(IJACSA) International Journal of Advanced Computer Science and Applications,

Vol. 13, No. 5, 2022

A Model for Classification and Diagnosis of Skin

Disease using Machine Learning and Image
Processing Techniques
Shaden Abdulaziz AlDera1 Mohamed Tahar Ben Othman2
Department of Computer Science BIND Research Group, IEEE Senior Member
College of Computer Department of Computer Science
Qassim University, Buraydah 51452 College of Computer, Qassim University
Saudi Arabia Buraydah 51452, Saudi Arabia

Abstract—Skin diseases are a global health problem that is techniques can help achieve high accuracy in skin diagnosing
difficult to diagnose sometimes due to the disease’s complexity, at the initial stage. Images processing plays an effective role in
and the time-consuming effort. In addition to the fact that skin diagnosis the skin diseases with the help of libraries such as
diseases affect human health, it also affects the psycho-social life OpenCV, Scikit-Image, and NumPy. Afterward, machine
if not diagnosed and controlled early. The enhancement of learning algorithms such as SVM, RF, and K-NN are used for
images processing techniques and machine learning leads to an the classification task. Combining these techniques will save
effective and fast diagnosis that help detect the skin disease early. time and reach a quicker and more trusted diagnostic than
This paper presents a model that takes an image of the skin typical procedures like patch tests and biopsy [4]. Due to the
affected by a disease and diagnose acne, cherry angioma,
limitation of the existing models that diagnose different skin
melanoma, and psoriasis. The proposed model is composed of
diseases, this proposed model studied four skin diseases. This
five steps, i.e., image acquisition, preprocessing, segmentation,
feature extraction, and classification. In addition to using the research work aims to build a model that provides an easy, fast
machine learning algorithms for evaluating the model, i.e., and efficient solution for skin disease diagnosis, i.e., acne,
Support Vector Machine (SVM), Random Forest (RF), and K- cherry angioma, melanoma, and psoriasis, using image
Nearest Neighbor (K-NN) classifiers, and achieved 90.7%, processing and machine learning techniques.
84.2%, and 67.1%, respectively. Also, the SVM classifier result of The building of the proposed model passes through multi-
the proposed model was compared with other papers, and mostly
steps of image processing including, acquiring images,
the proposed model’s result is better. In contrast, one paper
preprocessing images, i.e., resizing images, color
achieved an accuracy of 100%.
transformation, de-noising, and normalization, segmentation,
Keywords—Skin disease; image processing; classification; and feature extraction. In the end, train the model with
machine learning; diagnosis; SVM; RF; K-NN; acne; cherry traditional machine learning algorithms, e.g., SVM, RF, and K-
angioma; melanoma; psoriasis NN. Several papers conducted on this paper's topic had been
focused on the use of machine learning and image processing
I. INTRODUCTION to classify skin cancer. Thus, this paper proposed a model to
Nowadays, imaging is used in medical science extensively, diagnose other common diseases in addition to skin cancer.
so before any surgery or treatment decision a preliminary Furthermore, the proposed model tested cherry angioma
knowledge can be determined, and diagnosis can be done. For disease, which is very rarely tested in the previous research.
this, imaging in medicine has become a tool to start most of the This paper is organized into the following sections:
disease treatment cycle, starting from detection passing Section 2 reviews the previous work, Section 3 presents the
through evaluation, and ending with the treatment decision. methodology, Section 4 shows the obtained results, and
Skin disease is one of these medical areas where images play a Section 5 is the conclusion of this work.
role in detecting, diagnosing, and treating the disease [1]. In
recent years, skin diseases have increased and begun to be a II. PREVIOUS WORK
global health problem [2]. Those who suffer from skin diseases Many researchers have proposed a model that combines
without disease diagnosis may diminish their life quality and image processing and machine learning algorithms techniques
have a negatively psycho-social impact [3]. to classify and diagnose several skin diseases.
In fact, skin diseases are difficult to diagnose due to the Hameed, Shabut, and Hossain [5] implemented a system
complexity of human skin. Also, the lack of expertise may lead that classifies healthy, acne, eczema, psoriasis, benign, and
to misdiagnosis or overdue diagnoses. Diagnosis of skin melanoma (malignant) skin diseases. The system was built
diseases at the health center may take a long time and require based on image processing techniques. To enhance images, the
domain expertise, which causes physical and financial costs. authors used an algorithm called Dull Razor to remove hair
On the other side, machine learning and image processing from skin images and then applied the Gaussian filter to

252 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 13, No. 5, 2022

smooth it. After that, in the segmentation task, they firstly increase accuracy by more than 90%. Authors in [13] propose a
discarded any non-skin area, then applied Otsu’s thresholding web system to diagnose skin diseases in Ghana (Africa). Their
on the area of skin in order to segment the disease lesion. study includes a CNN to classify 254 images of diseases, i.e.,
Moreover, they used color spaces like red, green, and blue atopic dermatitis, acne vulgaris, and scabies. In the end, they
(RGB), hue, saturation, and value (HSV), luma, blue, and red reached an accuracy for each disease of 88%, 85%, and 84.7%,
(YCbCr) which means separate brightness i.e. luma from color, respectively. The proposed system just takes 0.0001 seconds to
and grayscale to extract color features. Also, they used diagnose, which may expedite more patients’ diagnoses than a
neighborhood gray-tone difference matrix (NGTDM) and Gray diagnosis in the clinic.
Level Co-occurrence Matrix (GLCM) techniques to extract the
texture features. In the end, they classified the diseases using Most of the previous studies prove the efficacy of using
SVM and obtained an accuracy of 83%. SVM and CNN. Furthermore, the studies evinced the image
processing plays a key role in helping to classify various skin
Additionally, Sinthura S. et al. [6] propose a method to diseases. Moreover, increasing the number of images may
detect skin diseases. Their method indicates using the adaptive positively affect the classification due to the increased training
filter to remove the noises, then converting it to grayscale model.
color. Besides, they used Otsu’s thresholding technique to
segment the disease lesion. Furthermore, they used GLCM to III. METHODOLOGY
extract the texture features. Finally, to validate their proposal, This paper demonstrates the classification of several types
they used the SVM classifier and achieved an accuracy of 89%. of skin disease to diagnose the lesions such as acne, cherry
In image classification-based color, the researchers in [7] train angioma, melanoma, and psoriasis. Accordingly, the processes
a model for detecting and classifying various skin diseases involved in identifying these skin diseases are preprocessing,
using the K-NN classifier. They use color models to extract segmentation, feature extraction, and classification. The
features, including the HSV and the lightness, red/green, and following points show the datasets used and explain the
blue/yellow (L*a*b) color models. Their results showed that proposed and techniques of this work.
the HSV color model is more efficient with 91.80% accuracy
than the L*a*b color model with 81.60% accuracy. Moreover, A. Dataset
Ahmed, Ema, and Islam [8] propose a new automated system Due to the privacy of medical records, collecting images is
using the Transductive SVM (TSVM) to classify 24 types of a challenging task. Therefore, the images gathered from
skin disease. The proposed system uses a hybrid genetic available resources: the dermnet NZ [14] and atlas
algorithm to segment the image. Also, they used ant colony dermatologico [15]. In this work, the dataset consists of 377
optimization (ACO-GA) and GLCM to extract its features. images of four different disease classes: acne, cherry angioma,
Their work achieved 95% accuracy. melanoma, and psoriasis. Fig. 1 shows a sample of each class.
Table I lists the number of images of each class.
A method was carried out to apply pre-trained
Convolutional Neural Networks (CNN) to extract features for 1) Diseases definition: In the following, a brief definition
skin diseases. The paper by ALenezi [9] proposes a system of each disease studied in this work, as mentioned in the
using the pre-trained CNN AlexNet to extract skin disease dermnet NZ website [14].
features and SVM to classify the diseases. The system was
a) Acne: It is a common chronic disorder, often
built with a dataset of 80 images of these skin types;
melanoma, psoriasis, eczema, and healthy skin. Her system confined to the face, but it may happen in the chest, back, and
was tested on 20 images and achieved an accuracy of 100%. neck. Acne may occur in children and adults of all ages.
The same method used by researchers [10] present an However, acne is caused due to a combination of several
intelligent expert system for classifying 9144 skin lesions, i.e., factors such as familial tendency, acne bacteria, and
acne, eczema, benign or malignant (melanoma), and healthy hormones. Acne could be characterized as blackheads and
skin images. For extracting the lesion’s features, they used a whiteheads.
pre-trained CNN model AlexNet. Their system result achieved b) Cherry angioma: The reason of cherry angioma is
an accuracy of 86.21% by using the SVM classifier, where unknown. It is very prevalent in males and females of any age,
divided dataset in the ratio 70:30 for train and test set. while it markedly increases in people from about the age of
Another study by Hajgude et al. [11] proposes a solution to 40. However, cherry angioma may be in red or purple, or blue
detect 408 images of eczema, impetigo, melanoma skin color. Also, it could be scattered overall body surface parts.
diseases, and a class named other images. They build the c) Melanoma: It is the gravest form of skin cancer. It
model using these techniques: median filter to remove the happened due to the uncontrolled growth of melanocytes
noises, the Otsu method to segment lesions, 2D Wavelet (pigment cells). Melanoma may occur at any age but is very
transform to extract features like entropy and standard rare in children. However, the features of melanoma could be
deviation, and GLCM to extract texture features like contrast having several colors like blue, brown, red, etc.
and correlation. They used SVM and CNN classifiers to d) Psoriasis: It is a chronic inflammation of a skin
classify the diseases and obtained an accuracy of 90.7% and
condition. It affects males and females at any age. It is
99.1%, respectively. Authors in [12] describe the CNN
characterized by symmetrically distributed, red color, scaly
classifier and the major libraries for image processing. Then,
they use CNN to classify and diagnose skin disease with an plaques with well-defined edges.
accuracy of 70%, even though using a large dataset may

253 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 13, No. 5, 2022

After that, due to the noises in skin images, a de-noising

technique called a median filter was applied. The median filter
is the most filter the researcher used according to the
advantage of preserving the edges of the image [17]. Further,
the color images converted to a grayscale color model for
segmentation and feature extraction tasks. Finally, the images
pixels’ values normalized between 1 and 0. In Fig. 3, column
Fig. 1. Dataset Samples A shows the original images with a size of 250X250. Next,
column B shows the images after applying the median filter.
TABLE I. DATASET DISTRIBUTION Also, column C shows images converted to grayscale.
Disease Dataset Total b) Segmentation: The segmentation task is a process of
Acne Dermnet NZ dataset, Atlas dermatologico 80 partitioning the lesion region from the skin. This segment
gains based on similarity or difference of pixels properties like
Cherry angioma Dermnet NZ dataset 37
color, sharpness, brightness, or intensity of an image [18][19].
Melanoma Dermnet NZ dataset, Atlas dermatologico 80 Based on this work, it’s a challenging task due to the several
Psoriasis Dermnet NZ dataset, Atlas dermatologico 180 diseases the proposed test. Also, the key problem is the entire
Total 377
lesion’s color may be similar, and the lesion’s boundaries may
be fuzzy, besides the complexity of the skin itself. To address
B. Proposed Methodology this problem, the Otsu’s thresholding is used to create a mask
This section presents the processes of this proposed model (binary image), then applied it to the grayscale image. Some
and the techniques used. The architecture of this model is of the results are shown in Fig. 3. Column D shows the binary
shown in Fig. 2. The procedure of the proposed model is masks, and column E shows the final segmentation results.
described in the following points:  Otsu’s: Otsu’s is the most popular threshold
segmentation technique, and it is applied to a grayscale
 Import the train set images and process it through: image. Unlike the manual threshold, it automatically
preprocessing the images, then segmenting the lesion compares the minimize weighted within gray classes
from the remaining normal skin, and after that variance to find the optimal threshold value. Since the
extracting its features. threshold value was determined, the lesion can be
 Import the test set images and process them in the same segmented from the normal skin region [20][21].
way as step 1.
 The features that extract from the train set images are
stored in a knowledge base.
 Compare the features that extract from the test set
image with the feature stored in the knowledge base.
 Classify images using machine learning algorithms i.e.,
SVM, RF, K-NN.
 Diagnosis of the disease.
 When the user uploads an image, it is will pass through
the same processing of the test set image.
1) Image processing: Image processing is a technique that
manipulates and analyzes an image received from a camera or
sensors. Therefore, the image processing’s main objective is to
enhance the image’s quality and extract its information in
order to be more interpretable by a human or machine
perception. Nowadays, many image processing techniques are
incorporated as they turn out to be strong computational
methodologies and strong potential to be effective in
healthcare and all fields [16]. In the following, describe the
image processing techniques used in this work.
a) Preprocessing: The first step in processing the skin
disease images is preprocessing in order to enhance images. In
this work, firstly, all the images were resized to 250X250. Fig. 2. Model Architecture.

254 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 13, No. 5, 2022

features. All the feature extraction techniques' parameters used

in this proposed model are listed in Table II.
 Gabor: Gabor filter is a linear filter used to extract
texture features. It is the most commonly used in image
processing and image texture analysis [23]. Essentially,
Gabor is a band-pass that extracts patterns in a signal at
specific frequencies [24].
 Entropy: It measures the expectation of the quantity of
information in the grayscale image. It calculates the
center pixel of all the neighboring pixels in the kernel’s
window [25].
 Sobel: Sobel is a filter used to detect the image’s edge
features. It has the power on processing with more
minor time consumption, less in loss of edge, and
strongly resists the noise [26]. In this work, the Sobel is
Fig. 3. Samples Results of Images Processing Techniques.
applied that utilized two kernels to obtain the
horizontal(Gx) and vertical(Gy) approximation of the
c) Feature Extraction: Feature extraction is a technique
derivation of the grayscale skin diseases images. These
that plays a significant role in image processing. It is used the kernels were convolved to overlay the image’s pixels.
image or the segmented lesion to extract the characteristics’
features that represent the information of that image for 2) Classification: The classification task classifies data
classification tasks. However, texture feature is a type of into distinct known classes using machine learning algorithms
feature that can be used in recognition of an image by to predict the disease. Once features are extracted, it is given
describing the visual image’s surface. Textures are complex as input to the classifier model. When the model is accurate, it
patterns composed of many characteristics, including size, is used to classify new images that are a member of the trained
color, brightness, slope, etc. [22]. In this work, feature disease classes. The proposed model used the traditional well-
extraction is performed using Gabor and Entropy techniques Known classifiers to conduct the experiments, i.e., SVM, RF,
to extract texture features and the Sobel technique for edge and K-NN.
TABLE II. THE PARAMETERS OF FEATURE EXTRACTION TECHNIQUES

Feature Extraction
Formula Parameters Value
Technique

( ) λ: wavelength of the cosine multiplier.

Pi/ 4 = 3.14/4
( ) θ: Frequency of alternations in degrees.
( ) ( ), where 4 * Pi = 4 * 3.14
ϕ: Phase offset of the sinusoidal function.
Gabor = x cos + y sin 0
: Sigma / standard deviation of the Gaussian envelope.
1 and 3
= - x sin + y cos γ: Spatial aspect ratio and specifies the ellipticity of the support
0.5
of Gabor function.

( ) ∑ ( ) ( ), where Logb=
Entropy Kernel 3X3
2

Sobel √ Two Kernels: Gx and Gy 3X3 and 3X3

Accuracy = (TP + TN) / (TP+FP+TN+FN) (1)

IV. EXPERIMENTAL RESULTS
This section shows the evaluated experiments to measure Precision = TP / (TP+FP) (2)
the performance of this model. The proposed model's Recall = TP / (TP+FN) (3)
experiments were implemented using the python language in
Spyder environment from anaconda, along with several F1-score = 2 * (precision * recall / precision + recall) (4)
libraries, i.e., Scikit-learn to perform machine learning Where TP is true positive, TN is a true negative, FP is false
algorithms, OpenCV, Glob, and Os to perform image positive, and FN is a false negative.
processing, Skimage to perform filters, and Matplot for the
visualization. B. Splitting Dataset and Classifiers Parameters Results
A. Evaluation Measure To detect the behavior of machine learning algorithms, it
needs to test the model on data that is not used in the training
The model performance was analysis with several process. Toward this, the dataset splits into two sets: training
measures. These measures are formulated as in (1), (2), (3), set and testing set. Table III lists three experiments of splitting
and (4): and lists the accuracy results of each classifier.

255 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 13, No. 5, 2022

TABLE III. THE SPLITTING DATASET WITH TRAIN AND TEST RESULTS

Splitting #images Classifiers Parameters Training Testing

SVM 92.3% 78%
Kernal = 'rbf'
Train: 70%
Train: 263 RF n_estimators=50 100% 73.6%
Test: 30%
n_neighbors=5
Test: 114 K-NN 73.7% 67.5%
SVM Kernal = 'rbf' 91.4% 77.8%
Train: 75% Train: 282
RF n_estimators=50 100% 74.7%
Test: 25% Test: 95
K-NN n_neighbors=5 72.6 68٪
SVM 91% 90.7%
Kernal = 'rbf'
Train: 80%
Train: 301 RF n_estimators=50 100% 84.2%
Test: 20%
n_neighbors=5
Test: 76 K-NN 73% 67.1%

TABLE IV. CLASSIFIERS PARAMETERS AND THEIR RESULTS

Classifiers Parameters Values Accuracy

'Linear': 81.5%
Kernel 'linear', 'rbf', 'poly', 'sigmoid'
'rbf': 90.7%
SVM Regularization(C) 1.0
'poly': 71%'
Degree 3
'sigmoid': 63%
30: 73.6%
40:80%
RF n_estimators 30,40,50,60
50: 84.2%
60: 82.8%
3: 64%
K-NN n_neighbors 3,4,5 4: 63%
5: 67.1%
D. Classification Experiment
Each ML classifier has several possible values of each
corresponding parameter that value may affect on accuracy The proposed model was validated using SVM, RF, and K-
result. However, the value that obtains the highest accuracy NN classifiers with the evaluation measures, i.e., accuracy,
result will be used in the proposed model. The possible precision, recall, and f1-score. Table VI lists all measurement
parameters and their values are illustrated with the accuracy results of each classifier. It is observed that the SVM has
results in Table IV. All these experiments were tested on the superiority in classifying skin diseases over others in accuracy,
dataset that split into 80% for training and 20% for testing. precision, recall, and f1-score of 90.7%, 91%, 90.8%, and
90.8%, respectively. At the same time, the K-NN has obtained
C. Classifiers Performance Matrix the worst results.
This section shows the confusion matrix of each classifier.
E. Comparison Results
It is represented in a table style to describe the model’s
classifier performance. However, it presents the prediction To our knowledge, there are no study experiments on the
results on the test set data. The matrix size is based on the same diseases of this work. Also, due to different and
number of classes; in this case, the matrix size is 4X4. The unavailable datasets, the proposed model was compared with
rows indicate the true class, while the columns indicate the the research that tested some of the diseases studied in this
prediction for each class. While The diagonal of the matrix work. Table VII details the comparison research [5], [6], [9],
points to the number of images that are correctly classified. and [10] toward their studied diseases and the image
Fig. 4 displays the confusion matrices of the classifiers of processing techniques they used.
SVM, RF, and K-NN as shown in A, B, and C, respectively. Basically, many techniques are available for image
For example, in SVM, a total of 16 acne images, it classified processing and classification tasks, and it used in several skin
14 images correctly and 2 images were misclassified as diseases classification research. Among that, the comparison
psoriasis. research in Table VII are probably the most papers close to the
From these confusion matrices, could measure the techniques used and to the diseases tested in this work.
accuracy, precision, recall, and F1-score of each disease with In the comparison papers, they present in two ways of
each classifier as shown in Table V. Among the three image processing: manual and automatic. In manual image
classifiers, SVM is superior performance on the accuracy rate processing, researchers [5], [6] follow the same processing
of cherry angioma, melanoma, and psoriasis. While acne steps with different preprocessing filters techniques; Dull
obtained the highest accuracy rate using the RF classifier. In Razor with Gaussian in [5], and Median in [6]. Further, they
contrast, K-NN has the worst results accuracy in all these used the same segmentation technique, i.e., Otsu’s. Another
diseases. essential point, studies have shown that extraction techniques
play a key role in extracting the appropriate features for the

256 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 13, No. 5, 2022

classification task. For that, [5] extract the texture features 83%. However, it was observed the proposed model is higher
using GLCM and NGTDM, and extract color features using than [5], [6], and [10] with 90.7% accuracy.
color spaces. Similarly, researchers [6] extract texture features
using GLCM. On the other hand, both researchers [9] and [10] Thus, probably the proposed model's performance is better
utilized the automatic image processing by CNN with a pre- since this proposed model used multi techniques to extract a
train AlexNet. combination of features, i.e., Gabor and Entropy for texture
features and Sobel for edge features. In contrast, the paper [9]
Therefore, Table VIII lists the SVM classifier accuracy of is superior to the proposed model with 100% accuracy. Despite
this proposed model with the comparison research model’s the fact, that they trained the model on 80 images, while the
accuracy that used the same classifier, i.e., SVM, and different proposed model was trained on 301 images. Also, the proposed
image processing methods. All these papers showed a model tested cherry angioma disease, whereas the other studies
promising high accuracy rate in diagnosing diseases above did not test it.

Fig. 4. The Confusion Matrices OF SVM, RF, AND K-NN Classifiers.

TABLE V. DISEASE CLASSIFICATION RATE

Accuracy Precision Recall F1-score

Acne 82% 82% 88% 85%
Cherry Angioma 100% 100% 88% 93%
SVM
Melanoma 94% 94% 94% 94%
Psoriasis 92% 92% 92% 92%
Acne 92% 92% 69% 79%
Cherry Angioma 83% 83% 62% 71%
RF
Melanoma 79% 79% 94% 86%
Psoriasis 85% 85% 92% 88%
Acne 58% 58% 75% 60%
Cherry Angioma 38% 38% 75% 52%
K-NN
Melanoma 82% 82% 56% 67%
Psoriasis 83% 83% 61% 71%

TABLE VI. CLASSIFIERS RESULTS

Classifier Accuracy Precision Recall F1-Score

SVM 90.7% 91% 90.8% 90.8%
RF 84.2% 84.8% 84.2% 83.8%
K-NN 67.1% 72.8% 67.1% 68.4%

TABLE VII. THE COMPARISON RESEARCH

Ref. Disease Segmentation technique Feature techniques

Hameed N et al. [5] Acne, Psoriasis, Melanoma, Benign, Eczema, and Healthy Skin. Otsu’s GLCM, NGTDM, and color spaces
Sinthura S. et al. [6] Acne, Psoriasis, Melanoma, and Rosacea. Otsu’s GLCM
ALEnezi N. [9] Psoriasis, Melanoma, Eczema, and Healthy Skin. - CNN: AlexNet.
Hameed N et al. [10] Acne, Melanoma, Benign, Eczema, and Healthy Skin. - CNN: AlexNet.
Proposed Model Acne, Cherry angioma, Melanoma, and Psoriasis. Otsu’s Entropy, Gabor, and Sobel.

257 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 13, No. 5, 2022

psoriasis,‖ Acta Derm. Venereol., vol. 97, no. 1, pp. 91–97, 2017, doi:
TABLE VIII. COMPARISON OF ACCURACY 10.2340/00015555-2483.
Ref. Accuracy [4] V. Pugazhenthi et al., ―Skin Disease Detection And Classification,‖ Int. J.
Adv. Eng. Res. Sci., vol. 6, no. 5, pp. 396–400, 2019.
Hameed N et al. [5] 83% [5] N. Hameed, A. Shabut, and M. A. Hossain, ―A Computer-Aided
Sinthura S. et al. [6] 89% diagnosis system for classifying prominent skin lesions using machine
learning,‖ 2018 10th Comput. Sci. Electron. Eng. Conf. CEEC 2018 -
ALEnezi N. [9] 100%. Proc., pp. 186–191, 2019, doi: 10.1109/CEEC.2018.8674183.
Hameed N et al. [10] 86.21% [6] S. S. Sinthura, K. R. Sharon, G. Bhavani, L. Mounika, and B. Joshika,
―Advanced Skin Diseases Diagnosis Leveraging Image Processing,‖
Proposed Model 90.7% Proc. Int. Conf. Electron. Sustain. Commun. Syst. ICESC 2020, no. Icesc,
pp. 440–444, 2020, doi: 10.1109/ICESC48915.2020.9155914.
The main problem encountered in developing the proposed [7] A.V.Ubale and P. L. Paikrao, ―Detection and Classification of Skin
model was the few of availability of the image of the diseases Diseases using Different Color Phase Models,‖ Int. Res. J. Eng. Technol.
tested. In addition, the few papers that study various types of e-ISSN, vol. 06, no. 07, pp. 3658–3663, 2019.
skin diseases. May training more images increase the accuracy [8] M. H. Ahmed, R. R. Ema, and T. Islam, ―An Automated Dermatological
and make the model more accurate to diagnose new images, as Images Segmentation Based on a New Hybrid Intelligent ACO-GA
Algorithm and Diseases Identification Using TSVM Classifier,‖ 1st Int.
well as selecting appropriate techniques to extract useful Conf. Adv. Sci. Eng. Robot. Technol. 2019, ICASERT 2019, vol. 2019,
features. no. Icasert, 2019, doi: 10.1109/ICASERT.2019.8934560.
[9] N. S. Alkolifi Alenezi, ―A Method of Skin Disease Detection Using
V. CONCLUSION Image Processing and Machine Learning,‖ in Procedia Computer
This paper proposed a model that provides the Science, 2019, vol. 163, pp. 85–92, doi: 10.1016/j.procs.2019.12.090.
classification of different types of skin diseases: acne, cherry [10] N. Hameed, A. M. Shabut, and M. A. Hossain, ―Multi-Class Skin
Diseases Classification Using Deep Convolutional Neural Network and
angioma, melanoma, and psoriasis. According to the previous Support Vector Machine,‖ Int. Conf. Software, Knowl. Information, Ind.
works in this area, there is a scarcity in the studies on these Manag. Appl. Ski., vol. 2018-Decem, 2019, doi:
diseases as most research focuses on skin cancer. This model 10.1109/SKIMA.2018.8631525.
was conducted through image processing techniques and [11] M. J. Hajgude, A. Bhavsar, H. Achara, and N. Khubchandani, ―Skin
machine learning algorithms on a total of 377 images. The disease detection using Image Processing with data mining and deep
dataset is divided into 301 images for the train set and 76 learning,‖ IRJET, vol. 6, no. 4, pp. 4363–4366, 2019.
images for the test set. Firstly, the processing techniques are [12] J. Rathod, V. Wazhmode, A. Sodha, and P. Bhavathankar, ―Diagnosis of
applied to images in several steps: preprocessing including skin diseases using Convolutional Neural Networks,‖ Proc. 2nd Int. Conf.
Electron. Commun. Aerosp. Technol. ICECA 2018, no. Iceca, pp. 1048–
resizing images, removing the noise using the median filter and 1051, 2018, doi: 10.1109/ICECA.2018.8474593.
converting the image to grayscale, then separating the infected [13] S. Akyeramfo-Sam, A. Addo Philip, D. Yeboah, N. C. Nartey, and I. Kofi
area using Otsu’s, and extracting its features using Gabor, Nti, ―A Web-Based Skin Disease Diagnosis Using Convolutional Neural
Entropy, and Sobel. Secondly, the model was evaluated using Networks,‖ Int. J. Inf. Technol. Comput. Sci., vol. 11, no. 11, pp. 54–60,
SVM, KNN, and RF classifiers in terms of accuracy, precision, 2019, doi: 10.5815/ijitcs.2019.11.06.
recall, and f1-score. However, the proposed model results show [14] ―DermNet NZ.‖ website: dermnetnz.org , license: https://creativeco
that the SVM accomplished higher accuracy with 90.7% than mmons.org/licenses/by-nc-nd/3.0/nz/legalcode (accessed Jun. 01, 2021).
RF and K-NN. At the same time, RF and K-NN achieved [15] S. F. da Silva and D. B. Calheiros., ―Dermatology Atlas.‖
atlasdermatologico.com.br (accessed Jun. 01, 2021).
84.2% and 67.1% accuracy, respectively. The result of the
[16] S. Perumal and T. Velmurugan, ―Preprocessing by Contrast Enhancement
proposed model using the SVM classifier achieved better Techniques for Medical Images,‖ International Journal of Pure and
accuracy than the comparison research’ accuracy. In contrast, Applied Mathematics, vol. 118, no. 18. pp. 3681–3688, 2018.
one paper outperformed the proposed model's accuracy. Since [17] L. Kapoor and S. Thakur, ―A Survey on Brain Tumor Detection Using
it was not possible to collect a skin disease dataset locally, Image Processing Techniques,‖ 2017 7th Int. Conf. Cloud Comput. Data
finding a public source with multi images of diseases was the Sci. Eng. - Conflu., pp. 582–585, 2017, doi: 10.1109/CONFLUE
biggest challenge during this work. This model can accomplish NCE.2017.7943218.
higher accuracy by using more dataset images. Moreover, [18] M. M. Lone and S. Hussain, ―A Survey on Image Segmentation
programmers can utilize the model to develop a smartphone Techniques,‖ Int. Res. J. Eng. Technol., vol. 05, 2018, doi: 10.1007/978-
3-030-32150-5_112.
application to diagnose these skin diseases easily and early.
[19] H. Seo et al., ―Machine learning techniques for biomedical image
REFERENCES segmentation: An overview of technical aspects and introduction to state‐
[1] L. Bajaj, H. Kumar, and Y. Hasija, ―Automated System for Prediction of of‐art applications,‖ arXiv, pp. 1–35, 2019.
Skin Disease using Image Processing and Machine Learning,‖ Int. J. [20] H. Sen and A. Agarwal, ―A comparative analysis of entropy based
Comput. Appl., vol. 180, no. 19, pp. 9–12, 2018, doi: segmentation with Otsu method for gray and color images,‖ Proc. Int.
10.5120/ijca2018916428. Conf. Electron. Commun. Aerosp. Technol. ICECA 2017, vol. 2017-
[2] T. Vos et al., ―Global, regional, and national incidence, prevalence, and Janua, pp. 113–118, 2017, doi: 10.1109/ICECA.2017.8203655.
years lived with disability for 328 diseases and injuries for 195 countries, [21] Y. Wang, ―Improved OTSU and adaptive genetic algorithm for infrared
1990-2016: A systematic analysis for the Global Burden of Disease Study image segmentation,‖ Proc. 30th Chinese Control Decis. Conf. CCDC
2016,‖ Lancet, vol. 390, no. 10100, pp. 1211–1259, 2017, doi: 2018, pp. 5644–5648, 2018, doi: 10.1109/CCDC.2018.8408116.
10.1016/S0140-6736(17)32154-2. [22] S. Kolkur and D. R. Kalbande, ―Survey of texture based feature
[3] E. Wojtyna, P. Łakuta, K. Marcinkiewicz, B. Bergler-Czop, and L. extraction for skin disease detection,‖ Proc. 2016 Int. Conf. ICT Business,
Brzezińska-Wcisło, ―Gender, body image and social support: Ind. Gov. ICTBIG 2016, 2017, doi: 10.1109/ICTBIG.2016.7892649.
Biopsychosocial determinants of depression among patients with

258 | P a g e
www.ijacsa.thesai.org
 (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 13, No. 5, 2022

[23] H. Ji, Z. Shen, and Y. Zhao, ―Digital Gabor filters do generate MRA- [25] F. Hržić, I. Ńtajduhar, S. Tschauner, E. Sorantin, and J. Lerga, ―Local-
based wavelet tight frames,‖ Appl. Comput. Harmon. Anal., vol. 47, no. entropy based approach for X-ray image segmentation and fracture
1, pp. 87–108, 2019, doi: 10.1016/j.acha.2017.08.001. detection,‖ Entropy, vol. 21, no. 4, 2019, doi: 10.3390/e21040338.
[24] S. T. H. Rizvi, G. Cabodi, P. Gusmao, and G. Francini, ―Gabor filter [26] C.-C. Zhang and J.-D. Fang, ―Edge Detection Based on Improved Sobel
based image representation for object classification,‖ Int. Conf. Control. Operator,‖ in International Conference on Computer Engineering and
Decis. Inf. Technol. CoDIT 2016, pp. 628–632, 2016, doi: Information Systems, 2016, vol. 52, doi: 10.2991/ceis-16.2016.25
10.1109/CoDIT.2016.7593635.

259 | P a g e
www.ijacsa.thesai.org