S.A. Engineering College,Chennai-77.

Department of Computer Science and Engineering,

DERMATOLOGY : AN INTELLIGENT SYSTEM FOR SKIN

LESSION RECOGNITION

TEAM MEMBERS
BELRIN ANIS ABRAHAM G (111920CS01013)

KARTHIK M (111920CS01303)

SUBASH D (111920CS010310)

GUIDED BY
Mr. BALAKRISHNAN C M.E., (Ph.D).,
 ABSTRACT

• Skin cancer is an abnormal growth of skin cells. It generally develops in areas that are exposed to the sun, but it
can also form in places that don’t normally get sun exposure.
• Skin cancers aren’t all identical, and they may not cause many symptoms. Still, unusual changes to your skin can
be a warning sign for the different types of cancer.
• Being alert for changes to your skin may help you get a diagnosis earlier. Accurate and precise diagnosis of
diseases has been a significant challenge and he recent advances in computer vision made possible by deep
learning has paved the way for disease diagnosis for skin cancer.
• It described the innovative solution that provides efficient disease detection and deep learning with convolutional
neural networks (CNNs) has achieved great success in the classification of various skin cancer diseases.
• A variety of neuron-wise and layer-wise visualization methods were applied using a CNN, trained with a publicly
available skin cancer disease given image dataset.
• So, it observed that neural networks can capture the colors and textures of lesions specific to respective diseases
upon diagnosis, which resembles human decision-making. And this model to deploy Django web framework.
 INTRODUCTION

• DATA SCIENCE:

Data science is an interdisciplinary field that uses scientific methods, processes, algorithms and systems to
extract knowledge and insights from structured and unstructured data, and apply knowledge and actionable insights
from data across a broad range of application domains.
• ARTIFICIAL INTELLIGENCE:

Artificial intelligence (AI) refers to the simulation of human intelligence in machines that are programmed
to think like humans and mimic their actions. The term may also be applied to any machine that exhibits traits
associated with a human mind such as learning and problem-solving.
• MACHINE LEARNING:
Machine learning is to predict the future from past data. Machine learning (ML) is a type of artificial intelligence
(AI) that provides computers with the ability to learn without being explicitly programmed. Machine learning focuses
on the development of Computer Programs that can change when exposed to new data and the basics of Machine
Learning, implementation of a simple machine learning algorithm using python. Process of training and prediction
involves use of specialized algorithms.

• DEEP LEARNING:
Deep learning is a branch of machine learning which is completely based on artificial neural networks, as neural
network is going to mimic the human brain so deep learning is also a kind of mimic of human brain. It’s on hype
nowadays because earlier we did not have that much processing power and a lot of data.
 REQUIREMENTS

1. Software Requirements:
Operating System : Windows / Linux
Simulation Tool : Visual Studio Code

2. Hardware requirements:
Processor : Pentium IV/III
Hard disk : minimum 80 GB
RAM : minimum 2 GB
 EXISTING WORK

• In this section, the Metadata Processing Block (MetaBlock), an attention-based mechanism approach that uses
the metadata to enhance the feature maps extracted from images in order to improve data classification.
• In order to evaluate the MetaBlock performance, it to five different Convolutional Neural Network (CNN)
trained on two different skin lesion dataset: ISIC 2019 and PAD-UFES-20.
• Also, the obtained results with the models without using metadata, using a concatenation baseline, and the
MetaNet. In brief, our model achieves the best performance, in terms of balanced accuracy, for 3 out of 5 CNN
models for both datasets.
• Also, it is more stable than the other two combination methods and the statistical test indicates it generally
performs better for both datasets Computer-aided skin cancer classification systems built with deepneural
networks usually yield predictions based only on images of skin lesions.
• Despite presenting promising results, it is possible to achieve higher performance by taking into account patient
demographics, which are important clues that human experts consider during skin lesion screening.
LIMITATIONS:
● Overreliance on Image Features
● Limited Contextual Information
● Potential Bias in Predictions
● Inability to Generalize
● Missed Opportunities for Personalized Medicine
 PROPOSED SYSTEM

• The design deep learning technique so that a person with lesser expertise in software should also be able to
use it easily.
• It proposed system to predicting skin cancer. It explains about the experimental analysis of our methodology.
Samples of images are collected that comprised of different skin cancer.
• Different number of images is collected for each disease that was classified into database images and input
images.
• The primary attributes of the image are relied upon the shape and texture oriented features. An efficient
disease detection and deep learning with convolutional neural networks (CNNs) has achieved great success in
the classification of various skin cancer diseases.
• A variety of neuron-wise and layer-wise visualization methods were applied using a CNN, trained with a
publicly available skin cancer disease given image dataset.
• The sample screenshots displays the skin cancer disease detection using color based classification model. And
to deploy this model web application.
ADVANTAGES:
• Increasing throughput & reducing subjectiveness arising from human experts in detecting the skin cancer.
• It is essential to detect a particular disease. In our country many farmers are not so educated to get correct
information about all diseases.
• As the volume of skin cancer cases continues to rise globally, there is a growing need for scalable diagnostic
solutions.
• Unlike human experts, automated classification systems are not subject to variations in expertise or subjective
interpretation.
 PROBLEM STATEMENT

• Data Collection: Gather a dataset of skin lesion images with corresponding labels indicating the type of skin
cancer (e.g., melanoma, basal cell carcinoma, benign nevus).

• Data Preprocessing: Preprocess the images, which may include resizing, normalization, and data
augmentation (e.g., rotation, flipping, zooming) to increase the diversity of the training set and improve model
generalization.

• Model Architecture: Design a CNN architecture using TensorFlow and Keras. This typically consists of
convolutional layers followed by pooling layers for feature extraction, and then fully connected layers for
classification.
• Model Training: Split the dataset into training, validation, and testing sets. Train the CNN model on the
training set, validate its performance on the validation set, and tune hyperparameters accordingly to prevent
overfitting.

• Evaluation: Evaluate the trained model's performance on the testing set using metrics such as accuracy,
precision, recall, and F1-score. Additionally, visualize the model's predictions and examine any
misclassifications to identify areas for improvement.

• Deployment: Once satisfied with the model's performance, deploy it for real-world use. This could involve
integrating it into a web application, mobile app, or any other platform where skin cancer classification is
needed.
 METHODOLOGY

 Data Acquisition:

Obtain a dataset of skin lesion images with corresponding labels indicating the type of skin cancer or whether the lesion is benign or
malignant. Datasets like ISIC (International Skin Imaging Collaboration) or HAM 10000 (Human Against Machine with 10000
training images) are commonly used for this purpose.

 Data Preprocessing:

● Resize images to a uniform size, commonly used sizes are 224x224 or 299x299 pixels.
● Normalize pixel values to the range [0, 1].
● Split the dataset into training, validation, and testing sets.
● Perform data augmentation to increase the diversity of the training set. Common augmentation techniques include rotation,
flipping, zooming, and shearing.

 Model Architecture Selection:

● Choose a suitable pre-trained convolutional neural network (CNN) architecture such as VGG16, VGG19, ResNet50,
InceptionV3, or Efficient Net.
● Optionally, fine-tune the pre-trained model to better suit the skin cancer classification task. This involves unfreezing some of the
layers of the pre-trained model and retraining them along with additional layers added for classification.
 Model Building:

● Build the CNN model

● If using transfer learning, load the pre-trained model and add layers for classification on top.
● Add a global average pooling layer to reduce the spatial dimensions of the feature maps.
● Add one or more fully connected (dense) layers for classification.
● Use dropout regularization to reduce overfitting.
● Choose an appropriate activation function for the output layer

 Model Compilation:

● Compile the model using an appropriate optimizer and a suitable loss function

 Model Training:

● Train the model on the training set.

● Monitor the model's performance on the validation set and adjust hyperparameters if necessary.
• Use early stopping to prevent overfitting by monitoring a validation metric

 Model Evaluation:

● Evaluate the trained model on the testing set to assess its performance.
● Calculate metrics such as accuracy, precision, recall, and F1-score.
● Generate a confusion matrix to visualize the model's performance across different classes.
● Analyze any misclassifications to identify patterns and potential areas for improvement.

 Deployment:

● Deploy the trained model for real-world use, such as integrating it into a web application, mobile app, or healthcare
system.
● Monitor the model's performance in production and update it periodically as needed.
ARCHITECTURE DIAGRAM

Fig 1: Architecture Diagram

 LITERATURE SURVEY

Paper No TITLE Author Name Abstract Algorithm

Skin cancer, particularly the malignant melanoma

The Power of Generative subtype, is widely recognized as a highly lethal
AI to Augment for form of cancer characterized by abnormal
01 Enhanced Skin Cancer Shafiq Ur Rehman melanocyte cell growth. However, diagnosing and CNN , GAN
Classification: A Deep classifying skin lesions, as well as automatically
Learning Approach recognizing malignant tumors from dermoscopy
images, present significant challenges.

Deep Skin: A Deep

Learning Approach for Skin cancer is one of the most rapidly spreading
Skin Cancer Classification illnesses in the world and because of the limited
02 Francesco Flammini resources available. Early detection of skin cancer CNN , Deep Learning
is crucial accurate diagnosis of skin cancer
identification for preventive approach in general.

An Interpretable Skin Skin cancer is a prevalent form of malignancy

Cancer Classification globally, and its early and accurate diagnosis is
Using Optimized critical for patient survival. Clinical evaluation of
03 Convolutional Neural Jungpil Shin skin lesions is essential, but it faces challenges CNN , Grad-CAM ,
Network for a Smart such as long waiting times and subjective Explainable AI
Healthcare System interpretations.
 LITERATURE SURVEY

Paper No TITLE Author Name Abstract Algorithm

Skin cancer is one of the most threatening cancers,

Recent Advances in which spreads to the other parts of the body if not
Diagnosis of Skin Lesions caught and treated early. During the last few years,
04 Using Dermoscopic Yali Nie the integration of deep learning into skin cancer Dermoscopy images,
Images Based on Deep has been a milestone in health care, and Deep learning
Learning dermoscopic images are right at the center of this
revolution.

The skin lesion types result in delayed diagnosis

A Deep Learning due to high similarity in early stages of the skin
Approach Based on cancer. In this regard, deep learning algorithms are
05 Explainable Artificial Douhadji Abalo well-recognized solutions; however, these black AI, Deep Learning
Intelligence for Skin box approaches result in lack of trust as
Lesion Classification dermatologists are unable to interpret and validate
the decisions made by the models.

The complex detection background and lesion

Automatic Skin Cancer features make the automatic detection of
Detection in Dermoscopy dermoscopy image lesions face many challenges.
06 Images Based on Kun Ding The previous solutions mainly focus on using CNN
Ensemble Lightweight larger and more complex models to improve the
Deep Learning Network accuracy of detection, there is a lack of research
on significant intraclass differences and inter-class
similarity of lesion features.
 LITERATURE SURVEY

Paper No TITLE Author Name Abstract Algorithm

Skin cancer is one of the top three perilous types

An enhanced technique of of cancer caused by damaged DNA that can cause
skin cancer classification death. This damaged DNA begins cells to grow
07 using deep convolutional Md Shahin Ali uncontrollably and nowadays it is getting CNN, RNN, LSTM
neural network with increased speedily. There exist some researches
transfer for the computerized analysis of malignancy in
skin lesion images.

Cosmetics consumers need to be aware of their

Human skin type skin type before purchasing products. Identifying
classification using image skin types can be challenging, especially when
08 processing and deep Sirawit Saiwaeo they vary from oily to dry in different areas, with
learning approaches skin specialist providing more accurate results. In CNN , Deep Learning
recent years, artificial intelligence and machine
learning have been utilized across various fields,
including medicine, to assist in identifying and
predicting situations.

Skin cancer detection

using dual optimization Skin cancer is the most perilous kind of cancer,
based deep learning which is a most important public health problem.
09 network E.Gomathi Skin cancer can be prevented and treated more CNN
effectively if malignant lesions are identified in its
early stages.
 LITERATURE SURVEY

Paper No TITLE Author Name Abstract Algorithm

The most common cancer in the world is skin

Segmentation of skin cancer. In recent years, one of the most important
cancer using Fuzzy U- challenges to public health has been melanoma,
10 network via deep learning A. Bindhu the most dangerous type of skin cancer. In this FRCNN, LinkNet-B7
paper, a novel MFO-Fuzzy U net has been
proposed to segmentation process to extract the
affected area in skin cancer image.

Interpretable deep learning We apply for the first-time interpretable deep

systems for multi-class learning methods simultaneously to the most
11 segmentation and Simon M.Thomas common skin cancers in a histological setting. As
classification of non- these three cancer types constitute more than 90% Deep Learning
melanoma skin cancer of diagnoses, we demonstrate that the majority of
dermatopathology work is amenable to automatic
machine analysis.

Skin cancer is an abnormal growth of skin cells. It

Skin Cancer Classification generally develops in areas that are exposed to the
12 using Tensorflow and Saranya S sun, but it can also form in places that don’t Deep Learning
Keras normally get sun exposure. Skin cancers aren’t all
identical, and they may not cause many
symptoms.
 LITERATURE SURVEY

Paper No TITLE Author Name Abstract Algorithm

Skin cancer is the most common type of cancer

Skin cancer classification worldwide, affecting a large population recently.
using explainable artificial To date, various machine learning techniques
13 intelligence on pre- Tarek Khater exploiting skin images have been applied directly CNN, ML Algorithms
extracted image features to skin cancer classification, showing promising
results in improving diagnostic accuracy.

For various forms of skin lesion, many different

Multi-features extraction feature extraction methods have been investigated
14 based on deep learning for Simon M.Thomas so far. Indeed, feature extraction is a crucial step CNN, GAN, LSTM,
skin lesion classification in machine learning processes. In general, we can RNN
distinct handcrafted and deep learning features.

In image classification, deep learning plays a

Synthetic medical image crucial role. Its branch is now extending to
augmentation: a GAN- medical image segmentation and classification,
15 based approach for V.Nirmala which has a wide breakthrough in different tasks CNN, GAN
melanoma skin lesion on computer vision. Primarily, the creation of
classification with deep large-scale annotated datasets for such
learning categorization remains a significant issue
Partial Demo of project
Partial Demo of project
 LIST OF MODULES

1. Manual Net

2. AlexNet

3. LeNet

4. Deploy
 REFERENCES

1.Kittler, H.; Pehamberger, H.; Wolff, K.; Binder, M. Diagnostic accuracy of dermoscopy. Lancet Oncol. 2002, 3, 159–165.

2.Rosendahl, C.; Tschandl, P.;Cameron, A.; Kittler, H. Diagnostic accuracy of dermatoscopy for melanocytic and nonmelanocytic pigmented lesions.

3. J. Am. Acad. Dermatol. 2011, 64, 1068–1073.

4.Jalalian, A.; Mashohor, S.; Mahmud, R.; Karasfi, B.; Saripan, M.I.B.; Ramli, A.R.B. Foundation and methodologies in computer-aided diagnosis

systems for breast cancer detection. EXCLI J. 2017, 16, 113.

5.Fan, H.; Xie, F.; Li, Y.; Jiang, Z.; Liu, J. Automatic segmentation of dermoscopy images using saliency combined with Otsu threshold. Comput. Biol.

Med. 2017, 85, 75–85.

 REFERENCES

6. Hasan, M.K.; Dahal, L.; Samarakoon, P.N.; Tushar, F.I.; Martí, R. DSNet: Automatic dermoscopic skin lesion segmentation. Comput. Biol.

Med. 2020, 120, 103738.

7. Korotkov, K.; Garcia, R. Computerized analysis of pigmented skin lesions: A review. Artif. Intell. Med. 2012, 56, 69–90.

8. Hasan, M.K.; Elahi, M.T.E.; Alam, M.A.; Jawad, M.T.; Martí, R. DermoExpert: Skin lesion classification using a hybrid convolutional neural

network through segmentation, transfer learning, and augmentation. Inform. Med. Unlocked 2022, 28, 100819.

9. Mishra, N.K.; Celebi, M.E. An overview of melanoma detection in dermoscopy images using image processing and machine learning. arXiv 2016,

arXiv:1601.07843.

10. Pacheco, A.G.; Krohling, R.A. Recent advances in deep learning applied to skin cancer detection. arXiv 2019, arXiv:1912.03280.