Skin Cancer Forecaster: Image-Based Cancer Detection Using Deep Learning

Project Title

Skin Cancer Forecaster: Image-Based Cancer Detection Using Deep Learning

Project Area of Specialization Computer ScienceProject Summary

Skin cancer in developing countries has a serious impact on people’s quality of life, causing lost productivity at work and discrimination due to disfigurement. Skin changes may also indicate the presence of more serious cancer diseases that need treatment immediately. Multiple types of skin cancer but the most dangerous one found in humans is melanoma. If skin cancer is not detected at early stages, then the survival rate decreases from 99% to 14%.  Such conditions are being completely ignored or given low priority because they all start from a single spot on your body that does not cause you any pain initially. But when it starts to grow then its outgrowth is so sudden that it causes severe health issues. Due to misconception and lack of knowledge, it is important to treat skin cancer and educate communities about how to prevent it. As the skin is the most important organ which serves many functionalities including organ protection, fluid preservation, temperature regulation, and eliminating toxins from the body by sweat excretion.

Visual information is the most important type of information perceived, processed, and interpreted by human beings. In the same context, image processing plays an important role in solving problems of different kinds, specifically medical related problems to improve the diagnosis of the medical disease and by detection through an image. In the modern era, computer-aided diagnosis (CAD) systems have become a necessity for the early detection of diseases. Health care applications should be provided to the public living in both remote as well as urban areas.

Skin cancer forecaster is a mobile-based application that will let users make a verdict about common skin cancer in the remote and urban areas of Pakistan. Skin cancer forecaster will help users to scan the area of skin and then it will predict the relevant skin cancer type through image processing. After predicting the type, it will recommend medicine according to the category of severity type. 

Project Objectives

1. To provide easy access to the user about skin cancer timely prediction on a mobile handset.
2. To tell the user about the severity level of relevant skin cancer type (malignant and benign).
3. To recommend medicines based on severity level.
4. To suggest measurable precautions to control skin diseases timely.

Project Implementation Method

Our proposed system will tackle the more severe cases of skin cancer problems using deep learning with more detailed research on pre-trained model VGG-16, YOLO, Alex-Net, Google-Net so that model can classify cancer with the best accuracy. Additionally, it will detect the category of relevant skin cancer diseases based on severity which will fall into either malignant or benign. Further medications will be suggested based on the severity level. This will help the user in the timely and efficient prediction of severe skin cancer disease if any, and assist the user in the treatment by suggesting medicines according to severity level and diabetic patients. A software-based approach to classify type and severity level (malignant or benign) of skin cancer diseases from the skin lesion image taken by normal mobile camera.

Steps involved in the process of skin cancer prediction are as follows:

• Image acquisition
• Image preprocessing
• Image segmentation
• Skin cancer type and severity classification
• Generate report

It will be a flutter based mobile application which facilitates all type of users who belongs to any operating system.

Benefits of the Project

Project will facilitate on the following basis.

• It will help to identify the issue before going to high severity.
• It will reduce the time of long medical testing methode.
• Easy access to every user who want to test the disease.

Technical Details of Final Deliverable

Project will following deliverables.

• Best model will be implemented in appkication.
• Mobiles have cameras but in desktop version you need high resolution camera.
• It will detect the the cancer type and generate report considering diabetes and other factors.
• Easy access to every user who want to test the disease.

Final Deliverable of the Project HW/SW integrated systemCore Industry HealthOther Industries IT , Medical Core Technology OthersOther TechnologiesSustainable Development Goals Life on LandRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	60000
API	Miscellaneous	1	5000	5000
Camera	Equipment	1	35000	35000
Sensors	Equipment	1	15000	15000
Google Publisher account	Miscellaneous	1	5000	5000