fog Computer algorithms must be computationally simple and clinically relevant, while the cloud executes complex algorithms for diagnosis, monitoring, and treatment.Data is deleted from the Fog computer after transmission or feature extraction.
3.1 Information Flow in Fog Data Architecture: Timestamped time-series data from wearable sensors is processed into features and patterns, with log files recording acquisition times and other operational factors.The paper identifies speech, ECG, and other clinical vital-sign data as supported inputs.

4. METHODS AND ALGORITHMS

The paper implements lightweight analytics and compression routines on an Intel Edison Fog computer to reduce and process telehealth data locally. Methods include DTW pattern mining, CLIP speech-feature extraction, and GNU zip compression, selected for embedded-resource constraints.

Dynamic Time Warping: DTW mines similar patterns in time-series data despite differences in series length and phase.It builds an adjacency matrix and finds the shortest path across it.
Clinical Speech Processing Chain (CLIP): CLIP filters speech data to compute clinically relevant metrics such as loudness and fundamental frequency.The implementation uses a coarse loudness estimate based on speech-signal amplitude because the Fog processor has limited computing resources.
Compression: GNU zip compresses data on the Fog processor before transmission, while the cloud decompresses it for subsequent processing.Higher-ratio alternatives such as bzip2 and 7-zip require more CPU power and RAM, making them less suitable for embedded Fog processors.
Onsite Processing: The Fog processor runs clinical speech processing, Pan-Tompkins ECG analysis, and DTW-based pattern matching locally.These routines are intended to reduce data transmission and storage while extracting clinically relevant information.
Power Consumption: Intel Edison measurements include 862 mW for DTW computations and 937 mW for encrypted Wi-Fi data transmission.The DTW computations were occasional and short, and data reduction also reduced transmission and cloud-storage power.

5. EXPERIMENTAL SETUP

The experimental setup uses an Intel Edison as the low-power Fog platform, with clinical requirements determining search patterns and DTW configured to transmit only the most similar match within a waiting window. Figure 3 compares percentage data reduction for speech-processing and compression methods.

Hardware Platform: The experiments use an Intel Edison Fog platform with rechargeable lithium-ion battery or plug-in AC power.The platform includes a dual-core 500 MHz Intel Atom CPU, 1 GB LPDDR3 memory, and 4 GB eMMC storage.
DTW Configuration: Clinical requirements determine the DTW reference pattern; a 40 ms speech segment is used in the speech-disorder case study.A waiting window sends only the most similar detected pattern to reduce network bandwidth and improve accuracy.

6. FOG DATA: CASE STUDIES

Two case studies evaluate Fog Data for speech monitoring in Parkinson’s disease and ECG monitoring. The speech study compares feature extraction, pattern mining, and compression, while the ECG study compares reduction and processing time.

Speech Motor Disorders: The speech case study sends smartwatch speech to an Intel Edison, which extracts features and pattern indices for cloud access by speech-language pathologists.The extracted features support remote access to loudness and pitch information for customizing speech exercises.
Speech Motor Disorders: More than 99% data reduction is achieved by DTW, which performs close to CLIP in the Parkinson’s disease speech case study.CLIP converts speech time-series into average loudness and average fundamental frequency; both CLIP and DTW are lossy.
Speech Motor Disorders: GNU zip provides reversible lossless compression but achieves lower data reduction than the lossy CLIP and DTW methods.The cloud can decompress GNU zip output to regenerate the original speech time-series.
ECG Monitoring: ECG processing time is always less than 15 seconds, supporting the architecture’s use for real-time ECG monitoring.The cited result explicitly presents this timing as validation of Fog Data for real-time monitoring.
ECG Monitoring: The ECG case study uses MIT-BIH Arrhythmia Database data to compare DTW and GNU zip reduction and to measure Pan-Tompkins QRS-detection execution time.The figure reports execution time in seconds on the Intel Edison Fog computer.

7. CONCLUSIONS

The paper validates Fog Data for speech-disorder and ECG telehealth applications and reports potential reductions in telehealth logistics, cloud storage, and edge-device transmission power. It also identifies additional speech features and in-home validation as future work.

Validated Applications: Fog Data is validated for two healthcare problems: speech disorders and ECG monitoring.The conclusion frames both applications as demonstrations of the architecture’s use.
Potential Benefits: The architecture has potential to reduce telehealth logistics requirements, cloud storage, and transmission power at edge devices.These are stated as potential benefits rather than established outcomes across all telehealth applications.
Future Work: Future work includes computing shimmer, jitter, and sensory pleasantness and conducting in-home validation studies of EchoWear.These additions would extend the speech features beyond loudness and fundamental frequency.