Final Report
Final Report
ENGINEERING
A MAJOR PROJECT
REPORT ON
BACHELOR
OF
TECHNOLOGY IN
K Bhavana
Certified that the project work entitled “Automated skin cancer detection using
deep learning techniques and image analysis” carried out under my guidance by
Pranay Vernekar (R21EN199), Priyanka S (R21EB020), K Bhavana (R21EB033)
a bonafide student of REVA University during the academic year 2024-2025, is
submitting the project report in partial fulfillment for the award of Bachelor of
Technology in Electronics and Communication Engineering during the academic
year 2024–25. The project report has been tested for plagiarism and has passed the
plagiarism test with a similarity score of less than 25%. The project report has
been approved as it satisfies the academic requirements regarding the Project work
prescribed for the said Degree.
External Examiner
Name of the Examiner with affiliation Signature with Date
1.
2.
Contents
Contents ii
List of Figures iv
1 Introduction 1
1.1 Background.............................................................................................................. 1
1.2 Problem Definition :............................................................................................... 2
2 Literature Review 4
3 Proposed work 9
3.1 Methodology …………………………………………............................... 9
3.2 Proposed workflow ..……………………………………........................... 9
3.2.1 CNN AND MOBILENET V2 MODEL TRAINING ………………..... 10
3.2.2 Code snippets (Keras)…………………………………………...................... 11
3.2.3 Dataset used …………………………………………............................ 11
3.2.4 Skin cancer types………………………………………………......................... 12
4 Future Scope 15
References 17
iii
List of Figures
iv
Chapter 1
Introduction
1.1 Background
Skin cancer is one of the most active types of cancer in the present decade [1].As
the skin is the body’s largest organ, the point of considering skin cancer as the
most common type of cancer among humans is understandable [2]. It is generally
classified into two major categories: melanoma and non melanoma skin cancer
[3]. Melanoma is a hazardous, rare, and deadly type of skin cancer. According to
statistics from the American Cancer Society, melanoma skin cancer cases are only
1% of total cases, but they result in a higher death rate [4]. Melanoma develops in
cells called melanocytes. It starts when healthy melanocytes begin to grow out of
control, creating a cancerous tumor. It can affect any area of the human body. It
usually appears on the areas exposed to sunrays, such as on the hands, face, neck,
lips, etc. Melanoma type of cancers can only be cured if diagnosed early; otherwise,
they spread to other body parts and lead to the victim’s painful death [5]. There as
various types of melanoma skin cancer such as nodular melanoma, superficial
spreading melanoma, acrallentiginous, and lentigomaligna [3].The majority of
cancer cases lie under the umbrella of non melanoma categories, such as basal cell
carcinoma (BCC), squamous cell carcinoma (SCC), and sebaceous gland
carcinoma(SGC). BCC, SGC, and SCC are formed in the middle and upper layers
of the epidermis, respectively. These cancer cells have a low tendency of
spreading to other body parts. Non melanoma cancers are easily treated as
compared with melanoma cancers. Therefore, the critical factor in skin cancer
treatment is early diagnosis [6]. Doctors ordinarily use the biopsy method for skin
cancer detection. This procedure removes a sample from a suspected skin lesion
for medical examination to determine whether it is cancerous or not. This process
is painful, slow, and time-consuming. Computer-based technology provides a
comfortable,
1
Chapter 1 Introduction 1.2. Problem Definition
less expensive, and speedy diagnosis of skin cancer symptoms. In order to
examine the skin cancer symptoms, whether they represent melanoma or non
melanoma, multiple techniques, non invasive in nature, are proposed. The general
procedure followed in skin cancer detection is acquiring the image, preprocessing,
segmenting the acquired preprocessed image, extracting the desired feature, and
classifying it, represented in figure. In 2018, there were approximately 1 million
non-melanoma skin cancers and 288,000 malignant melanoma (MM) cancers
reported worldwide [1]. Because of an aging population and limited health-care
resources, accurate diagnosis and detection are essential components of a
generalized skin cancer prevention strategy. The impact of immune therapies on
survival and cost, places additional strain over burdened healthcare system, raising
concerns about financial sustainability [2]. Dermatologists and general
practitioners face challenges in detecting skin cancer, particularly MM, early on.
Dermoscopy is considered the standard of care [3], but objective tests show that
dermatologists have a diagnostic sensitivity of 40% for MM detection[4] due to
the complexity of visual inputs embedded in a dermoscopy image [5].General
practitioners appear to benefit from the use Skin cancer [7]. Nonetheless,
laboratory studies reported clinical sensitivity ranging from 29% to 87% [8],a
discrepancy that could be attributed to the quality of the dataset input, classifying
the technology as experimental [9]. A recent prospective clinical observational
study [10] described a two-step approach for improving detection accuracy by
adding a second layer of sonification (visual data converted into sounds) to a DL
classifier. This dual DL used an advanced dermoscope, a relatively expensive
device, and a technique that was heavily reliant on physician experience [4],
making it less suitable for general primary care physician use.
Literature Review
4
Chapter 2 Literature Review Chapter 2 Literature Review
Proposed work
3.1 Methodology
Role of Deep Learning: Deep learning plays a pivotal role in the detection of skin
cancer by revolutionizing the accuracy and efficiency of diagnosis. Convolutional
Neural Networks (CNNs), a type of deep learning algorithm, can be trained on large
datasets of dermoscopic images to automatically detect and classify skin lesions into
benign or malignant categories. These models can learn to identify subtle patterns and
features characteristic of skin cancer, such as asymmetry, border irregularity, color
variation, and diameter. By leveraging transfer learning, data augmentation, and
ensemble methods, deep learning models can achieve exceptional accuracy (up to
95%) and sensitivity (up to 97%) in detecting melanoma, the deadliest form of skin
cancer. Moreover, deep learning-based systems can assist dermatologists in early
detection,
9
Reduce false positives and negatives, and improve patient outcomes. Additionally, AI-
powered skin analysis apps and telemedicine platforms can increase accessibility to
skin cancer screening, especially in under-served populations. Overall, deep learning
has transformed the field of skin cancer detection, enabling faster, more accurate, and
more accessible diagnosis, and ultimately saving lives.
2. Pre processing: Images are resized to uniform dimensions. Noise reduction tech-
niques like Gaussian filters are applied. Color space conversion from RGB to HSV
or LAB is performed. Image normalization enhances contrast. Artifact removal
improves image quality. Preprocessing enhances model performance.
4. Feature Extraction: Shape features like area and perimeter are extracted. Color
features like mean and standard deviation are calculated. Texture features using
Gabor filters and Local Binary Patterns are extracted. Feature extraction reduces
data dimensionality. Relevant features improve model performance. Feature
selection enhances classification accuracy.
5. Classification: Deep learning models like CNN and ResNet are trained. Traditional
machine learning algorithms like SVM and Random Forest are employed. Models
are trained on extracted features. Hyperparameter tuning optimizes performance.
Classification accuracy is evaluated. Models are fine-tuned for improved
performance.
9
Chapter 3 Proposed work 3.2. Proposed workflow
6. Classification: Deep learning models like CNN and ResNet are trained. Traditional
machine learning algorithms like SVM and Random Forest are employed. Models
are trained on extracted features. Hyperparameter tuning optimizes performance.
Classification accuracy is evaluated. Models are fine-tuned for improved
performance.
7. Prediction: Trained models predict skin cancer likelihood. Prediction probabilities
are calculated. Visualizations like heat maps and gradient-weighted class activation
mapping provide insights. Prediction accuracy is evaluated. False positives and
negatives are minimized. Model performance is continuously monitored.
CNN Model Training: Train a Convolutional Neural Network (CNN) for skin
cancer detection using dermoscopic images. The CNN consists of 2-3
convolutional layers with max pooling, followed by flatten, dense, and dropout
layers. Hyperparameters include Adam optimizer, categorical entropy loss, 0.001
learning rate, 32 batchsize, and 10 epochs. Preprocess images by resizing,
normalizing, and augmenting. Achieve high accuracy (greater than 90%) and
sensitivity (greater than 95%) for effective skin cancer detection.
MobileNetV2 Model Training: Train a MobileNetV2 model for skin cancer
Detection by fine-tuning pre-trained weights on dermoscopic images. Add custom
classification layers and adjust hyperparameters: Adamoptimizer, categorical cross-
entropy loss, 0.0001 learning rate, 32 batch size, and 5 epochs. MobileNetV2’s
efficiency and accuracy make it ideal for real-time skin cancer detection applications.
Achieve high performance with minimal computational resources.
Keypoints include:
• Available on Kaggle
Expected Outcome
Accurate Detection: The skin cancer detection project aims to achieve accurate
detection of skin cancer from dermatoscopic images. This outcome reduces false
positives and false negatives, improving sensitivity and specificity. Enhanced
diagnostic accuracy supports early diagnosis and treatment. The project lays the
foundation for automated diagnosis, revolutionizing skin cancer detection. By
leveraging AI-powered algorithms, clinicians can make informed decisions.
Early Diagnosis: Early detection of skin cancer increases treatment effective-
ness, improving patient survival rates and reducing mortality. Timely
interventions and treatments enhance quality of life for patients. The project
facilitates early diagnosis, reducing healthcare costs associated with late-stage
treatment. Patients benefit from prompt diagnosis, saving lives. This outcome
underscores the project’s life-saving potential.
Reduced Healthcare Costs: The project reduces health care costs by minimizing
unnecessary biopsies, referrals, and treatments. Optimized resource allocation
streamlines clinical workflows and decision-making. Enhanced diagnostic accuracy
decreases unnecessary medications and procedures. The project promotes cost-
effective healthcare solutions. By reducing waste, the project enhances over-all
healthcare efficiency.
Improved Patient Outcomes: Improved patient outcomes result from timely di-
agnosis and treatment. Patient satisfaction and experience improve significantly.
Reduced anxiety and stress associated with diagnosis enable better disease man-
agement. The project promotes patient engagement and education. Patients benefit
from personalized treatment plans, enhancing overall well-being.
Enhanced Clinical Decision Support: The project enhances clinical decision support
for dermatologists. Objective, data-driven insights facilitate differential diagnosis
and treatment planning. Clinicians benefit from AI-assisted diagnosis,
streamlining workflows. The project augments clinician expertise, promoting
high-quality care. By leveraging AI, clinicians make informed decisions.
Future Scope
• Integration with Clinical Systems: Incorporate the detection system into hospital
information systems and electronic health records for seamless integration with
existing workflows.
15
References
[1] A. A. Agrahari, Pradhumn and N. Subhashini, “Skin cancer detection using deep
learning,” Futuristic Communication and Network Technologies: Select Proceedings of
VICFCNT 2020. Springer Singapore, 2022.
[2] H. Nahata and S. P. Singh, “Deep learning solutions for skin cancer detection and
diagnosis,” Machine learning with health care perspective: machine learning and
healthcare (2020), pp. 159–182, 2020.
[3] W. Gouda, “Detection of skin cancer based on skin lesion images using deep learn-
ing,” vol. 10, 2022.
[5] W. Gouda, “Detection of skin cancer based on skin lesion images using deep learn-
ing,” vol. 10, 2022.
[6] Y. Li, “Skin cancer detection and tracking using data synthesis and deep learning,”
Workshops at the thirty-first AAAI conference on artificial intelligence, 2017.
[7] M. A. Hossin, “Melanoma skin cancer detection using deep learning and advanced
regularizer,” 2020.
[8] I. Kousis, “Deep learning methods for accurate skin cancer recognition and mobile
application,” Electronics 11.9 (2022).
[9] S. S. Han, “Keratinocytic skin cancer detection on the face using region-based
convolutional neural network,” JAMA dermatology 156.1, pp. 29–37, 2020.
[10] A. Ameri, “deep learning approach to skin cancer detection in dermoscopy images,”
Journal of biomedical physics engineering 10.6, 2020
18
SchoolofECE,RevaUniversity 16