Similarity Report ID: oid:27992:59096314

PAPER NAME

B6

WORD COUNT CHARACTER COUNT

6938 Words 42447 Characters

PAGE COUNT FILE SIZE

31 Pages 817.7KB

SUBMISSION DATE REPORT DATE

May 12, 2024 4:42 PM GMT+5:30 May 12, 2024 4:43 PM GMT+5:30

59% Overall Similarity

The combined total of all matches, including overlapping sources, for each database.
21% Internet database 49% Publications database
Crossref database Crossref Posted Content database
21% Submitted Works database

Excluded from Similarity Report

Bibliographic material Quoted material

Summary
 SKIN SIGHT
14
submitted in fulfillment of requirements for the award of the degree of

BACHELOR OF TECHNOLOGY
IN
INFORMATION TECHNOLOGY

Submitted By
Abhishek Laddha
(21124005)
Anugya Singh
(21124014)
Deeksha Verma
(21124029)
Kushagra Sharma
(21124058)

Under the Mentorship of

Dr. Vijay Kumar

5
DEPARTMENT OF INFORMATION TECHNOLOGY
DR. B.R. AMBEDKAR NATIONAL INSTITUTE OF TECHNOLOGY
JALANDHAR-144027, PUNJAB (INDIA)
May 02, 2024
 Undertaking

We declare that the project work presented in this report entitled Skin Sight, submitted to
the Department of Information Technology, Dr. B R Ambedkar National Institute of
Technology Jalandhar, for the award of the Bachelor of Technology degree in Information
Technology, is our original work. We have not plagiarized or submitted the same work for
the award of any other degree. In case this undertaking is found incorrect, we accept that
our degree may be unconditionally withdrawn.

Abhishek Laddha 21124005

Anugya Singh 21124014

Deeksha Verma 21124029

Kushagra Sharma 21124058

5
Department of Information Technology

Dr B R Ambedkar National Institute of Technology

Jalandhar, Punjab, India

May 02, 2024

 Certificate

This is to certify that the project report entitled Skin Sight submitted by Abhishek Laddha
(21124005), Anugya Singh (21124014), Deeksha Verma (21124029) and Kushagra
5
Sharma (21124058), to the Dr B R Ambedkar National Institute of Technology Jalandhar,
in partial fulfillment for the award of the degree of B. Tech in Information Technology has
been carried out under our supervision and that this work has not been submitted elsewhere
for a degree.

Dr. Vijay Kumar

Associate Professor

Department of Information Technology

Dr B R Ambedkar National Institute of Technology

Jalandhar, Punjab, India

May 02,2024

III
 Acknowledgments

39
We would like to acknowledge and give our most grateful and warmest thanks to our
supervisor Dr. Vijay Kumar for his continuous support and encouragement. His guidance and
experience were a key factor in the success of the execution of the project.

We would like to thank the members of the team for the tons of sleepless nights they have put
on this project.

Lastly we thank our classmates who helped us a lot in all the difficulties which we faced
throughout the project.
12
We also thank all respondent who have given their value time, views and authentic
12
information for this minor project. We thank each and everybody who has contributed
directly or indirectly to the successful completion of this project.

32
IV
 Table of Contents
Sr. No. Topic Page No.
16
1 Introduction

1.1 Introduction 1

1.2 Existing system 2

1.3 Literature Review 3

1.4 Limitations of the system 11

1.5 Problem Statement 11

1.6 Need of System 12

1.7 Hardware Specifications 13

1.8 Software Specifications 13

1.7 Scope of the Program 14

1.8 Objectives of the Project 14

1.9 Features of new System 14

2 Software Requirement Specification 15

3 Design 18

4 Database Design 20

5 Data Dictionary 21

6 Bibliography and References 24

26
V
 LIST OF TABLES

Table No. Table Name Page No.

34
1 Hardware Specification 13

2 Software Specification 13

3 User’s Database 20

4 User’s Record Database 20

29
VI
 LIST OF FIGURES

Figure No. Figure Name Page No.

1 Use Case diagram overview 18

2 Activity diagram overview 19

3 Home page 21

4 Book an appointment page 21

5 About Us page 22

6 Check your Skin page 22

7 Our Services page 22

8 FAQs page 23

9 Logo 23

VII
 INTRODUCTION

1.1 Introduction

Skin conditions and dermatological manifestations are prevalent health concerns globally,
20
affecting individuals of all ages and demographics. Timely and accurate diagnosis of these
2
conditions is crucial for effective treatment and management. However, it is very difficult to
provide better dermatological care to under-served or resource-poor regions in a cost-effective
manner owing to unavailability of efficient diagnostic tools, lack of connectivity, and poor
2
laboratory infrastructure etc. . Even, preliminary screening of a dermatological manifestation
seems to be an arduous task.

In recent years, Artificial Intelligence (AI) has revolutionized various sectors of healthcare,
11
including dermatology. AI-based tools for the diagnosis of dermatological manifestations have
emerged as promising solutions to improve accuracy, efficiency, and accessibility in skin
27
disease diagnosis. These tools utilize advanced machine learning algorithms to analyze images
of skin lesions and provide preliminary assessments, assisting healthcare professionals in
making informed decisions.

11
The AI-based tool for preliminary diagnosis of dermatological manifestations utilizes advanced
machine learning algorithms to analyze and classify skin conditions. By leveraging a vast
database of images and patient data, the tool can provide accurate and efficient preliminary
diagnoses, aiding healthcare professionals in their decision-making process.

1
1.2 Existing System
11
Here are some existing projects related to AI-based tools for the diagnosis of
dermatological manifestations:
1. DermDetect[1]:
 Description: DermDetect is a web-based platform that uses deep learning
35
algorithms to assist in the diagnosis of skin diseases. Users can upload
40
images of skin lesions, and the system provides preliminary assessments and
recommendations.
2. SkinVision[2]:
 Description: SkinVision is a mobile application that utilizes AI technology
to analyze photos of skin lesions and provide risk assessments for skin
cancer. It offers users a convenient way to monitor changes in their skin and
seek medical advice if necessary.
3. MetaOptima's DermEngine[3]:
 Description: DermEngine is a comprehensive dermatology software
platform that integrates AI-powered imaging, analytics, and documentation
tools. It assists healthcare professionals in diagnosing and managing skin
conditions effectively.
4. Miiskin[4]:
10
 Description: Miiskin is a skin tracking app that allows users to monitor
changes in their moles and skin lesions over time. It uses AI algorithms to
analyze images and provide insights into potential changes that may indicate
skin cancer or other conditions.
5. VisualDx[5]:
38
 Description: VisualDx is a clinical decision support system that includes a
comprehensive database of dermatological images and information. It helps
healthcare professionals in diagnosing and managing a wide range of skin
conditions using AI-driven image recognition technology.

1.3 Literature review

9
i. Deep learning for AI-based diagnosis of skin-related neglected tropical diseases: a pilot study[6]

2
 33 13
Summary: This study explores the potential of deep learning models in diagnosing skin-related
neglected tropical diseases (skin NTDs), particularly focusing on populations with darker skin
tones. Using photographs collected in West African countries, the researchers developed
13
convolutional neural network (CNN) models, specifically ResNet-50 and VGG-16, to classify
9
images into five targeted diseases: Buruli ulcer, leprosy, mycetoma, scabies, and yaws. They
13
found that the ResNet-50 model outperformed VGG-16, especially with larger training datasets.
Interestingly, using laboratory-confirmed cases for training only marginally improved accuracy,
13
suggesting clinical diagnosis alone can be reliable to some extent.
The study highlights challenges such as imbalanced datasets and heterogeneity in image quality
due to diverse capture conditions. Despite limitations, the findings suggest that deep learning
can aid in screening and diagnosing skin NTDs, potentially addressing healthcare gaps in low-
resource settings. Further research is needed to refine models and overcome challenges,
emphasizing the importance of diverse datasets and local expertise.
9
Principle Derived: The two models were able to correctly predict over 70% of the diagnoses,
and there was a consistent performance improvement with more training samples. The ResNet-
50 model performed better than the VGG-16 model. Model trained with PCR confirmed cases
of Buruli ulcer yielded 1-3% increase in prediction accuracy over training sets including
unconfirmed cases.

3
ii. Artificial Intelligence in Cosmetic Dermatology: A Systematic Literature Review[7]
3
Summary: Over the last ten years, the field of dermatology has experienced significant
advancements through the utilization of artificial intelligence (AI) technologies. The adoption
of such technologies is multifaceted, encompassing tasks such as screening, diagnosis,
treatment, and prediction of treatment outcomes. The majority of prior systematic reviews in
this domain were centered on medical dermatology, with the aim of detecting and managing
serious skin diseases such as skin cancer. However, the adoption of AI in cosmetic dermatology,
which focuses on improving skin conditions for cosmetic purposes, has not been
comprehensively reviewed. Therefore, the objective of this systematic review article is to
analyze the existing and recent research revolving around applications of AI in the field of
cosmetic dermatology. The study encompasses articles published between 2018 and 2023,
where a total of 63 publications are deemed relevant based on the established inclusion criteria,
divided into five categories based on utilization domains, namely cosmetic product
development, skin assessment, skin condition diagnosis, treatment recommendation, and
treatment outcome prediction. This systematic review article provides not only valuable
insights for researchers interested in exploring new research areas related to aesthetic medicine
but also applicable guidance for practitioners seeking to implement AI technologies to address
real-world challenges in cosmetic services.

3
 6
Principle Derived: A scheme for recommending cosmetic products that utilize Convolutional
Neural Networks (CNN). The study employed image analysis techniques to predict various
categories of consumer facial images based on skin health. This was achieved by extracting
relevant features such as shape, texture, and color from the photographs. The algorithm that
was proposed attained a success rate of 97.38% accuracy in recommending items on the test
data.
6
A ranking algorithm was developed with the aim of suggesting the optimal product based on
the specific needs of consumers and their individual skin types. ML techniques were employed
to evaluate skin images, genetic data, and other pertinent factors for the purpose of providing
personalized treatment recommendations in the field of aesthetic dermatology. This
methodology facilitated the development of a therapeutic regimen that is both effective and
tailored to specific customer requirements while considering the unique clinical manifestations,
medical background, genetic makeup, surroundings, and lifestyles. These techniques aim to
enhance the accuracy and efficiency of the system in aligning customers with the optimal
products and treatments based on their individual requirements, skin characteristics, and
dermatological conditions.
8
iii. Recent Advances in Diagnosis of Skin Lesions Using Dermoscopic Images Based on
Deep Learning[8]
8
Summary: Skin cancer is one of the most threatening cancers, which spreads to the other
parts of the body if not caught and treated early. During the last few years, the integration of
deep learning into skin cancer has been a milestone in health care, and dermoscopic images
are right at the center of this revolution. This review study focuses on the state-of-the-art
automatic diagnosis of skin cancer from dermoscopic images based on deep learning. This
work thoroughly explores the existing deep learning and its application in diagnosing
dermoscopic images. This study aims to present and summarize the latest methodology in
melanoma classification and the techniques to improve this. We discuss advancements in
deep learning-based solutions to diagnose skin cancer, along with some challenges and future
opportunities to strengthen these automatic systems to support dermatologists and enhance
their ability to diagnose skin cancer.
7
Principle derived: TensorFlow is an open-source machine learning platform focusing on
neural networks, which was developed by the Google Brain team. The main reason for
choosing TF over other DL frameworks is its popularity. TensorFlow is mighty and easy to
use and has excellent community support. Keras was designed by Google to enable fast
experimentation with neural networks. It is very user-friendly, modular, and extensible.
Keras also has the advantage of being simple, flexible, and powerful. Because of these
features, Keras is viewed by newcomers as the go-to DL framework. Since PyTorch was
developed by Facebook and offers an easy- to-use interface, its popularity has gained
momentum, par-ticularly in academia. PyTorch is the main competitor of TF. MatConvNet
is a toolkit based on CNN for Matlab, supporting both CPU and GPU. In fact, this toolkit not
only supports CNN, but also supports some other networks such as RNN, LSTM, etc. Caffe
is an early DL framework made with expression, speed, and modularity. It is ideal for

4
 feedforward neural networks and image processing tasks. Theano is based on python whose
development started in 2007. This library is good at dealing with multidimensional arrays.
With the strong rise of Tensorflow, Keras and Pytorch, MatConvNet, Caffe, Theano are
declining day by day, and fewer and fewer researchers use them. A convolutional neural
network (CNN), also known as ‘‘ConvNet’’, is a specific type of feed-forward neural
network with a stack of convolutional layers, each followed by pooling layers in order to
extract features from the input data and produce a set of high level feature maps at each level
of convolution. The feature maps information is summarized using pooling layers in order
to reduce the number of parameters and uses a fully connected layer to produce the final
classification . The CNN structure evolution summarized in this arti cle started with the
neurocognitive machine model. At the same time, the convolutional structure has appeared.
The LeNet CNN structure became available in 1998. However, the CNN’s edge began to be
overshadowed by hand-designed features such as support vector machine (SVM). With the
introduction of rectified linear unit (ReLU) and Dropout, as well as the historic opportunities
brought by graphics processing units (GPUs) and big data, CNN ushered in a landmark
breakthrough in 2012 - AlexNet. Figure 6 presents the evolution of the CNN
structure. Today, researchers rarely build models from start to finish. Common features of
classic models have been encapsulated in DL frameworks (such as TF or PyTorch).
Researchers only make some modifications on this basis. All the literature collected in this
study is based on the CNN model. Compared with traditional machine learning, the CNN
model has excellent feature representation (automatically learned from raw data). Currently,
the primary method of skin disease image recognition is to use a CNN in DL, and then to
use pooling for image recognition. The research work collected in this study adopted famous
CNN architecture, such as AlexNet , VGG (short for ‘‘Visual Geometry Group’’) ,
Inception , ResNet (short for ‘‘residual neural network’’) [20], DensenNet ,
EfficientNet and so on. Figure 7 plots the state-of-art models’ per- 251 formances in dataset
ImageNet from 2011 to 2021.
3
iv. Deep Learning-Based Dermatological Condition Detection: A Systematic Review With
Recent Methods, Datasets, Challenges, and Future Directions

Summary: Dermatological conditions are a global health concern affecting all age groups.
They encompass various skin issues such as rashes, moles, acne, blisters, hives, nodules,
cysts, macules, and papules. Early diagnosis of dermatological conditions is crucial to
prevent skin damage and chronic diseases. Recent advancements in artificial intelligence and
medical image processing, particularly through deep learning, have significantly improved
the precision and efficiency of dermatological disease detection by dermatologists. This
paper thoroughly examines deep learning-based methods for detecting dermatological
conditions from dermoscopic images. Specifically, it presents study of 22 methods that are
used to detect dermatological conditions such as basal cell carcinoma, melanocytic nevus,
seborrheic keratosis, psoriasis, benign keratosis, melanoma, acne, cold sore, warts, eczema,
hives, shingles, scar, skin tag, inflammatory hyper pigment, cyst, dark circle, blackhead, burn
and skin rash. It also covers clinical diagnostic methods for dermatological conditions and
the need for computer-aided diagnosis. In this paper, we have also proposed the

5
 categorization of deep learning-based dermatological condition detection methods. Moreover,
a comprehensive summary of these methods is presented. In addition, this paper summarizes
the majority of the datasets available for computer-aided detection of dermatological
conditions. Furthermore, it presents enormous challenges and potential future research
directions. This survey informs researchers about the latest advancements in deep learning-
based detection methods for dermatological conditions and allows dermatologists to stay
updated on technological breakthroughs in this area.

Principle Derived:
1
 SKIN DISEASE DIAGNOSIS USING PRE-TRAINED CNN
Computerized technique utilizing VGG-16, Inception, Xception, Alexnet, VGG-19, and
Googlenet has been proposed for diagnosing skin diseases. illustrates the major steps
involved in this technique. Firstly, the dermoscopic image undergoes preprocessing, which
includes downsizing and image augmentation methods. Following preprocessing, the
affected region is segmented using a hybrid watershed region-based image segmentation
method. This segmentation aims to determine the watershed line within the dermoscopic
image to identify prominent regions. The hybrid method combines both edge-based and
regionbased segmentation techniques. The segmented image is then used to train
convolutional neural network (CNN) models to predict one of the four classes: basal cell
carcinoma, melanocytic nevus, seborrheic keratosis, and psoriasis. The authors of the paper
observed that among the various CNN models tested, GoogleNet demonstrates superior
efficiency and accuracy while requiring minimal computational power. The performance
evaluation of this technique encompasses metrics such as accuracy, precision, sensitivity,
recall, and specificity.

 CLASSIFICATION OF SKIN DISEASE USING MOBILENET V2 WITH LSTM

4
Technique for skin disease classification is proposed, leveraging the integration of
MobileNet V2 with LSTM. The key advantage of this technique lies in the utilization of
LSTM to enhance classification accuracy by preserving the state information of the features.
The workflow of this technique is depicted in . Due to a slight classimbalance in the skin
disease dataset, data augmentation techniques such as rotation and transformation are applied.
MobileNet V2 is used to categorize the various skin diseases, and LSTM is utilized to
1
improve the model’s performanceby retaining feature information encountered in the initial
version of image classification. MobileNet V2 is employed for feature extraction, offering
computational efficiency suitable for low-power computing devices. The core structure of
MobileNet V2 is built on several abstraction layers, each of which is a component of a
distinct convolution layer that appears to be the quantized configuration for thoroughly
examining a problem’s difficulty. The complexity of 1 × 1 is referred to as point-wise
complexity. The general framework of the MobileNet V2 with the LSTM model is an
assortment of multiple convolutions and max pooling layers including the LSTM element
that is linked to the flattening layer of the model. The fully connected layer involves training
to correlate the recognized characteristics with pre- existing data. Finally, there is the
softmax layer, which calculates the probability of distinct skin disease categories. Following

6
 feature extraction, MobileNet V2 assesses the probability of skin disease, categorizing
dermoscopic images into six classes: melanocytic nevus (nv), benign keratosis-like lesions
(bkl), dermatofibroma (df), vascular lesions (vasc), basal cell carcinoma (bcc), and
melanoma (mel). The technique is evaluated by configuring hyperparameters, including
1
training and validation loss. Performance evaluation measures encompass sensitivity,
specificity, accuracy, Jaccard Similarity Index (JSI), and Mathew Coefficient Correlation
(MCC).

 SKIN CANCER DIAGNOSIS USING ANN, CNN, KNN AND GAN

A deep learning-based technique for diagnosing skin cancer has been presented in . The
1
workflow of this technique shows the seven different categories of skin cancer. In the first
step, the dermoscopic image is cleaned using hair removal, noise removal, and contrast
enhancement methods. Subsequently, the cleaned image is segmented using edge-based,
region-based, and morphological-based segmentation methods. Next, the segmented images
train the model to classify one of the dermoscopic images into seven skin cancer categories:
nevus, dermatofibroma, melanoma, vascular, pigmented Bowens, pigmented benign
keratoses, and basal carcinoma. The authors observed that among the deep learning models
(ANN, CNN, KNN, and GAN), GAN gives better classification accuracy. The performance
of this technique is assessed using accuracy, sensitivity, and specificity.

 SKIN DISEASE RECOGNITION USING DEEP NEURAL NETWORKS

In , a computer-aided diagnostic technique based on Deep Neural Networks (DNN) is
proposed for skin disease recognition. This technique leverages two publicly available skin
4
image datasets: ISIC and DermNet.The process begins with the preprocessing of
1
dermoscopic images, which includes noise removal, hair removal, andcontrast enhancement
methods. Subsequently, color and texture features are extracted from the preprocessed
images. These extracted features are then utilized in the classification stage. During
classification, the weight optimization method is employed with DNN to diagnose one of the
18 classes, including Seborrheic dermatitis, basal cell carcinoma, squamous cell carcinoma,
melanoma, acne, vitiligo, rosacea, dermatofibroma, eczema, cold sore, warts, actinic
keratosis, shingles, hives, sunburn, glandular rosacea, bullous pemphigoid, and perioral
dermatitis. The weight optimization method offers the advantage of minimizing prediction
errors and increasing classification accuracy. The performance evaluation of this technique
includes sensitivity, specificity, accuracy, Positive Predicted Value (PPV), Negative
Predicted Value (NPV), false positive rate, and false negative rate.

 DETECTION OF SKIN DISEASE USING A HYBRID COMBINATION OF LBP, GLCM AND

ANN
In , a skin disease detection method using a hybrid combination of LBP and GLCM with
ANN has been offered. The process of this method is depicted in . In this method, the
dermoscopic image is preprocessed using the Gaussian filter to filter unnecessary noise.
After preprocessing, the active count segmentation method is used to identify the region of
interest from the preprocessed image. Next, a hybrid combination of two methods with ANN
is used to extractfeatures from the segmented image. This feature extraction method extracts

7
 216 texture features to classify skin diseases into one of the six classes viz. melanoma (mel),
nevus (nv), dermatofibroma (df), vascular, basal cell carcinoma (bcc), actinic keratoses and
intraepithelial carcinoma (akiec), and benign lesions of keratosis (bkl). The performance of
this method is tested using accuracy, sensitivity, specificity, and recall.

 SKIN DISEASE DETECTION USING DENSENET121, INCEPTIONRESNETV2, AND

RESNET152V2
In , a technique for detecting skin diseases utilizing DenseNet121, InceptionResNetV2, and
ResNet152 V2 is presented. The workflow of this technique is depicted in . Initially, the
dataset exhibits imbalanced classes of skin diseases. Data augmentation techniques such as
flipping, rotation, translation, color space augmentation, image mixing, kernel filters, and
image space augmentation are applied to address the problem of imbalanced classes.
Following augmentation, the images undergo preprocessing involving resizing, color
normalization, and splitting techniques. Pretrained networks, including DenseNet121,
InceptionResNetV2, and ResNet152V2, are employed to extract features and diagnose one of
the five skin diseases: malignant melanoma, atypical nevus, dermal nevus, acral lentiginous
melanoma, and nevus.
1
The authors observed that ResNet152V2 outperforms the other pre-trained networks
regarding efficiency and classification accuracy. Performance evaluation of this technique
encompasses sensitivity, specificity, recall, precision, and f1-score as metrics.

 SKIN DISEASE CLASSIFICATION USING ALEX NET AND VGG-16

In a skin disease classification technique using AlexNet and VGG-16 has been proposed.
Two datasets, ISIC 2019 and PH2 are used to evaluate this technique. These datasets include
information on ten different skin diseases such as psoriasis, melanoma, melanocytic nevus,
1
benign keratosis, basal cell carcinoma, squamous cell carcinoma, actinic keratosis,
dermatofibroma, vascular lesion, and eczema. The main steps involved in this technique are
depicted in . First, the image is pre-processed using illumination correction, contrast, and
edge enhancement methods. These preprocessing methods remove artifacts such as hair,
noise, and ink markers from the dermoscopic image. After the preprocessing step, the image
is augmented using basic geometric transformation methods. The geometric transformation
methods include zooming, cropping, scaling, rotation, and translation. Next, the features are
extracted from the dermoscopic image using pre-trained CNN models: AlexNet and VGG-16.
The extracted features classify the dermoscopic image in one of the ten skin diseases. The
performance of this technique is assessed using accuracy, mean precision, true positive rate,
false positive rate, true negative rate, AUC, and ROC.

1
 ARAL MELANOMA DETECTION USING THE VGG-16 MODEL
A deep learning-based method is proposed for detecting acral melanoma, the most prevalent
1
type among Asians,and is often associated with a poor prognosis due to late detection. The
methodology of this method is presented diagrammatically in , outlining two main stages:
training and inference. Given the limited quantity of dermoscopic images available for
training, the authors employed augmentation techniques such as cropping and resizing. The
VGG-16 model was then trained for feature extraction using the augmented dermoscopic

8
 images. To enhance the model’s performance, the authors fine-tuned the VGG-16 model by
introducing sixteen additional layers, including thirteen convolutions and three fully
connected layers. This fine-tuning process not only improved feature extraction but also
mitigated overfitting. In the inference stage, the trained VGG-16 model was utilized to
diagnose the two classes: non-melanoma and acral melanoma. The performance of the
proposed technique was assessed using accuracy, sensitivity, specificity, AUC, and
Youden’s index values as evaluation metrics.

 AUTOMATIC MELANOMA DIAGNOSIS FROM DERMOSCOPIC IMAGES USING HAND-

CRAFTED FEATURES WITH CNN
A CNN-based method has the advantage that it helps gain better classification accuracy
1
compared to traditionalCNN. The method’s functioning is depicted in . Firstly, the affected
area is segmented by finding the relation between the pixels and neighborhood pixels using
the region growing technique. Next, the hand-crafted features are extracted, as described in ,
from the segmented image using CNN combined with the scattered wavelet transform
method. The extracted hand-crafted features are used to classify the dermoscopic image into
one of the three classes of melanoma: basal cell carcinoma, melanoma, and nevus. The
performance of this method is tested using accuracy, sensitivity, and specificity.

 COMPUTERIZED ANALYSIS FOR DIFFERENT TYPES OF PSORIASIS USING VGG-19

Psoriasis, a chronic disease caused by immune system dysfunction, is recognized as the most
common dermatological condition. illustrates the distinct types of psoriasis. The workflow
of the approach is depicted in. The approach begins with pre-processing the dermoscopic
image to eliminate distortions using image enhancement techniques. Next, the prominent
region is segmented by identifying the edges in the dermoscopic image through edge
detection methods. Following segmentation, features are extracted utilizing the VGG-19
model, encompassing colon, texture, and shape attributes. These extracted features are
subsequently employed to classify the dermoscopic image into one of six classes: plaque
psoriasis, guttate psoriasis, inverse psoriasis, pustular psoriasis, erythrodermic psoriasis, and
normal skin. On a dataset of 1838 photos, the suggested method’s performance is assessed,
1
and measures including accuracy,sensitivity, specificity, recall, and precision are used to
gauge its efficacy.
1
 AUTOMATED GRADING OF ACNE VULGARIS USING DENSENET15, INCEPTION V4,
AND RESNET-18
The processing steps involved in this method are illustrated in . Initially, the Region of
Interest (ROI) is segmented using the adaptive growth cut method. Following segmentation,
the segmented ROI is augmented using random combinations of right and left flipping,
contrast adjustment, scaling, and shifting techniques. Subsequently, the dermoscopic images
are utilized for training CNN models, specifically DenseNet15, Inception v4, and ResNet-18,
for feature extraction. Based on the extracted features, the dermoscopic image is classified
into two classes: acne and normal skin. Additionally, the dermoscopic image is further
classified into three severity groups of acne: IGA Grades 0-1 (clear/almost clear skin), IGA
Grade 2 (mild), and IGA Grades 3-4 (moderate to severe). The performance evaluation of

9
 this method involves 484 frontal facial images, each with a pixel size of 2000×3000,
acquired from 420 individuals. The method’s effectiveness is assessed using accuracy,
sensitivity, specificity, Pearson correlation, recall, and precision metrics.

 SKIN CONDITIONS IDENTIFICATION USING DENSENET15, INCEPTION V4 AND

1
RESNET18 Along with the dermoscopic image, the metadata of the patient is considered.
The features are extracted using the Inception V4 model to classify the dermoscopic image in
one of the 27 classes of skin diseases. Actinic keratosis, allergic contact dermatitis,
androgenetic alopecia, hidradenitis, basal cell carcinoma, lentigo, melanocytic nevus,
postinflammatory hyperpigmentation, psoriasis, squamous cell carcinoma, and seborrheic
keratosis are among the conditions that fall into these categories. The performance of this
method is evaluated using accuracy, sensitivity, and specificity. The results are compared
with 360 specialized dermatologists who have annotated the dermoscopic images.
M. AN ENSEMBLE LIGHTWEIGHT COMBINATION OF MOBILENETV1 AND
DENSENET-121 NETWORK FOR SKIN CANCER DETECTION The method’s workflow
is illustrated in . The technique consists of three main steps: pre-processing, feature
extraction, and classification. Initially, the dermoscopic image undergoes augmentation using
an image augmentation method. Subsequently, features are extracted from the augmented
image using a lightweight CNN architecture that combines MobileNet V1 and DenseNet-121
models. This combination offers the advantage of extracting highly discriminant features,
improving accuracy. benign or The extracted features are then used to classify the image into
melanoma classes. The technique’s performance is evaluated using various indicators such as
accuracy, Jaccard index, dice coefficient, specificity, and sensitivity.

1
 AUTOMATED DETECTION OF MELANOMA USING FULLY CONVOLUTIONAL RESIDUAL
NETWORK (FCRN)
1
The dermoscopic image undergoes preprocessing utilizing cropping and resizing techniques.
Subsequently, the region of interest is segmented using the Fully Convolutional Residual
Network (FCRN). FCRN offers significant advantages regarding segmentation accuracy,
memory efficiency, and computing time. The segmented images are then employed to train
the Deep Residual Network (DRN), distinguishing between benign and malignant melanoma.
DRN is crucial in extracting representative and specific features from the dermoscopic image,
4
enhancing classification accuracy. The performanceof the proposed method is evaluated
using multiple metrics, including sensitivity, specificity, accuracy, Jaccard index, and dice
coefficient.
.

4
 CLASSIFICATION OF MELANOMA USING DCN
The main objective of an automated approach for efficient melanoma detection using a Deep
Convolutional Neural Network (DCNN) is to achieve an accurate classification of melanoma
while utilizing a less complex DCNN architecture . The functioning of this method is
demonstrated in . A data normalization technique is employed to eliminate redundancy and
address the issue of non-normalized dermoscopic images in the dataset. Subsequently, a
noise removal technique is applied during preprocessing to ensure that the DCNN model

10
 operates effectively on artifact-free raw images. This technique helps remove unwanted
elements
such as hairlines, gel bubbles, ruler markings, and ink markers. The preprocessed image is
then augmented using rotation, scaling, and translation methods to enhance the dataset
diversity. The augmented images are utilized for training the DCNN model, enabling the
classification into two classes: melanoma and benign. The methodology involves
experimentation with hyperparameters such as learning rate, mini-batch size, epoch,
activation function, momentum, and regularization. The performance evaluation of this
method is conducted using metrics such as accuracy, precision, recall, specificity, and F1-
score.

v. Skin rash detection using DCNN

Summary: An automated method to detect skin rash using ResNet34, ResNet-50, and
InceptionV3 has been proposed in . The main steps of this technique are shown in . The
dermoscopic image is resized at a size of 224 × 224 using the image resizing method. After
resizing, the resized image is augmented using the horizontal flipping and rotation technique.
Next, the augmented images train the CNN models to detect one of the two classes: skin rash
and normal skin. The authors observed that among the CNN models (ResNet-34, ResNet-50,
and InceptionV3), ResNet50 showed better classification accuracy with less computation
time. The performance of this technique is evaluated using accuracy, sensitivity, and
specificity

1.4 Limitations of the System

 Complexities
 Low accuracy
 Not giving detailed description of diagnosis

1.5 Problem Statement

2
The Global Burden of Disease project has shown that skin diseases continue to be the 4th
leading cause of nonfatal disease burden worldwide. These conditions are often the presenting
face of more severe systemic illnesses, including HIV and neglected tropical diseases (NTD).
such as elephantiasis and other lymphedema-causing diseases. Additionally, skin disorders
pose a significant threat to patients' well-being, mental health, ability to function, and social
participation. However, it is very difficult to provide better dermatological care to under-
served or resource-poor regions in a cost-effective manner owing to unavailability of efficient
diagnostic tools, lack of connectivity, and poor laboratory infrastructure etc. Moreover, there
is also a scarcity of physicians with dermatological training. Even, preliminary screening of a
dermatological manifestation seems to be an arduous task. Thus, developing an Artificial
intelligence-based tool (through Image processing technique) for preliminary diagnosis of
numerous dermatological conditions will prove to be a boon in the health care system.

11
1.6 Need of the System

1. Improved Diagnostic Accuracy: Dermatological conditions can vary widely in

appearance and presentation, making accurate diagnosis challenging, even for experienced
dermatologists. AI-based tools offer the potential to enhance diagnostic accuracy by
leveraging advanced image analysis techniques and machine learning algorithms to
recognize subtle patterns and features in skin lesions.
2. Efficiency and Timeliness: Dermatology clinics often face high patient volumes and long
wait times for appointments. AI-based tools can help alleviate these challenges by
providing preliminary assessments of skin conditions rapidly and efficiently. This can
expedite the diagnostic process, reduce waiting times for patients, and enable earlier
interventions when necessary.
3. Access to Specialist Care: Access to dermatological specialists may be limited in certain
geographic areas, particularly in rural or underserved communities. AI-based tools can
extend the reach of dermatological expertise by enabling remote consultations and
31
assessments. Patients can upload images of their skin conditions for evaluation, receiving
expert guidance and recommendations without the need for in-person visits.
4. Support for Healthcare Professionals: Dermatologists and primary care physicians may
benefit from AI-based tools as decision support systems. These tools can assist healthcare
professionals in recognizing potentially concerning lesions, suggesting differential
18
diagnoses, and guiding treatment decisions. By augmenting human expertise with AI-
driven insights, healthcare professionals can make more informed and confident decisions
in clinical practice.
10
5. Research and Innovation: AI-based tools for dermatological diagnosis contribute to
ongoing research and innovation in the field of dermatology. By analyzing large datasets of
dermatological images, researchers can gain deeper insights into the underlying
mechanisms of skin diseases, identify novel diagnostic markers, and develop more effective
treatment strategies. This continuous learning and refinement of AI algorithms drive
advancements in dermatological care and contribute to improved patient outcomes.

12
1.7 Hardware Specifications

HARDWARE SPECIFICATION

Client Side

RAM 512 MB

GPUs NVIDIA GPUs

24
Processor(CPU) Intel Core Duo or Higher or Ryzen

Server Side

RAM 2GB or Higher

Table 1: Hardware Specification

1.8 Software Specifications

SOFTWARE SPECIFICATION
23
Client Side

Operating System Windows 7 and above

Web Browser Google Chrome or any other upto

date Browser
Server Side
28
Operating System Windows 7 and above

Web Technology: Python

Front-end: CSS,Javascript,HTML,BOOTSTRAP

Back-end ML/AI,Angular

Other Technologies Azure cloud,Rest API,Image

processing

Table 2: Software specification

13
 1.9 Scope of Program
2
The scope of the AI-Based Tool for Diagnosis of Dermatological Manifestations encompasses
10
the development of a user-friendly platform enabling individuals to upload images of skin
lesions for preliminary analysis. This platform will feature user authentication and management
functionalities to secure user accounts and facilitate login procedures. Users will be able to
upload images from various devices, with the system supporting common image formats for
secure storage and association with user accounts. Leveraging artificial intelligence algorithms,
the tool will analyze uploaded images to identify potential dermatological conditions,
generating preliminary diagnoses with confidence scores and providing detailed information on
detected conditions, including symptoms and treatment options. Integration with external
databases or knowledge bases will further enhance the tool by offering additional resources
related to specific diagnoses. The user interface will prioritize intuitive design and accessibility,
with performance optimization and robust security measures ensuring efficient processing,
scalability, and compliance with data protection regulations. Comprehensive documentation,
support services, and regular maintenance procedures will complement the deployment of the
tool, ensuring its reliability, usability, and effectiveness in assisting users with dermatological
concerns.

1.10 Objectives of the Project

 Improving Accessibility to Dermatological Care: Utilize AI-powered diagnostic tools to

provide timely and accessible evaluations for individuals, particularly those in underserved or
remote areas with limited access to dermatologists.
10
 Enhancing Diagnostic Accuracy: Develop AI algorithms capable of accurately identifying
and classifying dermatological manifestations based on skin images, patient history, and
clinical data.

 Streamlining Diagnostic Workflow: Integrate AI-based tools into clinical practice to expedite
the diagnostic process, optimize workflow efficiency, and reduce the time required for
diagnosis and treatment planning.

1.11 Features of new System

Several user- friendly coding have also adopted.This project shall prove to be satisfying all
the requirements of the organization.Achieving this objective is difficult using a manual
system as the information is scattered, can be redundant and collecting relevant information
may be very time consuming.All these problems are solved using this project.
Also,we are introducing several services along with the AI diagnosis i.e, Consultations with
corresponding Doctor by booking an appointment with them.
14
 Software Requirement Specification

1. Introduction
11
The AI-Based Tool for Diagnosis of Dermatological Manifestations aims to provide a user-
friendly platform for preliminary diagnosis of skin conditions using artificial intelligence (AI)
algorithms. The software will analyze images of skin lesions uploaded by users and provide
preliminary assessments, assisting healthcare professionals and individuals in identifying
potential dermatological conditions.
2. Purpose
The purpose of this software is to:
 Assist healthcare professionals in diagnosing dermatological conditions accurately and
efficiently.
 Provide individuals with preliminary assessments of skin conditions for informational purposes.
 Enhance accessibility to dermatological care by enabling remote consultations and assessments.
3. Scope
2
The AI-Based Tool for Diagnosis of Dermatological Manifestations will include the following
features:
 User registration and authentication.
10
 Image upload functionality for users to submit images of skin lesions.
 AI-based analysis of uploaded images to provide preliminary diagnoses.
 Display of diagnosis results with relevant information and recommendations.
 Integration with external databases or knowledge bases for additional information on
dermatological conditions.
 Secure storage of user data and images in compliance with privacy regulations.
4. Functional Requirements
19
4.1 User Registration and Authentication
 Users shall be able to create accounts by providing necessary information (e.g., name,
email, password).
37
 The system shall verify user credentials during login to authenticate users securely.
4.2 Image Upload
 Users shall be able to upload images of skin lesions from their devices or cameras.
25
 The system shall support common image formats (e.g., JPEG, PNG) for uploads.
 Uploaded images shall be securely stored and associated with the user's account.
4.3 AI-Based Analysis
 The system shall use AI algorithms to analyze uploaded images and identify potential
15
 dermatological conditions.
 Analysis results shall include preliminary diagnoses, confidence scores, and relevant
information about detected conditions.
 The system shall provide explanations for diagnosis results and recommend further actions
(e.g., seeking medical advice).
4.4 Diagnosis Display
 Diagnosis results shall be displayed to users in a clear and understandable format.
 The system shall present diagnosis details, including detected conditions, associated
symptoms, and treatment options.
17
 Users shall have the option to view additional information or resources related to
specific diagnoses.
30
5. Non-Functional Requirements
5.1 Performance
 The system shall process image uploads and provide diagnosis results within a reasonable
timeframe (e.g., seconds to minutes).
 The AI algorithms used for analysis shall be optimized for efficiency and scalability to
handle varying workload demands.
5.2 Usability
21
 The user interface shall be intuitive and user-friendly, allowing users to navigate the
application easily.
 The system shall provide clear instructions and guidance for users on how to upload images
and interpret diagnosis results.
5.3 Security
15
 User data and uploaded images shall be encrypted during transmission and storage to
protect privacy and confidentiality.
 Access controls and authentication mechanisms shall be implemented to prevent
unauthorized access to user accounts and data.
17
5.4 Compliance
 The system shall comply with relevant data protection regulations (e.g., GDPR) and
healthcare privacy standards.
 Any third-party data sources or APIs used for diagnosis or information retrieval shall be
vetted for compliance with legal and ethical standards.
6. System Architecture
11
The AI-Based Tool for Diagnosis of Dermatological Manifestations shall follow a client-server
architecture, with the following components:
16
 36
 Client-side: Web-based interface for users to interact with the application and upload images.
 Server-side: Backend server responsible for image analysis, diagnosis generation, and
communication with external databases or APIs.
7. Glossary
22
 AI: Artificial Intelligence
 SRS: Software Requirements Specification
 GDPR: General Data Protection Regulation

8. References

 Wang, L., Kim, S., & Chun, S. (2023). Skin disease detection using artificial intelligence.
ResearchGate. [Online].
Available:
https://www.researchgate.net/publication/360743295_Skin_disease_detection_using_artificial_
intelligence
 Zhu, T., Yang, J., Hu, J., & Ji, W. (2023). Artificial intelligence for skin disease diagnosis: A
comprehensive review. medRxiv. [Online].
Available: https://www.medrxiv.org/content/10.1101/2023.03.14.23287243v1.full

17
 DESIGN

3.1 Entity Relationship Diagram

ER model is a conceptual data model that views the real world as entities and relationships. A
basic component of the model is the Entity-Relationship diagram which is used to visually
represent data objects.

3.2 Use Case Diagram

Use case diagrams model behavior within a system and helps the developers
understand of what the user require. The stick man represents what’s called an actor.
Use case diagram consists of use cases and actors and shows the interaction between
the use case and actors.

Figure 1 Use case Diagram Overview

18
3.3 Activity Diagram
An activity diagram is a visual representation of the flow of activities or actions within a
system or process. It uses symbols like rounded rectangles for activities, arrows for control
flow, and decision diamonds for conditional branching. Initial and final nodes denote the start
and end of the process, while swimlanes organize activities by actor or system component.
Activity diagrams help model complex processes, workflows, or software behaviors, aiding in
understanding and optimizing systems.

Figure 2: Activity diagram overview

19
 Database Design

User:

Field Name Data Type Constraint Description

Email int Primary Key Email-id of user

Phone number int Phone number of User

Password varchar Password

Table 3: User’s Database

User’s_Record:

Field Name Data Type Constraint Description

Phone No. int Primary Key Phone no. of User
F_name Varchar First name of User
Skin_type Varchar Skin type of User
Diagnosed disease Varchar Disease diagnosed of
the User
Email Varchar Email id of User

Table 4:User’s Record Database

20
 SCREEN SHOTS

 Homepage:

Figure 2: Home page of Website

 Book an appointment section:

Figure 3: Book an appointment Section

21
  About Skin Sight:

Figure 4: About Us Page

 Check your Skin:

Figure 5: Check your Skin page

 Our Services:

Figure 6: Our Services Page

 Frequently Asked Questions:
22
 Figure 7: FAQs page

 Logo:

Figure 8: Logo

23
 BIBLIOGRAPHY AND REFERENCES

 Unsplash. (n.d.). [Online]. Available: https://unsplash.com/

 [1] DermEngine, "DermDetect Testimonial," [Online]. Available:
https://www.dermengine.com/dermdetect-testimonial. [Accessed: May 1, 2024].

 [2] SkinVision, [Online]. Available: https://www.skinvision.com/. [Accessed: May 1, 2024].

 [3] Miiskin, [Online]. Available: https://miiskin.com/. [Accessed: May 1, 2024].

 [4] VisualDx. [Online]. Available: https://www.visualdx.com/. Accessed: May 1, 2024.

 [5] Ismailpromus. (n.d.). Skin Diseases Image Dataset. Kaggle. [Online]. Available:
https://www.kaggle.com/datasets/ismailpromus/skin-diseases-image-dataset
 [6] Wang, L., Kim, S., & Chun, S. (2023). Skin disease detection using artificial intelligence.
ResearchGate. [Online]. Available:
https://www.researchgate.net/publication/360743295_Skin_disease_detection_using_artifici
al_intelligence
 [7] Zhu, T., Yang, J., Hu, J., & Ji, W. (2023). Artificial intelligence for skin disease
diagnosis: A comprehensive review. medRxiv. [Online]. Available:
https://www.medrxiv.org/content/10.1101/2023.03.14.23287243v1.full
 [8] Teng, Y., Lin, Q., & Li, Y. (2022). Current trends and challenges in AI-based
dermatological diagnosis. BMC Medical Imaging, 22(1), 45. [Online]. Available:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9693628/

24
 Similarity Report ID: oid:27992:59096314

59% Overall Similarity

Top sources found in the following databases:
21% Internet database 49% Publications database
Crossref database Crossref Posted Content database
21% Submitted Works database

TOP SOURCES
The sources with the highest number of matches within the submission. Overlapping sources will not be
displayed.

Stephanie S. Noronha, Mayuri A. Mehta, Dweepna Garg, Ketan Kotecha...

1 30%
Crossref

sih.gov.in
2 3%
Internet

doaj.org
3 3%
Internet

Stephanie S. Noronha, Mayuri A. Mehta, Dweepna Garg, Ketan Kotecha...

4 3%
Crossref

NIIT University on 2023-05-28

5 3%
Submitted works

Pat Vatiwutipong, Sirawich Vachmanus, Thanapon Noraset, Suppawon...

6 3%
Crossref

Yali Nie, Paolo Sommella, Marco Carratu, Matteo Ferro, Mattias O'Nils, ...
7 3%
Crossref

researchgate.net
8 2%
Internet

Sources overview
 Similarity Report ID: oid:27992:59096314

medrxiv.org
9 1%
Internet

University of Wales Institute, Cardiff on 2024-04-29

10 <1%
Submitted works

bannariamman on 2024-04-24
11 <1%
Submitted works

Symbiosis International University on 2013-03-03

12 <1%
Submitted works

ncbi.nlm.nih.gov
13 <1%
Internet

coursehero.com
14 <1%
Internet

Fiji National University on 2023-08-13

15 <1%
Submitted works

itspubs.ucdavis.edu
16 <1%
Internet

Chicago State University on 2023-08-04

17 <1%
Submitted works

RMIT University on 2024-04-29

18 <1%
Submitted works

Middlesex University on 2024-01-14

19 <1%
Submitted works

Zequan Zhao, Yajun Mi, Hafeez Ur Rehman, Enqi Sun, Xia Cao, Ning Wa...
20 <1%
Crossref

Sources overview
 Similarity Report ID: oid:27992:59096314

kkwagh on 2024-05-03
21 <1%
Submitted works

Harrisburg University of Science and Technology on 2023-09-16

22 <1%
Submitted works

Majan College on 2016-01-09

23 <1%
Submitted works

Swinburne University of Technology on 2022-06-06

24 <1%
Submitted works

University of Southern Mississippi on 2024-05-09

25 <1%
Submitted works

dspace.bracu.ac.bd:8080
26 <1%
Internet

fastercapital.com
27 <1%
Internet

gecgudlavalleru.ac.in
28 <1%
Internet

tudr.thapar.edu:8080
29 <1%
Internet

ecs.csun.edu
30 <1%
Internet

University of Greenwich on 2024-04-28

31 <1%
Submitted works

mafiadoc.com
32 <1%
Internet

Sources overview
 Similarity Report ID: oid:27992:59096314

Arshia Eskandari, Mahkame Sharbatdar. "Efficient diagnosis of psorias...

33 <1%
Crossref

Informatics Education Limited on 2013-08-26

34 <1%
Submitted works

Manu Goyal, Thomas Knackstedt, Shaofeng Yan, Saeed Hassanpour. "...

35 <1%
Crossref

University of Greenwich on 2006-05-16

36 <1%
Submitted works

Victoria University on 2023-06-30

37 <1%
Submitted works

Yonsei University on 2023-04-20

38 <1%
Submitted works

ir.juit.ac.in:8080
39 <1%
Internet

University of Greenwich on 2023-12-21

40 <1%
Submitted works

Sources overview