IoT Empowerment in Healthcare: Detailed ECG Analysis and Prediction by using 2D Gaussian Filter

Q: What is the clinical evidence from this study?

Population: Vital signs monitoring and ECG classification (n=3). Intervention: IoT-based patient monitoring system and Inception V3 deep learning model vs. CE registered devices (pulse oximeter, mercury thermometer). Primary outcome: Validation accuracy of the Inception V3 deep learning model for ECG classification.

Q: What does this research mean for the field?

An IoT-enabled monitoring system combining 2D Gaussian noise filtering with the Inception V3 deep learning model classifies ECG signals with 96.90% validation accuracy. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Asif Raza; Khaliq Ahmed; Maheen Danish; Kashif Sheikh; Fozia Noor

doi:10.32350/icr.41.04

June 12, 2024Innovative Computing ReviewOpen Access

IoT Empowerment in Healthcare: Detailed ECG Analysis and Prediction by using 2D Gaussian Filter

View Full Paper

Ask AI

Bookmark

Share

Key result

Inception V3 deep learning model classifies ECGs with ~97% accuracy in an IoT monitoring prototype.

n=3

Why the study?

Remote and portable patient monitoring systems are increasingly important due to population growth, rising health issues, and limited access to medical facilities.

Key Points

To develop and validate a portable, cost-effective IoT-based patient monitoring system that records vital signs and classifies ECG signals using deep learning.
Built a hardware prototype with an ESP32 microcontroller, a DS18B20 temperature sensor, a MAX30100 module for SpO2, and custom sensors for heart rate and ECG, transmitting data to an LCD and a cloud SQL database.
Cleaned ECG images from the CPEIC dataset using two-dimensional Gaussian filtering to eliminate image noise and classified cardiac patterns using an Inception V3 convolutional neural network.
Conducted 2-minute continuous measurement trials on three human volunteers to compare prototype heart rate and temperature outputs against a CE-registered pulse oximeter and a mercury thermometer.
The Inception V3 model achieved a training loss of 0.1315 with a training accuracy of 96.66% on the ECG dataset.
On the validation set, the deep learning model achieved a validation loss of 0.1146 and a validation accuracy of 96.90%.
Volunteer testing demonstrated prototype pulse readings of 76.3, 79.6, and 65.6 bpm (benchmark pulse oximeter: 78, 82, and 67 bpm) and average temperatures of 35.77°C, 36.13°C, and 37.7°C.

Structured PICO

P

Population

Three volunteers underwent testing of a prototype IoT-based patient monitoring system for vital signs, alongside validation of an Inception V3 deep learning model for ECG classification.

I

Intervention

IoT-based patient monitoring system measuring ECG, SpO2, heart rate, and body temperature, utilizing an Inception V3 deep learning model and 2D Gaussian filter for ECG classification.

C

Comparator

CE approved devices (pulse oximeter, mercury thermometer) for vital signs; no comparator for the machine learning model.

O

Outcome

Model accuracy, precision, recall, and F1-score for ECG classification; sensor accuracy compared to CE approved devices.surrogate

An IoT-based patient monitoring system using an Inception V3 deep learning model can accurately classify ECG signals and reliably measure vital signs.

Limitations

Very small sample size (n=3) for hardware prototype testing
Need for better integration and interoperability among devices from different manufacturers
Data security and privacy concerns regarding sensitive patient data
Many RPM systems lack user-centric design, potentially leading to low adoption rates
Prototype failed to incorporate advanced level sensors because of their high cost.

Implication

May support affordable IoT-ECG monitoring development; leaves open prospective clinical validation before adoption.

Abstract

Currently, healthcare sector is the most dynamic sector in terms of introducing new technologies and services. An innovative advancement in this sector is the remote or portable monitoring of patients, which is proving to be very beneficial in a world with a rapidly expanding population, rising health issues, and limited access to medical facilities. A patient monitoring equipment is often used to quantitatively measure a patient's vital signs, including blood pressure, temperature, ECG, heart rate, and SpO2. This study attempts to enhance the future of healthcare by developing such a patient monitoring system. The ECG dataset was downloaded from CPEIC. The device created for this study can measure four parameters, namely ECG, SpO2, heart rate, and body temperature. The values are displayed on an LCD screen and the device is IoT-based, allowing data transmission to a web application for easy and universal access. To address the issues encountered, the device is designed to be cost-effective, dependable, and portable. The ECG output of this IoT-based model was thoroughly examined and tested using a deep learning model known as Inception V3 to determine the accuracy and dependability of the network. The model obtained phenomenal training loss of 0.1315 and a training accuracy of 96.66%. On the validation set, it achieved a validation loss of 0.1146 and a validation accuracy of 96.90%. Two-dimensional Gaussian elimination was used to remove noise from ECG images.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Raza et al. (2024) studied Vital signs monitoring and ECG classification (n=3). IoT-based patient monitoring system and Inception V3 deep learning model vs. CE registered devices (pulse oximeter, mercury thermometer) was evaluated on Validation accuracy of the Inception V3 deep learning model for ECG classification. The Inception V3 deep learning model achieved a training accuracy of 96.66% and a validation accuracy of 96.90% for ECG classification, while the IoT prototype accurately measured vital signs.

synapsesocial.com/papers/6ac2fd227180a622ddd4dfc7 https://doi.org/10.32350/icr.41.04

Topics

Artificial intelligence in cardiology

View Full Paper

Ask AI

Bookmark

Share

IoT Empowerment in Healthcare: Detailed ECG Analysis and Prediction by using 2D Gaussian Filter

Key Points

Structured PICO

Limitations

Implication

Abstract

Discussion

Cite This Study

Topics

IoT Empowerment in Healthcare: Detailed ECG Analysis and Prediction by using 2D Gaussian Filter

Key Points

Structured PICO

Limitations

Implication

Abstract

Discussion

Cite This Study

Keep reading

Topics

Keep reading