Volume 5, Issue 6, June – 2020 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Dermatological Disorder Detection Using

Machine Learning
Ashlesha Gaikwad, Meghna Sonayallu, Shivani Tilekar, A.S.Deokar.
Department of Computer Engineering,
AISSMS College of Engineering PUNE

Abstract:- Skin diseases are considered one of the biggest human skin. A recognition system based on machine
scientific troubles in 21st century because of its especially learning should prove to be very beneficial in improving the
complex and luxurious prognosis with problems and accuracy in current work. The image processing methods are
subjectivity of human interpretation. In cases of deadly the approach given in this for extraction of the skin feature.
illnesses like Melanoma prognosis in early tiers play a Convolutional neural networks are to be used for automated
critical part in determining the possibility of getting detection of diseases in skin. The proposed method can be
cured. The software of automated strategies will assist in impactful and effective for skin disease detection, and it
early diagnosis specifically with photographs with appears to be an important approach. This system acquires
variety of analysis. Hence, in this system we present a various images of skin for extraction of features. The
completely automated machine of skin sickness analysis phase, pre-processing unit is used to remove noise,
recognition via lesion images, a device intervention in to convert gray scale using Gaussian filter, to convert binary
evaluation to traditional clinical personnel based images. Convolutional Neural Network (CNN) will be used
detection. This system is designed into 3 levels in our system for disease prediction where we have the input
compromising of statistics series and augmentation, unit of skin images training data set. Next have a hidden
designing version and subsequently prediction of disease. layer that acts on this training dataset to evaluate the results
This proposed system uses more than one AI algorithms train model of the output unit. This whole CNN works by
like Convolutional Neural Network and naive Bayes taking into account the elements, namely the matrix feature
classifier and amalgamated it with image processing of images, for designing a train model for recognition of
tools to shape a higher shape, leading to better accuracy. skin disease. When dealing with real-time skin disease
detection, will face limitations that will not produce results
Keywords:- Convolutional Neural Network, Naive Bayes with higher accuracy. In the future, to solve this constraint,
classifier, Dermatological Disorders, Machine Learning. must work with real-time skin disease data collection.

I. INTRODUCTION II. REVIEW OF LITERATURE

THE proposed system is for the detection of human Hassan Yasser et. al [1] stated that human pores and
pores and skin sicknesses primarily based on image skin coloration has been studied as biometric indicator, most
processing. Detection of disorder consists of strategies of preceding researches studies centered in the use of skin
which include acquisition of skin samples, pre-processing of color to locate the face, human and human movement. Goal
samples, segmentation of images, and extraction of of this article is to assess a version in human skin shade as
functions. The proposed system makes use of Convolutional completely unique code for identity reason. Fifty picture of
neural network and Naive Bayes classifier to discover the the nostril were captured the usage of a cell digital camera
tactics used to pick out human skin diseases. This system with a decision of three mega pixel. Area of the nose
additionally addressed several strategies of segmentation location has been the region of interest to symbolize a place
and extraction of capabilities used with extra accuracy in the of less direct publicity to the solar light as the relaxation of a
identity of human pores and skin ailment. The current face. The color pics had been transformed from RGB to HSI
research for the diagnosis of human skin disease is entirely layout in an effort to isolate the impact of the mild. Intensity
based on machine learning and neural network through in the course of shots. The snap shots had been analyzed
which human skin disease identification and detection is using mathematical and statistical methods ( Mean, Median,
conducted. For better accuracy and high reliability, the need wellknown deviation, kurtosis, skewness, and gray level co-
for human skin dataset reaches in size. On the basis of prevalence matrices ( GLCM ).
machine learning and image processing, system developing
human skin disease detection and prediction technology. Satishkumar L Varma et. al [2] states that human skin
The system will work on the limited size of the image detection coloration is crucial in numerous programs. There
dataset of the human skin downloaded from isic Archive. are various pores and skin primarily based packages in
The system use Convolutional neural network to achieve several areas namely gesture evaluation, face reputation,
high detection accuracy. Also use Naive Bayes classifier for character tracking, and nudity detection, pornography
performance evaluation. The proposed technique is used to filtering, website filtering, content based picture retrieval.
classify and identify the various skin diseases that infect Skin detection includes looking of pores and skin colored

IJISRT20JUN711 www.ijisrt.com 1380

Volume 5, Issue 6, June – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
pixels. There is numerous skin areas found in image or III. SYSTEM ARCHITECTURE / SYSTEM
video. It identifies the coloration pixel as a color of human OVERVIEW
pores and skin. It is one of the pre-processing steps to locate
human parts in pics. It has programs in numerous regions This system is going to overcome existing problem of
use pores and skin for photo processing. Skin detector skin disease diagnosis and quality loss of human skins due
transforms pixel into appropriate shade space. Skin is used to multiple diseases. The machine learning approach is used
as classifier to differentiate pixel as skin or non-pores and to accurate detection of human skin diseases. In current
skin pixel. Skin classifier uses threshold. Different color work done system is going to work on image processing
spaces are used for class over given database. techniques with the advantages of Convolutional neural
network feature of accuracy in image classification and
Shanmugavadivu Pichai et. al [3] affords A new recognition. System uses human skin dataset for training
combinatory and Multi-Colour theme System for the digital purpose to create train model which will further used to
pictures is bestowed during this paper. This pre-processing classify human skin diseases accurately. Overall working of
algorithmic rule finds application in external body part system is shown in fig.1.
detection in addition as in recognition. The Multi-Colour
theme System (MCSS) for Human Skin Detection in digital
pictures aims at object localization, supported principle of
neighborhood constituent process. This method thanks to its
process accuracy and performance is well-tried to possess a
footing over the competitive strategies. It conjointly assures Fig 1- Architecture of system
pay-off on reduced process quality, compared to facial
textures/ geometrics-based skin detection strategies. an The overall system design consists of following
intensive comparative analysis is disbursed to judge the modules:
result and influence of various multi-color bands for the 1. Human skin image acquisition
human skin classification. 2. Image processing
3. Feature extraction
Pratik Dubal et al. [4] gives Skin most cancers is the 4. Feature mapping and model generation
maximum commonplace form of most cancers, which This system implement the web based software
influences the existence of millions of people each yr. About application in python for skin disease prediction using image
three million humans are recognized with the disease each processing .Software performs the operations such as:
year in the United States by myself. The fee of survival
decreases steeply as the t he disorder progresses. However, 1) Human skin image acquisition
detection of pores and skin cancer in the early ranges is a Open-CV (Open Source Computer Vision) is a library
difficult and steeply-priced technique. In this examine, we of programming functions used for dynamic image
suggest a method that detects and identies pores and skin processing with computer vision. In our implementation
lesions as benign or malignant primarily based upon pics system uses open compute vision for taking human skin
taken from fashionable cameras. The photographs are images for further processing. After getting human skin
segmented, functions extracted by making use of the ABCD images processing applied on it for removing noise from it.
rule and a Neural Network is trained to classify the lesions
to a high diploma of accuracy. 2) Image processing
Then human skin image will send to image processing
P. B. Manoorkar et. al [5] states the basic approach of phase. In image processing image gets transformed in gray
detecting these pores and skin illnesses is thru visual format by removing noisy data in it using Gaussian filter.
inspection followed by way of biopsy and pathological After gray conversion phase image thresholding by setting
examination. If the medical doctor finds the arrival of lesion RGB colour values to zero and preserving only black and
dubious then commonly visible inspection approach is used white values. Gray to binary conversion is done by using
for diagnosis however all malignant lesions are not OTSU’s method. After getting black formatted image hand
identified via visible inspection. Now, there are not any shape get extracted from image. The exact shape of hand
usually established equipment that physician can use to at will get by forming edge using canny edge detection
once find the pores and skin ailment within the health method. After processing phase extract disease infected area
facility. Most shape of visible inspection could help to from input image.
prevent misdiagnosis of BCC and other varieties of pores
and skin diseases.

IJISRT20JUN711 www.ijisrt.com 1381

Volume 5, Issue 6, June – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
3) Feature extraction V. RESULTS
After getting proper shape of human skin disease
infected area features get extracted from it by using method This system accurately detects and classifies human
named as pixels weight calculations. The image pixels get skin diseases using convolutional neural network and
transfered in matrix by using weight gradient functions only machine learning approaches. It is very important for the
on drawn area of diseases. Feature extraction done on all successful detection of skin diseases and this can be done
human skin image dataset for training model creation and using image processing. Using Feature extraction and
drafting. The model creation done by using deep learning classification techniques to extract the features of infected
(CNN) algorithm. skin and the classification of human skin diseases. In our
system we can upload the image of infected area as shown in
4) Feature mapping and model generation fig 3 , so it will identify what type of disease person having
Human skin dataset is going through image processing in his/her infected body part. After identifying the disease it
and subsequent layers of feature extraction. After getting will provide to the user with its symptomsas shown in fig 4 .
image features these statistical features get mapped on In this way the system gives the user-friendly environment
machine files which is trained model. The runtime testing to the users.
image gets mapped with pre trained model and respective
outcomes will be generated. After generation of outcome
those results is nothing but our desire human skin disease
detection and prediction results.

IV. SYSTEM ANALYSIS

1) Convolutional Neural Network (CNN):-

In this system, CNN algorithm is used which takes
human skin disease as dataset for creating machine model.
After getting skin images it will processed using image
processing method for feature evaluation. System can
extract the different features from that images regardless of
their size and shapes in it consist by using series of
Fig 2:- Homepage of system
mathematical methods to determining the skin diseases.
Each layer in CNN has capability to find weights of images
by using matrix evaluations which converts input to output
with valuable methods. Layers of CNN used to determine
skin disease from dataset images and give prediction by
preserving higher accuracy and less time.
Steps are as follows:-
a) Input skin disease dataset ( ISIC’s )
b) Image processing by using ML libraries
c) Feature Extraction from images in the form of
bottlenecks generated during train model.
d) Model mapping by using machine learning framework.
e) Skin disease recognition by providing skin disease and
healthy skin images.
f) Performance evaluation by using test dataset images of Fig 3:- Uploading Input Image
human skin disease.

2) Naive Bayes algorithm:-

This Naive Bayes algorithm is divided down into 5
parts:
a) Separate by Skin Disease Classes.
b) Summarize Skin disease Dataset.
c) Summarize Data By Class called detected dis-ease
category.
d) Gaussian Probability Density Function.
e) Class Probabilities for each type of disease. All above
steps will give the basement that we need to implement
Naive Bayes classifier in our implementation and apply it
to our own predictive framework for skin disease Fig 4:- Disease Detected.
detection.

IJISRT20JUN711 www.ijisrt.com 1382

Volume 5, Issue 6, June – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
VI. CONCLUSION [10]. Jainesh Rathod, Vishal Waghmode, Aniruddh Sodha,
Dr. Prasenjit Bhavathankar, “Diagnosis of skin
Thus implemented system is a web based software diseases using Convolutional Neural Networks”, IEEE,
application in python for skin disease prediction using image November 2018.
processing. System uses operations such as: [11]. Yanhui Guo, Amira S. Ashour, Lei Si, Deep P
1) Human skin image acquisition Mandalaywala, “Multiple Convolutional Neural
2) Image processing Network for Skin Dermoscopic Image Classification”,
3) Feature Extraction Institute of Electrical and Electronic Engineers(IEEE),
4) Feature mapping and model generation. 2018.
[12]. Archana Ajith, Vrinda Goel, Priyanka Vazirani, Dr. M.
FUTURE WORK Mani Roja, “Digital Dermatology Skin Disease
Detection Model using Image Processing”, Institute of
The future research will be extended for further Electrical and Electronic Engineers(IEEE), 2017.
improvement in skin disease recognition accuracy and work
for real time human skin disease recognition.

ACKNOWLEDGMENT

This work is supported in a Human Skin Disease

research field in India. Authors are thankful to Faculty of
Engineering and Technology (FET), Savitribai Phule Pune
University, Pune for providing the facility to carry out the
research work.

REFERENCES

[1]. Hassan Yasser, “Human Skin Color Code Recognition:

A Case Study 2013 INTERNATIONAL
CONFERENCE ON COMPUTING, ELEC-TRICAL
AND ELECTRONIC ENGINEERING (ICCEEE). c
2013 IEEE.
[2]. Satishkumar L Varma, “Human Skin Detection Using
Histogram Process-ing and Gaussian Mixture Model
Based on Color Spaces, 9781-5386-1959-9/17 c 2017
IEEE.
[3]. Shanmugavadivu Pichai, “Human Skin Detection in
Digital Images using Multi Colour Scheme System” ,
(ICIIP) 978-1-50906734-3/17 c 2017 IEEE.
[4]. Pratik Dubal, “Skin Cancer Detection and
Classication” , 978-153860475-5/17/ c 2017 IEEE.
[5]. P. B. Manoorkar, Analysis and Classification of
Human Skin Diseases , 978-1-5090-2080-5/16 c 2016
IEEE.
[6]. Sabah Bashir, Navdeep Sharma “Remote Area Human
skin Disease Detection Using Image Processing”,
IOSR Journal of Electronics and Communication
Engineering , ISSN : 2278-2834 Volume 2, Issue 6
2012, PP 31-34.
[7]. Smita Naikwadi, Niket Amoda” ADVANCES IN
IMAGE PROCESSING FOR DETECTION OF
HUMAN SKIN DISEASES” , International Journal of
Application or Innovation in Engineering and
Management , Volume 2, Issue 11, November 2013.
[8]. Sanjay B. Patil et al. “ DISEASE SEVERITY
MEASUREMENT USING IMAGE PROCESSING”,
International Journal of Engineering and Technology
Vol.3 (5), 2011, 297-301.
[9]. Anand.H.Kulkarni, Ashwin Patil R. K.” Applying
image processing technique to detect human skin
diseases”, International Journal of Modern Engineering
Research, Vol.2, Issue.5, Sep-Oct. 2012 pp-36613664.

IJISRT20JUN711 www.ijisrt.com 1383