Reference no: EM132998128
Assignment - Project - SKIN CANCER DETECTION
Skin cancer accounts for the largest number of cancer incidence and mortality rates in the world. In the United States in 2015, for example, approximately 350,000 of melanoma cases were identified, with around 60,000 deaths.
Skin cancer is known as an abnormal condition in which skin cells multiply rapidly due to DNA mutation or generic defects. In medical science, the disease is classified as melanoma and non- melanoma, of which melanoma is one of the deadliest diseases with 98% incurable worldwide. Although the mortality is considerably high, there is a high chance for a cure if the disease is detected early. In Queensland, Australia, for instance, the incidence of thin melanoma in young people is decreasing due to benefits of primary prevention efforts of detections in the early stages.
In this project, students will investigate and develop an image processing technique to detect and classify the skin conditions, classified as Benign or Malignant. Students are encouraged to apply DSP techniques for image processing to improve the classification performance.
Project Tasks - The project should be conducted in following tasks:
1. Data collection
The project will be conducted based on the dataset of skin lesion in ISIC archive (The ISIC 2020 Challenge Dataset | ISIC 2020 Challenge Dataset.
2. Literature Review
As a group, conduct a thorough literature review on related works and present your findings in the written report.
3. Algorithm Development
Each group will develop their own classification algorithm based on its literature review. The proposed and developed algorithm must be able to classify the image as benign or malignant. Your algorithm must be robust against scale, orientation, lighting variance, and strong enough to identify the variety of skin moles.
The key idea of the unit project is to help students further developing the knowledge and skills in Digital Signal Processing, particularly in Image Processing. Applying image processing techniques to enhance the image quality will play a key role in improving the performance of developed algorithm regardless of a machine learning technique was used. Note that there are no requirements on how to proceed the development of the algorithm; students may want to:
Remove the skin hair to improve the mole visibility.
Study the characteristics of skin moles of both benign and malignant. e.g. shape, color, size, distribution, and so on.
Develop and test the algorithm using the studied characteristics.
Study on how to enhance image quality using image processing techniques and further improve the algorithm with noisier images (images with various background, blurry images, ...).
Your algorithm must meet the following criteria:
(a) Written in MATLAB. There may be multiple script files but should have a main script to run the algorithm, named as main.m.
(b) Prompt the user for a folder which contains a set of input images in JPEG format.
(c) Perform the image classification task on all images within the specified folder. The classification will be solely based on the input images. You must not use the filename or provide any additional information to assist the algorithm.
(d) Display the classification results on the screen by showing each image filename and its associated result.
(e) Take less than 10 seconds to execute each classification (preferably instantaneous). For example, if there are 10 image files in the folder, then your upper limit is 100 seconds.
Submit your algorithm to the unit Blackboard site by the due date. The accuracy of your algorithm will be evaluated by the lecturer.
4. Report
A group must work together and write a 4-page double-column paper documenting the work that you have achieved. Report:
must use the given template.
do not amend any of the given headings.
should not exceed the 4-page limit.
include the confusion matrix (from testing set), the success rates of your training and testing, and the execution speed must be presented.
5. Poster Presentation
One poster presentation per group. Prepare a 5-min 3-slide poster presentation.
Construct your poster in PowerPoint format according to Table 1. While all essential contents must appear in your poster, you may add further information as you see fit and within the 3- slide limit.
Poster must be submitted by the due date. Penalty for late submission will be applied according to the school policy.
Table 1: Essential contents to include in your poster.
Slide #
|
Essential Contents
|
1
|
Cover Page
- Title
- Names and student ID of all students
- Team number
- Unit code
- Semester and year
|
2
|
Proposed Methodology
- Flowchart / Block Diagram
|
3
|
Results and Discussion
- Confusion matrix (from testing set)
- Execution speed
|
Approximately 15 minutes are allocated to each group. The person in charge of the presentation submission will deliver a 5-minute presentation with the help of the other two team members. There will be a 5-minute Q&A after each presentation, and another 5-minute buffer for change- over. All team members must be able to address questions from the audience.
Note - Only technical work is required with explanation. Report and Presentation is not required.
Attachment:- DSP Project Assignment File.rar