Detection Of Melanoma Using Deep Learning by CNN Model

Abstract

Malignant Melanoma is a disease that death rate is quite high among the skin cancer types. It is very important to do the correct diagnosis of this cancer type whose danger level is high. Asymmetric shape, heterogeneous color, bigger diameter than 6 mm, and untidy borders that are used to diagnose melanoma by the dermatologists visually are important parameters. Conventional diagnosis method for skin cancer detection is the Biopsy method. lt is done by removing or scraping off skin and that sample undergoes a series of laboratory testing. lt is painful and time-consuming one. So for helping patients for detecting cancer computer-based detection uses image processing techniques and deep learning (CNN) algorithms.

I. Introduction

Skin cancer is the most common and prevalent type of cancer over the world. Over 3.5 million cases of Melanoma, 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. In fact, a person 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. Early detection of skin cancer yields a ten-year survival rate of 94%. However, this survival rate drops drastically as cancer progresses and reaches the next stages. Ten-year survival rates come to 15% in the case of Melanoma when it is detected in the final stage. However, early detection of skin cancer is an expensive affair. As skin lesions look quite similar to each other, it is difficult to determine whether a lesion is benign or malignant. Extensive analysis needs to be performed to identify the category of the lesion. Traditionally, an image using a special device, known as a dermatoscope, is taken to study the lesion closely. Unfortunately, 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.

Dermatoscopes are expensive and not widely available with dermatologists. One of the challenges of visual screening is the visual similarity between skin diseases. In the last few years, significant advancements have taken place in the domain of computer vision. With the advent 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. To detect unusual mole characteristics that could indicate skin cancer or melanoma, a search for moles with irregular borders, shapes, colors, and moles with more than 6mm in diameter is done. To assess the soreness of the skin and classify it either as melanoma or benign, numerous techniques, including genetic algorithms, artificial neural networks (ANNs), CNNs, ABCDE rule, and support vector machines (SVMs), have been proposed. . All these techniques have been verified as cost-effective, highly efficient, and less painful than conventional medical techniques. However, in many computer vision problems, it becomes undeniable that both CNNs and deep learning are the technique of choice. The CNN is designed in such a manner to differentiate melanoma from solar lentigo and seborrheic keratosis, which is often difficult. Deep learning is a class of machine learning algorithms that uses multiple layers to progressively extract higher-level features from the raw input. Convolutional Neural Network (CNN) is a deep learning algorithm used in the classification and identification of the disease. CNN is one of the main categories to do the image recognition and image classifications.

II. Skin Cancer Detection Technology Pipeline

Image processing is a method to perform some operations on an image, in order to get an enhanced image or to extract some useful information from it. Image processing methods are becoming increasingly sophisticated and the tendency is to develop as much automation as possible. At times, you want to get rid of distortions caused by lights and shadows in an image. Normalizing the RGB values of an image can at times be a simple and effective way of achieving this. When normalizing the RGB values of an image, each pixel’s value is divided by the sum of the pixel’s value over all channels. So if we have a pixel with intensified R, G, and B in the respective channels, its normalized values will be R/S, G/S, and B/S (where, S=R+G+B).

• Image Acquisition
• Image Preprocessing
• Image Segmentation
• Feature Extraction
• Image Classification

III. Deep Leaning

Deep learning (also known as deep structured learning or hierarchical learning) is part of a broader family of machine learning methods based on artificial neural networks. Deep learning is a class of machine learning algorithms that uses multiple layers to progressively extract higher-level features from the raw input. For example, in image processing, lower layers may identify edges, while higher layers may identify the concepts relevant to a human such as digits or letters, or faces.

Convolutional Neural Network

A Convolutional Neural Network (CNN) is an algorithm in deep learning which consist of a combination of convolutional and pooling layers in sequence and then followed by fully connected layers at the end a like multilayer neural network. CNN is a class of algorithms which is motivated to take advantage of any 2d structure in data. Hence CNN is one of the preferred algorithms for image classification. CNN also proves promising for task related to natural language processing. CNN leverages the local feature of an image to achieve higher accuracy for classification. CNN is one of the neural networks which is easy to train and has less number of hyperparameter as compared to ordinary fully connected neural networks.

The network will consist of several convolutional networks mixed with nonlinear and pooling layers. When the image passes through one convolution layer, the output of the first layer becomes the input for the second layer. And this happens with every further convolutional layer.

The nonlinear layer is added after each convolution operation. It has an activation function, which brings nonlinear property. Without this property, a network would not be sufficiently intense and will not be able to model the response variable (as a class label).

The pooling layer follows the nonlinear layer. It works with width and height of the image and performs a downsampling operation on them. As a result, the image volume is reduced. This means that if some features (as for example boundaries) have already been identified in the previous convolution operation than a detailed image is no longer needed for further processing, and it is compressed to less detailed pictures.

VGG16: It is a convolutional neural network model proposed by K. Simonyan and A. Zisserman from the University of Oxford in the paper “Very Deep Convolutional Networks for Large-Scale Image Recognition”. The model achieves 92.7% top-5 test accuracy in ImageNet, which is a dataset of over 14 million images belonging to 1000 classes.

The activation layer is used for show how the neuron layers inside the model is connected. There are 1024 neuron layers are connected. ADAM optimizer is an adaptive learning rate optimization algorithm that is been specially for training deep nural networks. This algorithm leverages the power of the adaptive learning rates method to find the individual learning rates of each parameter.

Conclusion

Due to the consequences of inadequate detection and of excellent detection accuracy, the detection of skin cancer has been deemed challenging. The implications of artificial intelligent methods and the efficient utilization of soft computing skills would negate the issues related to detection inaccuracy the proposed CNN model was tested on different use cases and produced better accuracy results than other methods. Hence, our proposed model produced the best results.

Reference

1. W. K. Saa'd, 'Method for detection and diagnosis of the area of skin disease based on colour by wavelet transform and artificial neural network', Al-Qadisiya Journal for Engineering Sciences, Vol. 2,No.4, 2009
2. Seema Kolkur, Dhananjay Kalbande, Vidya karkar, “Convolution Neural Network for Feature Extraction in Skin Disease Detection”, Journal of Advanced Research in Applied Artificial Intelligence and Neural Network Volume 5, Issue 1&2 - 2018, Pg. No. 1-5.
3. Jainesh Rathod, Vishal Waghmode, Aniruddh Sodha, ​Dr. Prasenjit Bhavathankar, “Diagnosis of skin diseases using Convolutional Neural Networks”, ICECA IEEE Conference Paper,2018
4. Sujay S Kumar Varun Saboo, “Dermatological Disease Detection Using Image Processing and Machine Learning”, IEEE Conference Paper, 2016
5. Jyostna More, Prachi Bhadale, Girija Lonkar, “Method for Skin Disease Detection Using Color Images”, International Journal of Engineering Science and Computing, 2019
6. Enakshi Jana, Dr.Ravi Subban, S,Saraswathi,“Research on Skin Cancer Cell Detection using Image Processing” IEEE Conference Paper,2017 [bookmark: page2][bookmark: page3]