Early Detection of Melanoma Skin Cancer using Image based segmentation and Classification

Project Title

Early Detection of Melanoma Skin Cancer using Image based segmentation and Classification

Project Area of Specialization Artificial IntelligenceProject Summary

Skin is the largest organ of the body and the most vulnerable to DNA damage caused by the ultraviolet rays of sun. It is just not possible to completely avoid sun exposure, which is why proteins that repair DNA damage are important to prevent skin cancers like Melanoma, Basal, and Squamous. Melanoma of the skin is the 19th most commonly occurring cancer in the world, with approximately 300,000 new cases worldwide in 2018. However, this is likely to be an underestimated. The number of peoples diagnosed per year with different type of skin cancer is projected to increase over the next 20 years[6]. Basically technology is playing a vital role in Health care Industry. It facilitates human beings with quick solutions at low cost and everywhere in the world. Our idea is an early detection of Melanoma skin cancer which diagnosis skin cancer using image processing and deep learning algorithms over the cloud services. Our solution is a mobile application which captures the photo and send it to cloud for analysis and the respond back low/high detection rate to the users. The challenge of developing a mobile application is that; to detect skin cancer with high accuracy and targeted audiences (untrained people). In addition to that a lot of research is required on the health system and user population. finally, mobile health apps will be useful tool for user to reduce skin cancer management cost and avoid skin cancer to cause lethality

Project Objectives

Our project uses an innovative algorithm to analyses skin spots, enabling peoples and the health system to minimize cases of skin cancer and positively touch the lives of millions of people around the globe.

Our project aims to save millions of lives in the next decade. Doctors use our application as second opinion or first opinion to identify skin cancer.

our project is to save skin healthcare cost management as well as time and it is an ease tool for self-monitoring for the peoples to diagnosed the unusual spot (mole) of the skin using their android phone with high camera.

Project Implementation Method

A mobile application which captures the image of the mole on skin and sent it to server for Early Detection of Melanoma skin cancer using image-based segmentation and classification. In Digital image processing we use different image segmentation techniques such as (threshold based segmentation,clustering, edge detection and etc) to calculate Total Dermoscopy score index for skin cancer detection. And simultaneously we have trained model with deep learning logarithms to analyze for early detection of Melanoma Skin Cancer within few seconds. we compare both results and then send result to user High/Low. In worst case, if we have obtained inverse results between both techniques then we sent it to dermatologists group to analyze. But we can achieve more than 90% accuracy because of big data-set available publicly. And we are using neural network which increases the accuracy as the data-set increases. Both technique are reliable for early detection but our approach is hybrid approach for reliable service to people to diagnosed the melanoma skin using their mobile phone by capture the image of unusual spot(moles) of their skin at any where in the world.

Benefits of the Project

Best self-monitoring and assessment tool for the people of top 20 countries where Melanoma Skin cancer prevalence is high.

peoples who are too busy in their professional life and dislike to wait in queue and get doctor appointment that’s why they always skip their skin healthcare. When the reach at final stage of melanoma skin cancer then the consultant with doctor then only biopsy is solution.

It saves you from physical meetup with doctor when our report result is High on that time people will consultant with doctor.

Our project is for those who do not want to show the spotted area at private place of the skin to dermatologist.

Our project saves the lives, cost of skin heath-care, time and suggest you to consultant with doctor at right time when you are at an early stage which is curable and cost saving.

Technical Details of Final Deliverable

Fully functional working android application which captures image and send it to local server using REST API. On local server we will have model completely trained with deep learning algorithms and image processing technique for feature extraction from input image and properly perform calculation to produce Total dermoscopy score (TDS) index and simultaneously verified and validated same input image with deep learning model with 85% accuracy to diagnose the melanoma. Our application will posses Human Computer Interface (HCI) good design principles and user friendly GUI which will be ease for low-tech and high-tech people to use it by login or sign up on it and capture the image and easily get report within few seconds. Report results will be show Low/High risk if result is high it will suggest to consultant with dermatologist. Our application will maintain user history. And beside this all the diagnosed cases are added in our local database and it will increase the accuracy of our deep learning model as people send the image for diagnosis.

Final Deliverable of the Project Software SystemCore Industry ITOther Industries Health Core Technology Artificial Intelligence(AI)Other Technologies Cloud Infrastructure, OthersSustainable Development Goals Good Health and Well-Being for PeopleRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	80000
Graphic Card MSI NVIDIA Getforce GTX 1060	Equipment	1	38000	38000
SSD drive 125GB	Equipment	1	9000	9000
Samsung Galaxy J4 Plus	Equipment	1	23000	23000
RAM 8GB DDR4	Miscellaneous	1	5500	5500
Printing charges	Miscellaneous	1	4500	4500