Uploaded by Aya A.elsamad

Poster 12332 T

advertisement
What about your heart?
Rahma Mohamed – Aya Abdelsamed – Mariam Ramadan
Alexandria STEM School, G.12, Semester 1,
G.No. 12332, 2020-2021.
Key words: Cardiovascular disease, instructions,
Artificial intelligence, Machine learning.
Abstract:
The current Speech establishing technology would accordingly create a
whole new era of human interactions with machines. Solving the
challenges of public health issues, urban congestion, and Improving the
scientifically and technological environment with artificial intelligence
was a great opportunity. The considerable challenge that we focused on
was the public health issues. Cardiovascular disease are the first cause of
death globally taking an estimated 17.9 million lives each year. This belief
has lit our mind with the idea of developing diagnoses for (CVD) by
creating a website with 13 parameters which are (Age, gender, cp,
Trestbps,chol, FBS, restecg, thalach, exang, old peak, Slope, ca, Thal).
According to those parameters, the patient enters the AI model to decide
if the patient has CVD. If the users have CVD, they will be given some
instructions that they will need according to their situation to avoid heart
attack without wasting time waiting or money. The solution's success will
be tested by some design requirements of high accuracy, easy to use with
simple interference, and using AI instead of traditional programming. The
prototype that will be tested consists of AI coding with an interface of
software website. Thankfully, our project shows a considerable success in
testing and achieving the design requirement as our project's accuracy was
89.06 ± 0.03% and had achieved the requirement of being AI not
traditional programming by deciding without any intervention and easy to
use by any person. In conclusion, the diagnostic will help the patients in
many ways and solves particular grand challenges mentioned before, also
achieved the design requirements.
Introduction:
Multiple risk factors characterize the Egyptian health system as critical
challenges such as the healthcare, urban congestion and technology
systems in Egypt. Still, the most critical one is health care because people's
health and life are more precious than anything else. The most dangerous
challenge in the health care system is misdiagnosis. Medical errors
estimated 40,000 to 80,000 deaths occur each year in Egypt hospitals
related to misdiagnosis. After searching about prior solutions, there is a
website called "intelligent Heart disease prediction system". It has AI
capabilities in heart disease diagnosis prediction without consulting a
doctor by entering some available data such as blood cholesterol and blood
pressure; besides, the user can get the result of diagnosing in 10 seconds.
Nevertheless, it has some limitations. For instance, it has limited accuracy,
and it doesn't have specific diagnosing in heart disease type.
Consequently, we decided to work on this idea's limitations and make
developments based on some design requirements such as increasing the
accuracy, easiness in using and the solution must demonstrate AI coding,
not general coding. It has been focused on cardiovascular diseases because
it is one of the uppermost risky diseases. CVDs are a group of disorders
of the heart and blood vessels, which are the number 1 cause of death
globally, taking an estimated 17.9 million lives each year for many
reasons. One of them is misidentification. Therefore, it is necessary to
provide strategies for improving the diagnosis of CVD in the future.
Currently, artificial intelligence (AI) may have the potential to solve this
problem. The chosen solution is creating a website for cardiovascular
disease predictions and give the patient some benefit instructions to avoid
a heart attack. The solution execution splits into three-part; the first part,
creating machine learning (AI) code with specially chosen parameters
such as (gender, cp, restbps, restecg….etc), second, collecting the
database for Model's training, third, making understandable interface
(website) including the instructions using python language.
Finally, the website achieved the design requirements as it has high
accuracy, which is 89.06± 0.03%. It is easy in use because of using general
understandable words available in medical tests which facilitates
transferring data from tests to the website. The project demonstrates AI
coding, not general coding by testing the code; it gives us the results
according to the database and algorithms without any conditions from the
code.
Materials and Methods:
Materials:
Name
Python
programming
language
Streamlit library
Anaconda
navigator
Spider
Jupyter
notebook
free integrated
development
environment that is
included with
Anaconda.
An open-source
web application
that allows you
to create live
code.
Figure
Description high-level
open-source
programming Python library that.
language with
dynamic
semantics.
Methods:
The problem was identified as the deterioration of public health and
improving Egypt's healthcare technology by using artificial intelligence.
This problem's chosen solution was to make an AI model for diagnosing
if the patient has cardiovascular heart disease or not by some parameters
and algorithms and chosen website interface. Three design requirements
were set for the website: (Easiness to use, AI code not traditional
programming, and accuracy).
According to methodology, there are many steps we followed:
1- Installing anaconda navigator to be used as a platform for writing and
running the algorithms.
2- Selecting the parameters that the AI model will diagnose according to
it.
3- Electing the dataset used in training and testing the model.
4- Writing functions and AI algorithms by using python programing
language.
5- Dividing the dataset into two parts: the first part of the data set was
selected to train the model with, and the second part was selected to test if
the model works appropriately.
6- Selecting the instructions for each parameter according to the standard
and unnormal ranges of each parameter.
7- Creating a website as an interface that the user will deal with by using
streamlit library.
8- Testing the system as a user by using the dataset that the model did not
deal with before and comparing the result of the model with the dataset
result.
9- Testing if the system meets the design requirements or not.
10- Developing and modifying the model.
Test plan:
After deciding how to solve the problem of health care with the help of
artificial intelligence and making the website, the model was tested to
show if the website can be used by the patients or not. Testing the
prototype included 3 stages to test the design requirements.
(1)Accuracy: which is the main design requirement it was calculated by
an AI algorithm as shown in (fig.1) and the results was 89.06 ± 0.03%.
Which is a positive indicator about the success of the project and a huge
improvement to the previous solutions.
Fig.1: the accuracy algorithm
(2) AI codes not traditional programing: the AI codes was testing by
taking the decision without any human intervention, unlike the traditional
programming, so it can give you the results according to your parameters,
this design requirement was tested by splitting the data using these code
(fig.2) into: 1- training dataset which the model train to diagnose with and
equal (239) users' data. 2- Testing dataset that we test the model with and
equal (64) users' data.
Fig.2: splitting data code
(3) Easy to use: this was tested by making a survey and show people's
opinion which shows that the website is easy to use.
Results:
1- Accuracy: The result was 89.06 ± 0.03% by doing the calculation
according to the data in (fig.3) using the algorithm which is (X_test,
y_test)*100.
Fig.3: Graph describes the amount of applicants who rate the
easiness in use the website from 1-5.
2- AI programming not general code: the model gives the user the
predictions according to the dataset and algorithms and it doesn’t need
any conditions or orders
3-Easy to use: the survey showed that the project is very easy to use as
shown in (fig.4).
Fig.4: the amount of True and False predictions diagnostic cases
Analysis:
Like many other countries, Egypt is suffering from many problems that
affect the economy, environment, social life and sustainability. These
problems represent an obstacle forward its development. The most
common issues that face Egypt during these years are health care problems;
thus; we are trying hard to solve these problems by creating new solutions.
The problem was chosen to be solved the problem of cardiovascular
disease misdiagnosis, which increases the death rate due to heart attack and
increases the congestion of patients in hospitals.
How does our prototype solve this problem?
The website will solve this problem with an accurate diagnosis of
cardiovascular disease. Furthermore, it will reduce the death rate by giving
instructions to reduce the probability of a heart attack. Consequently, it will
reduce the congestion in hospitals as it’s full of viruses such as coronavirus
disease.
AI model:
Both the model and the website were written by python programming
language. The model includes a dataset of real cardiovascular disease
patients. Data splitting was used to split the data set used into two parts.
The first part is used to train the model, while the second part for validation
is used to test the model if it works appropriately. Moreover, there is an AI
code for calculating the tested model's accuracy, as shown in (fig 5). Then,
the model was connected to the interface.
Fig.5: the most important codes in the model
The website:
The interface (fig 6) contains 13 parameters (age, sex, cp, trestbps, chol,
FBS, restecg, thalach, exang, oldpeak, slope, ca, thal). The user will enter
the data in numerical value or choose the data from a select box, then the
user’s data will be diagnosed by AI code, and the result will appear on the
website. If the user is sick, the website will give the patient some
instructions (fig 7) according to his case, to avoid a heart attack. For
example, the patient has a high blood pressure level, so the website will
give the patient some medical instructions from (WHO) to lower his
pressure level.
Fig.7: instructions sample
Fig.6: the interface of website
Design requirements:
After testing, we have achieved the selected design requirements: high
accuracy, ease in use, and AI code, not traditional programming.
1High accuracy: the model has a suitable number of a dataset for
training, making it has high accuracy. It was calculated by substituting
matrix values of the model in the law (fig.8) where TP = True Positives,
TN = True Negatives, FP = False Positives, and FN = False Negatives. ,
the result was 89.06 ± 0.03% with an error which is excellent performance.
Fig.8: Law of accuracy
2Easiness in use: the website provides simple, understandable words
as it gives the meaning of each abbreviation, in additions, the website
requires numerical data or choose from selection box as it doesn’t require
significant effort in entering data. It was determined by asking users how
easy to deal with the website by giving it a rate from 1-5, and the average
was 4.5, which is a considerable value.
3Creating AI code, not traditional programing: the model was tested.
It gives the results according to its data set and algorithms without human
intervention, making it AI code because traditional code wants conditions
and statements to give the results like IOT coding.
After constructing the interface, some modifications were made
based on the comments and feedback on the project.
1The website's accuracy increased from 78 to 89.06 ± 0.03% by
increasing the dataset from 200 to 300, which makes the model has large
training sets and produce high accuracy results.
2The number of users who rated the easiness in use 4 points out of 5
has increased from 60% to 85% because we have modified the interface by
writing the meaning of each medical test name wasn't clear to users.
Data Analysis:
According to the chart (fig.9), the percentage of the applicants surveyed
affected by the application is about 93.8%, which ensures that our website
is effecting positively on the Medical field.
Fig.9: Percentage of Applicants
surveyed affected by the application
Learning transfer:
BI.3.13: studying the structure and function of the components of the
circulatory and respiratory systems helped in writing the instructions the
patients must follow according to the relation between the systems of the
body.
CH.3.01: studying the scientific methodology improved the research
methods to find how to solve errors of the project; the uncertainty concept
increased the accuracy of the project.
MA.3.02: studying the local max and minimum and inflection points helps
us to analyse the graphs related to the function and the maximum and
minimum ranges of the parameters.
ES.3.02: Analyse data sets that provide evidence for plate motions,
including GPS and it helped us in seeing example of AI works and how
effectively it helps in our daily use.
ST.3.01: concept the sample distribution and central limit theorem helps in
creating data analysis graph shows people who are affected by the website
and rate it through the survey.
Conclusion:
To conclude, we choose to solve the challenge of public health issues with
the help of artificial intelligence. We focused on our solution on the
diagnosis of cardiovascular (CVD) heart disease, which is one of the most
dangerous diseases by creating a prototype which consists of an AI model
with a website interface. The website consists of 13 parameter indicators,
which are indicators of CVD and that the model takes the decision
according to each parameter's ranges. If the user has CVD, he will be given
some instructions according to its situation to avoid a heart attack. The
parameters are well-chosen according to their effects on the blood pressure,
cholesterol, fasting blood sugar and others that affect the heart. The
solution has achieved the design requirements which are high accuracy as
our accuracy is 89.06 ± 0.03%, which was calculated by AI code that is
suitable for calculating the accuracy, easy to use by any person as the
easiness of the interface. The user's parameters in their health tests and the
codes are AI, not traditional programming. Thankfully, the project shows
a great success comparing to other solutions which have low accuracy and
must be paid to show the results.
Recommendation:
We recommend to the future researchers who want to develop and work on
the project to:
1- Insert more data sets to the original one to train the AI model on
more data set, thus increasing efficiency, but it must be
authenticated.
2- Increase the parameters that the model decides according to them,
but they must be well chosen and must have a strong medical reason
to choose them.
3- The AI model interface can be changed into an application rather
than a website to be published to users and spread more between the
users.
4- The website may provide many languages (Arabic-FrenchSpanish…. etc.).
Literature cited:
Chellammal, S. & Sharmil, R. (2019). Recommendation of attributes for
heart disease prediction using correlation measure [PDF file].
International Journal of Recent Technology and Engineering
(IJRTE), Vol.8(253). 870-875. Retrieved November 15, 2020 from
https://www.ijrte.org/wpcontent/uploads/papers/v8i2S3/B11630782
S319.pdf
Hoyt, R. (2018). Heart disease prediction. Retrieved December 1, 2020
from https://data.world/informatics-edu/heart-disease-prediction
Mayo Clinic. (2021). Artificial intelligence (AI) in cardiovascular
medicine. Retrieved November 10, 2021 from
https://www.mayoclinic.org/departmentscenters/aicardiology/overvi
ew/ovc-20486648
Walker, J. & Halliday, D. & Resnick, R. (2011). Fundamental of physics
(10th ed.) [PDF file]. USA: Wiley & Sons. Retrieved January
5,2021.from
https://salmanisaleh.files.wordpress.com/2019/02/fundamentals-ofphysics-textbook.pdf
World Health Organization. (2017). Cardiovascular diseases (CVDs).
Retrieved October 30, 2020 from https://www.who.int/newsroom/fact-sheets/detail/cardiovascular-diseases-(cvds)
Acknowledgment:
At the end, we would like to thank everyone who had helped us in this
project. Special thanks to: Mrs. Eman Ali, our capstone leader Mrs Mervat
Hamed and our capstone teacher Mrs Walaa Mostafa for the great job that
they did with us.
For further information:
Please don't hesitate to contact us at:
•
Aya.1218108@stemalex.moe.edu.eg
•
rahma.1318147@stemalex.moe.edu.eg
•
Mariam.1218121@stemalex.moe.edu.eg
Download