Sie sind auf Seite 1von 6

GRD Journals- Global Research and Development Journal for Engineering | Volume 5 | Issue 10 | September 2020

ISSN- 2455-5703

Pulse Oximeter Design to Predict COVID-19


Possibilities on Patient’s Health using Machine
Learning
Dharmendrasinh Revar
Master of Engineer
Department of Electronic and Communication Engineering
Gujarat Technological University

Jayvirsinh Sevaniya Vedant Joshi


Master of Engineer Master of Engineer
Department of Electronic and Communication Engineering Department of Electronic and Communication Engineering
Gujarat Technological University Gujarat Technological University

Abstract
Today world is globally suffered from corona (COVID-19). Till the date no proper and effective vaccines are found that gives
proper relaxation from COVID-19. Pulse oximeter is the non-invasive device that can collect data of SPO2 and Heart Rate. Based
on data Machine Learning have abilities to predict the corona patient’s situations. The circuitry is used for making pulse oximeter
is MAX30100 MODULE, NODE MCU ESP8266 WI-FI MODULE and 16*2 JHD LCD DISPLAY. The pulse oximeter data is
communicated with blynk cloud platform. This Paper mainly provide facilities to predict the patient’s situations with help of ML
(machine learning) algorithms. This paper provides actual values and predicted values of SPO2 and Heart Rate comparison and
also predicts the values of these parameters at future time with graph. This paper mainly classifies the COVID-19 Patient’s
conditions on normal, at risk and critical situations.
Keywords- COVID-19, Pulse Oximeter, Machine Learning, Machine Learning, Support Vector Machine (SVM),
Decision Tree Algorithm

I. INTRODUCTION
Pulse oximeter is the device that work on the principle of Pulse oximetry. Pulse oximetry is the method for monitoring the people’s
oxygen saturation (SpO2) without harming their body. The Standard SpO2 level is between 95 to 100 percentages. SpO2 level is
mainly depends on R (ratio) value. R value is the ratio value of Red light to Infrared light. The wave length of Red is at 660nm
and IR (infrared) is at 940nm. Pulse oximeter checks to see if people’s blood oxygen level is within a healthy range or not. When
your oxygen level is down from normal range, it is called hypoxemia. When person’s oxygen level falls down from normal range,
the person’s lungs starting to become narrow, it is res started to slower down the blood flow by the lungs. When these things
happen, more stress has been placed on the heart. This could be led to Serious Heart failure problem for people. Machine leaning
is the way to teach the Algorithms to learn from its past experience and updated itself. Machine learning is a part of AI (artificial
intelligence). Machine Learning algorithms are build model in Joblib file on sample data. On the sample Data the partition should
be possible for train and test data.so that, it is used later to measure the accuracy of the algorithms.
There are three types of Machine Learning given below:
– Supervised Machine Learning
– Unsupervised Machine Learning
– Reinforcement Machine Learning
Based on the Data the machine learning algorithms are two types given below:
– Classification
– Regression
The programmer decides to base on data they want to choose classification or regression. Normally when we want to
classify some values at that time Classification algorithms are used and when we want to predict continues values like temperature
at that time regression algorithms are used.

A. Components Needed
– JHD 162A 16*2 LCD display
– ESP8266 (NODEMCU) Wi-Fi module
– Bread board

All rights reserved by www.grdjournals.com 9


Pulse Oximeter Design to Predict COVID-19 Possibilities on Patient’s Health using Machine Learning
(GRDJE/ Volume 5 / Issue 10 / 002)

– Male to Male, Female to Male


– And Female to Female connecter
– MAX 30100 moduleI2C Module

II. OVERVIEW & HARDWARE IMPLEMENTATION


The software for writing the code is Arduino IDE (Integrated Development Environment). The code is similar to C language code.
The sketch written in Arduino IDE when it complied successfully after that code is uploaded to Node MCU. The compiling code
is given in Fig.3.5. The results of Spo2 and Heart rate shown in display on Fig.3.6.
For Compilation of code the following libraries are added compulsory:
– MAX30100_PulseOximeter
– Wire.h
– ESP8266WiFi.h
– LiquidCrystal_I2C.h

Fig. 1: Prototype of Pulse Oximeter

A. Preparation of App on Blynk Platform


Blynk is open source IoT platform that provide connectivity to connect the IoT devices to them. Blynk is also provide the facility
to making our own app as per our device configuration. The data in Blynk app is secured and reliable with authentication key.
Features are provided by Blynk Platform is very useful and reliable.

Fig. 2: Code compilation of LCD, Node MCU and MAX30100

Fig. 3: Pin Diagram

All rights reserved by www.grdjournals.com 10


Pulse Oximeter Design to Predict COVID-19 Possibilities on Patient’s Health using Machine Learning
(GRDJE/ Volume 5 / Issue 10 / 002)

III. METHODOLOGY & SOFTWARE IMPLEMENTATION

A. Implementation of Machine Learning


The dataset (SPO2, Heart Rate, Age and gender) gathered by the pulse oximeter of various person can be used for the training and
testing of the model. Based on that the idea of, which machine learning algorithm should be taken or which machine learning
algorithm gives the best result on the dataset can be selected. The Machine Learning implementation here is split into three parts.
(1) Prediction of SPO2 and/or HR with others features given (Supervised Machine Learning) (2) Prediction of SPO2 and/or HR at
particular time (Supervised Machine Learning) (3) Prediction of person’s health based on their possibilities for COVID-19.

Fig. 4: Pulse Oximeter App Output on Blynk IoT Platform

IV. RESULT

A. SPO2 and/or HR Predictions


The dataset consists of the features listed above and, on that dataset, the various machine learning algorithms are applied. The
analysis of various algorithms performance on the dataset is shown in table below. The table shows the Support Vector Machine
(SVM) algorithm gives us the best results as compared to other algorithms listed in the table 1. Based on SVM we build a model
that predicts the person’s SPO2 and/or HR. The accuracy and model prediction of the SVM algorithm is as shown in figure below.
The table of the prediction of SPO2 and HR is as shown in table respectively below. The SVM algorithm is one of the supervised
learning algorithms of ML. It simply gives the result of predicted value based on the best decision boundary. The predicted value
gives the idea of what necessary precautions should be taken to take care of particular predicted value. The person hence can make
self-assessment of health.
Table 1: ML algorithm analysis on dataset
NO. ALGORITHM RMSE MAE ACCURACY
1. Linear Regression (LR) 3.15 2.19 96.58
2. Support Vector Machine (SVM) 3.87 2.67 97.13
3. K Nearest Neighbor (KNN) 5.94 3.96 95.82
4. Decision Tree (DT) 4.88 3.23 96.54
5. Random Forest Regression (RFR) 3.86 2.68 97.17

Fig. 5: Model Accuracy of SVM

All rights reserved by www.grdjournals.com 11


Pulse Oximeter Design to Predict COVID-19 Possibilities on Patient’s Health using Machine Learning
(GRDJE/ Volume 5 / Issue 10 / 002)

Fig 6: Model output

Table 2: SP02 predictions and actual value


SPO2
NO. ACTUAL PREDICTED
1. 96.25 98
2. 96.4166666666666 100
3. 88.9166666666666 98
4. 97.0833333333333 98
5. 91.1666666666666 99.96
6. 99 100
Table 3: HR predictions and actual value
HR
NO. ACTUAL PREDICTED
1. 93.0844545454545 100
2. 97 89
3. 78.8185 76
4. 83.53391 89
5. 77.01358 76
6. 73 70

B. Prediction of SPO2 and/or HR at Particular Time


Only knowing the value of SPO2 and/or HR is not sufficient as person cannot get the idea about his/her health for future time. The
future time prediction values of SPO2 and/or HR gives the idea about the health of person in near future as algorithm measures the
COVID-19 parameters (SPO2 and HR). The accuracy and comparison table of algorithm-based prediction for particular time is
shown in figure below. This can be proven useful in medical science for detecting the person’s health for next day and according
to it doctors can feed the dose of medicines. By this way the advance health prediction of COVID-19 patient can also be done.
Table 4: ML algorithm analysis on dataset for HR
HR
NO. ALGORITHM RMSE MAE ACCURACY
1. Linear Regression (LR) 12.7969 7.68 90.54%
2. Decision Tree (DT) 13.1155 7.68 89.81%
3. Random Forest Regression (RFR) 13.1844 8.15 90.02%
Table 5: ML algorithm analysis on dataset for SPO2
HR
NO. ALGORITHM RMSE MAE ACCURACY
1. Linear Regression (LR) 0.6708 0.44 99.54%
2. Decision Tree (DT) 0.7589 0.39 99.59%
3. Random Forest Regression (RFR) 0.7171 0.36 99.62%
4. K Nearest Neighbour (KNN) 0.7950 0.40 99.58%

Fig. 7: SPO2 variations graph vs Time

All rights reserved by www.grdjournals.com 12


Pulse Oximeter Design to Predict COVID-19 Possibilities on Patient’s Health using Machine Learning
(GRDJE/ Volume 5 / Issue 10 / 002)

C. Prediction of Covid-19 Possibilities


Based on SPO2 and HR ratings measured from pulse oximeter, the machine learning can be used to predict the chances of or say
possibilities of that particular person being infected from COVID-19. The various categories are taken based on the readings and
from the standard medical measurements. The person is said to be safe or say “Normal” from COVID-19 symptoms if SPO2 level
is said to be in between range of 95 to 100, and Heart Rate (HR) is said to be in between the range of 60 to 100. Same as Normal
condition, “At risk” from COVID-19 and “Critical” from COVID-19 conditions are also categorized by the algorithm. “Heart Rate
at Risk”, “Heart Rate Critical”, “SPO2 at Risk”, “SPO2 Critical” are also categorized in the model output. The person’s age is
most mattered feature in this prediction as SPO2 and HR for healthy human are different for different ages. The results of some
predictions from model is shown in figure below. The Decision Tree algorithm is used for the classification of this part. The table
of predicted disease name corresponding to predicted numbered output in model is shown below. The model can accurately predicts
based on the training and can be useful for fast COVID-19 testing without using any kit and doctors advice.

Fig 8: Patient age 71 Fig 9: Patient age 42

Fig 10: Patient age 66 Fig 11: Patient age 26

Fig 12: Prediction vs Time (All Samples) Fig 13: Patient age 31
Fig 8 - 13: SPO2 Actual & Predicted vs Time for different age patient

All rights reserved by www.grdjournals.com 13


Pulse Oximeter Design to Predict COVID-19 Possibilities on Patient’s Health using Machine Learning
(GRDJE/ Volume 5 / Issue 10 / 002)

Table 6: Model output corresponding meanings


No. Disease Name Model Output
1. Corona at Risk 0
2. Corona Critical 1
3. HR Critical 2
4. HR at Risk 3
5. Normal 4
6. SPO2 Critical 5
7. SPO2 at Risk 6
If the Heart Rate and SPO2 values are different than normal value, then person should bring bake that value to normal otherwise
it should produce serious diseases (i.e. techy cardia, die cardia etc.)

Fig 14: Prediction of Normal condition from Corona

Fig 15: Prediction of Critical Condition from Corona

Fig 16: Prediction of at-Risk condition from Corona

V. CONCLUSION
Taking readings of Heart Rate (HR) and SPO2 level of the person from the Pulse Oximeter reduces the effort of doctors and as it
is simple to use, Pulse Oximeter can be used at home. The readings can be entered into the learned model to predict the SPO2 and
HR of the person for the next sometime to get the idea of the precautions should be taken by the caretakers. Based on the entered
readings and patient’s age, the COVID-19 chances of that particular person will also be predicted with the categories are High risk,
Critical and Normal. This will reduce the human effort and required no pre-requisites, hence easy to use, accurate. This will save
the human time and efforts.

REFERENCES
[1] Dharmendrasinh revar, D.K Jhala and Hemal Nayak “Design and implementation of pulse oximeter to monitor and Predict Patient’s health”, Compliance
Engineering journal, Volume-11, Issue-8, August-2020
[2] Rodriguez LR, Kotin N, Lowenthal D, Kattan M: A study of pediatric house staff's knowledge of pulse oximetry. Pediatrics 1994; 93:810-813
[3] Kelleher JF: Pulse oximetry. J Clin Monit 1989; 5:37-62
[4] Grace RF: Pulse oximetry. Gold standard or false sense of security? Med J Aust 1994; 160:638-644
[5] Tremper KK, Barker S J: Pulse oximetry. Anesthesiology 1989; 70:98-108
[6] L. Schallom, C. Sona, M. McSweeney, and J. Mazuski, “Comparison of forehead and digit oximetry in surgical/trauma patients at risk for decreased peripheral
perfusion,” Heart & Lung: The Journal of Acute and Critical Care, vol. 36, no. 3, pp. 188–194, 2007.

All rights reserved by www.grdjournals.com 14

Das könnte Ihnen auch gefallen