This essay has been submitted by a student. This is not an example of the work written by professional essay writers.

Classification of Malignant Melanoma Using Convolutional Neural Networks

downloadDownload printPrint

Remember! This is just a sample.

You can get your custom paper by one of our expert writers.

Get custom essay

121 writers online

Download PDF

Human Cancer is one of the most dangerous diseases in the world which is mainly caused by genetic instability of multiple molecular alterations. Among many forms of human cancer, skin cancer is the most common one among people. The main objective is to identify skin cancer at an early stage and will study and analyze them through various techniques named as segmentation and feature extraction. Here, mainly focus on malignant melanoma skin cancer, (due to the high concentration of Melanoma- Hier offer skin, in the dermis layer of the skin) detection. This ABCD rule is used with dermoscopy technology for malignant melanoma skin cancer detection. In this project, different steps are performed for melanoma skin lesion characterization including Image Acquisition Technique, pre-processing, segmentation, defining features for skin Feature Selection which determines lesion characterization, and classification methods. The Feature extraction by digital image processing method includes symmetry detection, Border Detection, color, and diameter detection and also used minutiae extraction to extract the texture-based features. Here, proposed the deep Neural Network to classify the benign or malignant stage. 


A skin lesion is a part of human skin that has an abnormal growth or appearance compared to the skin around it. One of the most common causes of a skin lesion is an infection in or on the skin. One example is a wart. Warts are caused by the virus that is transmitted by touch. The main two types of skin lesions are primary and secondary. Primary skin lesions are abnormal skin conditions present at birth or acquired over one’s lifetime. Birthmarks are primary skin lesions. Other types are blisters, Macule, nodule, Papule, Pustule, Rash, wheels, etc. Secondary skin lesions are the results of irritated or manipulated primary skin lesions. For example, if someone scratches a mole until it bleeds, the resulting lesion, a crust, becomes a secondary skin lesion. 

The most common secondary skin lesions include crust, Ulcer, Scale, Scar, and skin atrophy. Skin Lesions are broadly categorized as benign and malignant. Benign skin lesions are those which are non-cancerous whereas malignant skin lesions are cancerous. Common Types of Benign Skin Lesions include Melanocytic nevus, Seborrheic keratosis, Acrochordon, Dermatofibroma, and Cherry angioma. Common Types of Malignant Skin Lesions include Basal cell carcinoma, Melanoma, and Squamous cell carcinoma. Dermoscopy is one of the non-invasive diagnostic techniques for the in vivo observation of pigmented skin lesions, which allows better visualization of the surface and sub-surface structures. Dermoscopy acts as an aid in the diagnosis of skin lesions. The dermoscopic images containing lesions can be given to the CAD system for detection and classification of the skin lesion. Accuracy of the segmentation of skin lesions is a vital process in such a CAD system. Precise segmentation enhances the accuracy of the CAD system which reduces the false classification due to inaccurate segmentation. Pre-processing of images is the first important step in image processing applications. It ensures that the image is enhanced as compared to that of the input image which results in the desired segmentation results. With the fact that the malignant skin lesions are difficult to segment, different pre-processing techniques which can be applied before segmentation of the dermoscopic images containing malignant skin lesions are looked upon. Common pre-processing techniques which include hair removal, median filtering, and contrast stretching in combination with Histogram Specification are applied on dermoscopic images containing malignant skin lesions before segmentation. 

Skin cancer is the most common type of cancer in the world. Over 3.5 million Melanoma cases, Basal Cell Carcinoma, and Squamous Cell Carcinoma are diagnosed every year. This is more than the combined counts of breast cancer, lung cancer, and colon cancers present in the world. In fact, a person gets affected and falls victim to Melanoma every 57 seconds. As it is with every variety of cancer, early screening, and detection of skin cancer is the most hopeful sign of making a full recovery of the patients. Early detection of skin cancer from the population yields a ten-year survival rate of 94%. However, this survival rate drops drastically with cancer progressing and reaching the next stages. Ten-year survival rates result to a meager 15% in the case of Melanoma when it is detected in the final stage. However, early detection of skin cancer from the population is an expensive affair. As the skin lesions look quite similar to each other, it is difficult to identify whether a lesion is benign or malignant. Extensive analysis needs to be performed for the identification of the category of the lesion. Traditionally, an image using a special device which is the dermatoscopy is taken to study the lesion closely. Unfortunately, dermatoscopes are expensive and not widely available to dermatologists. One of the challenges of visual screening is the visual similarities between the skin diseases. In the past few years, significant advancements have taken place in the domain of computer vision. With the advent of new algorithms, it has become possible to differentiate between clinically similar skin conditions. These algorithms do not require the images to be taken from special-purpose devices, such as dermatoscopes, and can be applied on images obtained from general-purpose cameras.

In this study, we present an approach to pre-process and segment lesions, extract features from the segmented lesions and train an artificial neural network, which would then classify the lesions into their respective categories. We consider three variants, each of benign and malignant lesions. The benign lesion category includes Melanocytic Nevi, Seborrheic Keratoses, and Acrochordon, whereas the malignant lesion category comprises of Melanoma, Basal Cell Carcinoma (BCC), and Squamous Cell Carcinoma (SCC). 



Digital image processing means the manipulation of images by the computer, which is a relatively recent development in terms of man’s ancient fascination with visual stimuli. In its short history, it has been applied in practice to every type of image with varying degrees of success. The inherent subjective appeal of the pictorial display attracts attention from the scientists and also from the layman. Digital image processing is similar to other glamour fields, which suffer from myths, misconnections, misunderstandings, and misinformation. It is a vast umbrella under which falls diverse aspects of optics, electronics, mathematics, photography graphics, and computer technology. Several factors combine indicating a lively future for digital image processing. One of the major factors is the declining cost of computer equipment. Several new technological trends promise further to promote digital image processing. These include parallel processing mode, which is made practical by low-cost microprocessors, and the use of charge-coupled devices (CCDs) for digitizing, storage during processing and display, and large low cost of image storage arrays. 


2.3 Image Acquisition: 

Image Acquisition is to acquire a digital image and to do so requires an image sensor and the capability to digitize the signal produced by the sensor. The sensor could be a monochrome or color TV camera that produces an entire image of the problem domain every 130 seconds. The image sensor could also be a line scan camera which produces a single image line at the same time. In that case, the motion of the object past the line. The scanner produces a two-dimensional image. If the outputs of the camera or other imaging sensor are not in digital form, then an analog to digital converter digitizes it. The nature of the sensor and the obtained image are determined by the application. 

2.4 Image Enhancement: 

The simplest and most appealing area of digital image processing is Image enhancement. The main idea behind the enhancement techniques is to bring out the obscured details and highlight certain features of an image. An example of an enhancement technique is increasing the contrast of an image, to make it look better. We should keep in mind that enhancement is a very subjective area of image processing. 

2.5 Image restoration: 

An area that also deals with improving the appearance of an image is Image restoration. Image restoration is objective, which tends to be based on the mathematical or probabilistic models of image degradation whereas enhancement is subjective, which is based on human subjective preferences regarding the constituents of a good enhancement result. For example, contrast stretching is considered as an enhancement technique because it is based primarily on pleasing the aspects it might present to the viewer, whereas removal of an image blur by applying a deblurring function, which is considered a restoration technique. 

2.6 Color image processing: 

The use of color in an image processing is motivated by two principal factors, where the first one is, color is a powerful descriptor that often simplifies object identification and extraction from a scene and the second one is, humans can discern thousands of color shades and intensities, compared to only two dozen shades of gray. This second factor is particularly important while performing manual image analysis. 

2.7 Segmentation: 

Segmentation refers to the partition of an image into its constituent parts or objects. Generally, autonomous segmentation is one of the most difficult techniques in digital image processing. A rugged segmentation procedure performs the process of a long way toward a successful solution of imaging the problems that require objects to be identified individually. On the other hand, weak segmentation algorithms which include erratic segmentation, almost always guarantee eventual failure. In general, the more accurate the segmentation is, the more likely recognition gets to succeed. 

2.8 Image Compression

Digital Image compression is used for reducing the amount of data required to represent a digital image. The underlying basis of the reduction process is to remove the redundant data. From the mathematical viewpoint, this results to a statically uncorrelated data set from a 2D pixel array. The redundancy of data is a mathematically quantifiable entity but not an abstract concept. If n1 and n2 denote the number of information-carrying units in two data different sets that represent the same information, the relative data redundancy of the first data set (which is characterized by n1) can be defined as, Where called as compression ratio. It is defined as = In the image compression technique, the three basic data redundancies that can be identified and exploited are Coding redundancy, interpixel redundancy, and phychovisal redundancy. Image compression gets achieved when one or more of these redundancies are reduced or eliminated. Image compression is mainly used for the transmission of images and storage. One of the image transmission applications is broadcast television, re-mote sensing through satellites, aircraft, radars, teleconferencing, computer communications, and facsimile transmission. Image storage is required commonly for educational and business documents, medical images that arise in computer tomography (CT), magnetic resonance imaging (MRI) and digital radiology, motion pictures, satellite images, weather maps, geological surveys, and so on.

2.9 Image Compression Types 

There are two types of compression techniques in image compression. 

  1. Lossy Image compression 
  2. Lossless Image compression 

2.9.1 Lossy Image compression: 

The Lossy compression technique provides higher levels of data reduction and results in a less than perfect reproduction of the original image. It provides a high compression ratio. The lossy image compression technique is used in the applications such as broadcast television and facsimile transmission, in which a certain amount of error is an acceptable trade-off for increased compression performance. Originally, PGF has been designed to the quick and progressive decode lossy compressed aerial images. A lossy compression mode has been preferred widely because in an application like a terrain explorer texture data (e.g., aerial orthophotos) is usually done by a mid-mapped filter and therefore lossy mapped onto the terrain surface. Additionally, decoding lossy compressed images is usually faster than decoding lossless compressed images. 

2.9.2 Lossless Image compression: 

Lossless Image compression is the most acceptable amount of data reduction. It provides a low compression ratio while compared to the lossy format. In Lossless Image compression the techniques are composed of two relatively independent operations: (1) devising an alternative representation of the image in which its inter-pixel redundancies are reduced and (2) coding the representation to eliminate coding redundancies. The applications of Lossless Image compression are medical imaginary, business documents, and satellite images. 

2.10 Thresholding: 

Thresholding is the simplest method of image segmentation. From a grayscale image, thresholding can be performed to create binary images. The simplest thresholding method is by replacing each pixel in an image with a black pixel if the image intensity is less than some fixed constant or a white pixel if the image intensity is more than that constant. Automatic thresholding will work the best when a good background to foreground contrast ratio is existed, which means the picture must be taken in good lighting conditions with minimal glare. 


3.1 Convolutional Neural Networks 

For addressing this problem, bionic convolutional neural networks are proposed to reduce the number of parameters and adapt the network architecture specifically to vision tasks. Convolutional neural networks are generally composed by a set of layers that can be grouped by their functionalities. 

3.2 Convolution 

Convolution Layer The 2D convolution process is performed on the inputs. The “dot products” between weights and the inputs are “integrated” across the “channels”. Filter weights are shared across receptive fields. The filter has the same number of layers as input volume channels have, and output volume has the same “depth” as the number of filters.

3.2.1 Activation Layer 

Used to increase the non-linearity of the network without affecting receptive fields of convolution layers. Prefer ReLU, results in faster training. Leaky ReLU addresses the problem of vanishing gradient. 

3.2.2 Softmax 

It is a special kind of activation layer, usually at the end of FC layer outputs which an be viewed as a fancy normalizer (Normalized exponential function), producing a discrete probability distribution vector. It is very convenient when it is combined with cross-entropy loss. 

3.2.3 Pooling Layer 

Convolutional layers provide activation maps where the pooling layer applies non-linear down-sampling of activation maps. In Pooling, the trend is to use a smaller filter size. 

3.2.4 FC Layer 

It is a regular neural network that can be viewed as the final learning phase, which maps extracted visual features to desired outputs. Usually adaptive to the classification or encoding tasks. Common output is a vector, which is then passed through soft-max as input to represent the confidence of classification. These outputs can also be used as “bottleneck”. 


Below are the results obtained when testing a benign and malignant skin image. Fig 4.1 and Fig 4.2 are the outputs obtained for the benign and malignant skin images respectively. Fig 4.1 Benign Skin Output Fig 4.2 Malignant Skin Output 4. 


To reduce skin cancer, people must get the proper in-formation that needed to be informed including the choices about sun protection. The policies must support these efforts and the youth must be protected from the harms of indoor tanning and adequate investments are needed to be made in the skin cancer research and surveillance. Achieving those goals would not be a small task. It requires dedication, ingenuity, skill, and the concerted efforts of many partners in prevention across many different sectors. Many of those partners are already enthusiastically involved, but greater coordination and support are needed to increase the reach of their efforts. The next steps are the strategies outlined in this paper. We must act with urgency to stop the ever-increasing scale of skin cancers in the world. 


  1. Yogesh Angal ; Anita Gade, “Development of library management robotic system”, 2017 Interna-tional Conference on Data Management, Analytics and Innovation (ICDMAI) 
  2. Benjamin Meunier, “Library Technology and Innovation as A Force For Public Good A Case Study From UCL Library Services”, 2018 5th International Symposium on Emerging Trends and Technologies in Libraries and Information Services (ETTLIS)  
  3. Jin Feng Zhang ; Chang Ji Wen, “The universi-ty library management system based on radio frequency identification”, 2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI) 
  4. Felecia ; S. Halim ; D. Wulandari,`Library facility layout design for digital native generation`, 2017 IEEE International Conference on Industrial Engineering and Engineering Management(IEEM) 
  5. Lou, Z. et al. (2015), “Color Constancy by Deep Learn-ing”, British Machine Vision Conference. DOI:10.5244C.29.76 
  6. Codella, N. et al. (2018), “Skin Lesion Analysis Toward Melanoma Detection 2018: A Challenge Hosted by the International Skin Imaging Collabora-tion (ISIC)”. arXiv:1902.03368 
  7. Finlayson, G.D. et al. (1998), ‘Comprehensive Colour Normalization’, Proc. European Conf. on Computer Vison (ECCV), vol. 1, pp. 475–490.
  8. Ronneberger, O., Fischer, P., Brox, T., (2015), ‘U-Net: Convolutional Networks for Biomedical Image Seg-mentation’, Medical Image Computing and Computer-Assisted Intervention, vol. 9351, Springer. 
  9. Jegou, S. et al. (2017), ‘The One Hundred Layers Ti-ramisu: Fully Convolutional DenseNets for Semantic Segmentation’, IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). DOI:10.1109cvprw.2017.156 
  10. He, K., Girshick, R., Dollár, P. (2018), ‘Rethinking Im-ageNet Pre-Training’. arXiv:1811.08883[cs.CV]

Remember: This is just a sample from a fellow student.

Your time is important. Let us write you an essay from scratch

experts 450+ experts on 30 subjects ready to help you just now

delivery Starting from 3 hours delivery

Find Free Essays

We provide you with original essay samples, perfect formatting and styling

Cite this Essay

To export a reference to this article please select a referencing style below:

Classification of Malignant Melanoma Using Convolutional Neural Networks. (2022, May 24). GradesFixer. Retrieved June 29, 2022, from
“Classification of Malignant Melanoma Using Convolutional Neural Networks.” GradesFixer, 24 May 2022,
Classification of Malignant Melanoma Using Convolutional Neural Networks. [online]. Available at: <> [Accessed 29 Jun. 2022].
Classification of Malignant Melanoma Using Convolutional Neural Networks [Internet]. GradesFixer. 2022 May 24 [cited 2022 Jun 29]. Available from:
copy to clipboard

Sorry, copying is not allowed on our website. If you’d like this or any other sample, we’ll happily email it to you.

    By clicking “Send”, you agree to our Terms of service and Privacy statement. We will occasionally send you account related emails.


    Attention! This essay is not unique. You can get a 100% Plagiarism-FREE one in 30 sec

    Receive a 100% plagiarism-free essay on your email just for $4.99
    get unique paper
    *Public papers are open and may contain not unique content
    download public sample

    Sorry, we could not paraphrase this essay. Our professional writers can rewrite it and get you a unique paper.



    Please check your inbox.

    Want us to write one just for you? We can custom edit this essay into an original, 100% plagiarism free essay.

    thanks-icon Order now

    Hi there!

    Are you interested in getting a customized paper?

    Check it out!
    Don't use plagiarized sources. Get your custom essay. Get custom paper

    Haven't found the right essay?

    Get an expert to write you the one you need!


    Professional writers and researchers


    Sources and citation are provided


    3 hour delivery