International Journal of Engineering Applied Sciences and Technology, 2021

Vol. 5, Issue 9, ISSN No. 2455-2143, Pages 173-177

Published Online January 2021 in IJEAST (http://www.ijeast.com)

THYROID DETECTION USING MACHINE

LEARNING
Chandan R Chetan Vasan
Department of ECE Department of ECE
DSU, Bangalore, Karnataka, India DSU, Bangalore, Karnataka, India

Chethan MS Devikarani H S
Department of ECE Department of ECE
DSU, Bangalore, Karnataka, India DSU, Bangalore, Karnataka, India

Abstract—The Thyroid gland is a vascular gland classification in healthy patient as accurate as possible.
and one of the most important organs of a human If this is not done, misclassification can lead to a
body. This gland secretes two hormones which help healthy patient getting unnecessary treatment.
in controlling the metabolism of the body. The two
types of Thyroid disorders are Hyperthyroidism The Thyroid gland is an endocrine gland present in the
and Hypothyroidism. When this disorder occurs in human neck beneath the Adam’s apple which help in
the body, they release certain type of hormones into secretion of thyroid hormone that influence the rate of
the body which imbalances the body’s metabolism. metabolism and protein synthesis. The thyroid
Thyroid related Blood test is used to detect this hormones are useful in counting how briskly the heart
disease but it is often blurred and noise will be beats and how fast we burn calories. The thyroid
present. Data cleansing methods were used to make secretes two types of active hormones called
the data primitive enough for the analytics to show levothyroxine (T4) and triiodothyronine (T3). These
the risk of patients getting this disease. Machine hormones help in regulating the body temperature.
Learning plays a very deciding role in the disease These also aid in energy-bearing and transmission in
prediction. Machine Learning algorithms, SVM - every part of the body and decisive in protein
support vector machine, decision tree, logistic management. Iodine is considered as the main
regression, KNN - K-nearest neighbours, ANN- building block of the thyroid gland. It’s prostrated in
Artificial Neural Network are used to predict the few specific problems. Undersupply of these
patient’s risk of getting thyroid disease. Web app is hormones can lead to hyperthyroidism. There are
created to get data from users to predict the type of many originations related to hyperthyroidism and
disease. underactive thyroids. There are various kinds of
medications like thyroid surgery is liable to ionizing
Keywords— SVM, KNN, ANN, Logistic Regression, radiation, continual tenderness of the thyroid,
Flask. deficiency of iodine and lack of enzyme to make
thyroid hormones.
I. INTRODUCTION
II. DEFINITION OF PROBLEM
The evolvement computational biology is used in
healthcare industry. It allows collection of stored According to statistics, thyroid disorders are on the
patient data for the prediction of the disease. There are rise in India. Approximately 1 in 10 Indian adults
prediction algorithms which are available for the suffer from thyroid problem. It has been estimated that
diagnosis of the disease at early stages. The medical around 42 million peoples suffer from thyroid disease.
information systems are rich of datasets but there are Predicting thyroid disorder by doctor is a tedious
only few intelligent systems which can easily analysis process which might lead to negative prediction, only
the disease. Over a period of time, the machine experienced doctor can examine the case properly. To
learning algorithms have started playing a crucial role assist doctors machine learning can help them in
in resolving the complex and non-linear problems in diagnosis of disease and reduces their burden.
the developing model. In any disease prediction
models are used to override the features that can be
selected from different datasets which can be used in

173
 International Journal of Engineering Applied Sciences and Technology, 2021
Vol. 5, Issue 9, ISSN No. 2455-2143, Pages 173-177
Published Online January 2021 in IJEAST (http://www.ijeast.com)

A. Objective IV. SYSTEM ANALYSIS AND DESIGN

 The main objective is to develop a system which
can predict the type of thyroid disease that patient A. METHODOLOGY
is affected from.
 To predict thyroid disease with usage of minimum For predicting Thyroid disease analyzing blood report
number of parameters. is required to analyze and predict disease. Thyroid
 To predict all possible types of Thyroid diseases. blood test data set analysis will be conducted using
various supervised machine learning classifier
III. LITERATURE SURVEY techniques. Based on the accuracy of different
algorithm, best accuracy algorithm will be chosen to
fetch the result.
[1] Ankita Tyagi and Rikitha Mehra “Interactive
Thyroid Disease Prediction System Using Machine For first part, thyroid data set is taken from UCI
Learning Techniques” 5th IEEE International repository. The dataset of hyperthyroidism and
Conference on Parallel, Distributed and Grid hypothyroidism is used where hyper and hypo are the
Computing(PDGC-2018), 20-22 Dec, 2018, Solan, two labels. These data set need to be checked before
India. feeding it to training. There may be presence of null
data or unnecessary data, this should undergo data
In this work they use different classification cleaning to remove such data.
algorithms- Decision Tree, Support Vector Machine,
Artificial Neural Network, k-Nearer-Neighbor Cleaned data is used as training data and test data,
algorithm. Based on the data set obtained from UCI which is fed as input to the algorithm.
Repository, classification and prediction was The algorithm extracts the features from different
performed and accuracy was obtained based on output dataset to classify the data according to the labels. To
produced. They have analyzed accuracy of algorithms check the accuracy of the prediction, test data is fed to
used and comparison is made to find best technique the algorithm.
with high accuracy.
Based on the feature extracted, probability will be
Sunila Godara. [3] They have used Logistics generated for test data by comparing the features of
Regression and SVM machine learning Technique to both. Highest probability value will be classified to
analyze Thyroid Dataset. Comparison was made that particular label whether it is hyperthyroidism or
between these two algorithms based on Precision, hypothyroidism.
Recall, F measure, ROC, RMS error. Logistic
Regression turned out has best classifier. Web app is developed using python Flask (back end)
and web application is designed using HTML5 ( front
YongFeng Wang. [2] Thyroid Nodule is diagnosed for end) and CSS, where the chosen ML model will be
benign or malignant type using Ultrasound images of linked with web app and HTML. The users blood test
thyroid by image analysis - radiomics and deep data will be entered in web app front end and the back
learning based approaches. Comparison is made end will process the data using model and result will
between these two approaches. The classification be designed on screen.
accuracy, sensitivity, and specificity of applying
radiomics based method are 66.81%, 51.19% and
75.77%, respectively, while the evaluation indexes for
the deep learning based method trained to the testing
samples are 74.69%, 63.10% and 80.20%,
respectively. Deep learning turned out as best
approaches.
Hitesh Garg. [4] Feed Forward Neural Network is used
for feature extraction and segmentation from
Ultrasound images to predict the tumors. The accuracy
and other factors were measured and all the average
values were above 86%.

174
 International Journal of Engineering Applied Sciences and Technology, 2021
Vol. 5, Issue 9, ISSN No. 2455-2143, Pages 173-177
Published Online January 2021 in IJEAST (http://www.ijeast.com)

V. EVALUATION AND TESTING

A. Result
 Heat map

A heat map (or heatmap) is a data visualization

technique that shows magnitude of a phenomenon as
color in two dimensions. The variation in color may be
by hue or intensity, giving obvious visual cues to the
reader about how the phenomenon is clustered or
varies over space. Correlation between parameters of
our dataset is interpreted and pictorial view is obtained
as shown in fig 3.

Fig 1: work flowchart

B. Data set
 A new_thyroid dataset is accessible in UCI
machine learning repository and this set has
215 samples and 5 features.
 5 lab tests are used to try to predict whether a Fig 3: Heat Map
patient’s thyroid to the class is normal,
KNN algorithm has been implemented and 93.84%
hypothyroidism or hyperthyroidism. The
diagnosis (the class level) was based on a accuracy score is obtained as shown in fig 4.
complete medical record.
 This dataset has 150 instance of normal class,
35 instance of hyperthyroidism class and 35
instance of hypothyroidism class.
 Class attribute, T3 resin uptake test, total T4
, total T3 , TSH , difference of TSH value
after injection of 200 micrograms of
thyrotropin releasing hormone (as shown in
fig 2). Fig 4: KNN algorithm accuracy

SVM algorithm has been implemented and 95.38%

accuracy score is obtained as shown in fig 5.

Fig 5: SVM algorithm accuracy

Fig 2: dataset header
ANN algorithm has been implemented and 75.38%
accuracy score is obtained as shown in fig 6.

175
 International Journal of Engineering Applied Sciences and Technology, 2021
Vol. 5, Issue 9, ISSN No. 2455-2143, Pages 173-177
Published Online January 2021 in IJEAST (http://www.ijeast.com)

Fig 6: ANN algorithm accuracy

Decision Tree algorithm has been implemented and

92.3% accuracy score is obtained as shown in fig 7.
Fig 9: bar graph of score

Since accuracy obtained from logistic regression

model (96.92%) was highest, this model will be
considered for our prediction model. To save this
model for interfacing with web app, joblib library is
used and file “thyroid_prediction_lr.pkl” will be
created.
Fig 7: Decision Tree algorithm accuracy Web app is created to interface user and the trained
model. Python Flask coding is used to create web app
Logistic Regression has been implemented and
and html is used to design web page.
96.92% accuracy score is obtained as shown in fig 8.
Fig 10 shows the data entered into the parameter box
of web app created, which will be used at back end by
model to predict thyroid disease.

Fig 8: logistic regression algorithm accuracy

Fig 10: result example

For the data given through the web page as shown in

fig 10, after processing the data by the model, result
will be displayed in the same web page as shown in fig
11.

176
 International Journal of Engineering Applied Sciences and Technology, 2021
Vol. 5, Issue 9, ISSN No. 2455-2143, Pages 173-177
Published Online January 2021 in IJEAST (http://www.ijeast.com)

[5] L. Ozyılmaz and T. Yıldırım,(2002). “Diagnosis of

thyroid disease using artificial neural network
methods,” in: Proceedings of ICONIP’02 9th
international conference on neural information
processing (Singapore: Orchid Country Club, pp.
2033–2036).
[6] K. Polat, S. Sahan and S. Gunes,(2007) “A novel
hybrid method based on artificial immune recognition
system (AIRS) with fuzzy weighted preprocessing for
thyroid disease diagnosis,” Expert Systems with
Applications,(vol. 32, pp. 1141-1147).
Fig 11: example result2 [7] F. Saiti, A. A. Naini, M. A. Shoorehdeli, and M.
Teshnehlab,(2009) “Thyroid Disease Diagnosis Based
VI. CONCLUSION on Genetic Algorithms Using PNN and SVM,” in 3rd
Thyroid Detection using Machine Learning is a project International Conference on Bioinformatics and
idea that aims a smart and precise way to predict BiomedicalEngineering. ICBBE 2009.
thyroid disease. We have made use of logistic
regression algorithm to train our dataset and to predict [8] G. Zhang, L.V. Berardi,(2007) “An investigation
thyroid disease with more accuracy. Here the machine of neural networks in thyroid function diagnosis,”
is trained to detect whether the person normal, hyper- Health Care Management Science,1998, (pp. 29-37.)
hypothyroidism based on the user’s input. So when
user enters data in web app the data will be processed [9] V. Vapnik,(2012).Estimation of Dependences
in backend (model) and the result will be displayed on Based on Empirical Data, Springer, New York.
the screen. Our objective was to give society an [10] Obermeyer Z,(2016). Emanuel EJ. Predicting the
efficient and precise way of machine learning which future— big data, machine learning, and clinical
can be used in applications aiming to perform disease medicine. N Engl ; (375:12161219).
detection.
[11] Breiman L.(2001) StatisticalModeling: the two
Further development can be do by using image cultures.Stat Sci. ;16:199-231..
processing of ultrasonic scanning of thyroid images to
predict thyroid nodules and cancer, which cannot be [12] Ehrenstein V, Nielsen H, Pedersen AB, Johnsen
recognized in blood test report. SP, Pedersen L. (2017) Clinical epidemiology in the
era of big data: new opportunities, familiar challenges.
By combining both the results, thyroid disease Clin Epidemiol. ; 9:245-250
prediction can cover all thyroid related diseases.
[13] S. Godara and R. Singh,(2016) "Evaluation of
VII. REFERENCE Predictive Machine Learning Techniques as Expert
Systems in Medical Diagnosis", Indian Journal of
[1]Ankita Tyagi and Ritika Mehra. (2018).“Interactive Science and Technology, (Vol. 910).
Thyroid Disease Prediction System using Machine
Learning Techniques” published on ResearchGate. [14] Sunila, Rishipal Singh and Sanjeev Kumar.(2016)
"A Novel Weighted Class based Clustering for
[2] YongFeng Wang,(2020). “Comparison Study of Medical Diagnostic Interface." Indian Journal of
Radiomics and Deep-Learning Based Methods for Science and Technology (Vol 9).
Thyroid Nodules Classification using Ultrasound
Images” published on IEEEAccess.  https://archive.ics.uci.edu/ml/datasets/Thyroid+
Disease
[3] Sunila Godara,(2018). “Prediction of Thyroid  https://www.tutorialspoint.com/python_data_scie
Disease Using Machine Learning Techniques” nce/index.html
published on IJEE.  https://www.tutorialspoint.com/flask/index.htm
[4] Hitesh Garg,(2013). “Segmentation of Thyroid
Gland in Ultrasound image using Neural Network”
published on IEEE.

177