Security, Privacy and Ethics
Coursework 2
B: Evaluating the potential risks and benefits of AI technologies on privacy and personal
Learning outcomes assessed: data C: Understanding the importance of fairness in AI systems and its implications Overview
Part 1 (Individual Report: 70 marks Artificial intelligence has great effect on modern lives. In this coursework, the theme on framework designof new medical image-based bias-mitigated and fair computer-aided diagnosis system will be investigatedand explored. The coursework consists of two parts. In Part 1, you needto complete a report based given
theme. In Part 2, you need to explain your design via video presentation.) Healthcare industry is rich of electronic (digital) medical data from different modalities. Deep learning hasrevolutionized the use of machine learning in healthcare industry by leveraging on the model’s automaticfeature extraction and learning. To date, deep models have been applied for numerous computer-aideddiagnosis tasks such as prediction,detection and classification. Despite its promising outlook, deep learningbased computer-aided diagnosis models still fail to earn the trust of medical doctors. In fact, there are
reports of inaccurate missed diagnoses due to bias error, lack of understanding about the underlyingmechanism of deep learning, and miscalibration in real practice. Therefore, there is an open call for fairand transparent computer-aided diagnosis model for trustworthy smart healthcare.The aim of this task is to empirically assess the current status of computer-aided diagnosis, technologies,
challenges and solutions in smart healthcare, and research questions in AI fairness to design a towards innovative bias-mitigated and fair deep learning medical image-based computer-aided
diagnosis model framework design with predefined traits. Hence, you are required to equip yourselveswith the understanding about the specific domain of medical imaging. Besides, you need to apply the knowledge acquired from the lectures and tutorials to complete this coursework. You also need to doliterature review to identify further relevant information that is helpful to develop your report content.
Task Instructions:
(1) You are required to study the medical image-based artificial intelligence computer-aided diagnosisby using deep learning in smart healthcare domain. Therefore, literature review is needed. For
beginner, you can refer to the suggested review paper to understand the domain of smartwith imagebased computer-aided diagnosis by using deep learning. Please note thatno mark will be given tothe literature review nor content extract from the given review paper. However, this effort shallserve as your first steep for your proposed towards innovative bias-mitigated and fair deep
learning medical image-based computer-aided diagnosis model framework design.
(2) Write a report on your proposed towards innovative bias-mitigated and fair deep learning medical image-based computer-aided diagnosis model framework design. he reportshould be written in a clear and concise manner with no more than 1,500 words+/-5%. in total
Approximate
word count MotivationYou are required to select your project domain from one of thefollowing areas of interest:COVID-19 detection
- Thorax abnormality detection
- Late age-related macular degeneration detectioYou are required to explain the motivations associated with yourselected area of interest and how AI-based computer-aideddiagnosis can help to improve the diagnosis delivery in yourselected area of interest.150Case Study 1 Privacy and ethical challenges associated with large-scalemedical image datasets for Artificial Intelligence-basedmputer-aided diagnosis models50You are required to identify an open access large-scale datasein your selected area of interest and study the dataset details.Then, you are required to:
Explain the relevant ethical issues associated witmedical image acquisition and sharing for opendevelopment of computer-aided diagnosis• Evaluate the relevant privacy risks associated withmedical image data sharing of your selected area ofinterest in accordance to the Health InsurancePortability and Accountability Act (HIPAA)
- Describe the innovative and rationale data governancemeasures to achievebettertransparency,communication security and sensitive digital medicaldata protection of your selected area of interest.Case Study 2 Bias and Fairness in Artificial Intelligencebased computer-aideddiagnosis models450You are required to perform analytics approach on existing
computer-aided models to help address this case study. Pleasenclude necessary evidence of analysis such as detection results,graphs and plots from the python programmingcodes tosupport your justification.代寫Security, Privacy and Ethics Please note that no mark will diagnosismodelassociated with your selected area of interest. The newmodel should critically cover these traits:Innovative bias-mitigation strategy in new AI-basedcomputer-aided diagnosismodel’s algorithm setting
- Innovative fairness practice in new AI-based computeraided diagnosis model’s social equality setting
- Incorporation of innovative transparency approach inyour computer-aided diagnosis model’s algorithm and
ohort criterion.FrameworkDesignYou are required to propose and explained a new designframework based on the motivation, case study 1 and 2. Thenew design framework should be illustrated in a detailed overallmodel diagram. The design framework should encompass thefollowing characteristics:Selection of most appropriate AI backbone models formaximum bias-mitigation compared to other state-ofrt
- Use of relevant explainable AI for transparenalgorithmic mechanismUse of relevant evaluation metrics to evaluate the AIfairness improvement450Important: Do not repeat existing information that is in the research papers. This will onlycontribute to low mark. Instead, you need to synthesize your own ideas/opinions based on yourunderstanding and present them in your own words. Part 2 (Individual Presentation: 30 marks)
Task Instructions:
(1) Prepare and record a short individual presentation video of 5 minutes+/-5%. Your presentationshould be clear, should be in no more than 10 Powerpoint slides and should not take beyond
5 minutes+/-5%. The presentation should address the followings:
To introduce and explain the significance of your proposed towards innovative bias
mitigated and fair deep learning medical image-based computer-aided
diagnosis model framework design.To explain how your proposed model design can effectively become General Data
Protection Regulation (GDPR) and IEEE “Human Standards” with Implications for AI compliance in order to promote your design to overseas healthcare marketsuccessfully.
Report Format:
Cover Page: This should include the Assessment Number, Assessment Title, Student Name, Student IDand Student EmailBody of the report: This should include all the relevant section headings to address each section asindicated above and marking rubrics.References: Both yourin-text and the references included in the “References” section at the end of thereport should adhere strictly to theIEEE reference style.Formatting requirement:
- Use multiple spacing: 1.08 and spacing after: 8pt;
- Use a standard 12-point font, font type: Tahoma
- Use “Justify” body text
- Put your page numbers at the top right (except the cover page)
- Most importantly, always run a spelling and grammar check; however, remember, such checksmay not pick up all errors. You should still edit your work manually and carefully.
Referencing:
It is compulsory to use IEEE reference style for citing and referencing research. Reference list is excludedfrom the imposed word limit.
Presentation Format:
Students are not requested to submit their presentation slides to the submission system. However, theymust present their Powerpoint presentation slides clearly throughout the video presentation period.Otherwise, the presentation will not be evaluated and ZERO marks will be given.All video presentation must be uploaded to the Mediasite and attach the video presentation linkat the lastpage of report. It is student responsibility to attach the link properly and apply the right accessibility settingin Mediasite to ensure examiners can access to the video presentationlink in their computers during
of circumstances, and has the right to give mark for the inaccessible video presentation link attached in the report.Marking Criteria Part 1: Individual Report (70 Marks)
The following criteria will be used to assess the assignment. This report is marked for the whole group.Outstanding: Report format is well-organized throughout including heading styles, fonts, and margins,
figure/table/diagram are critically interpreted and discussed, writing flows extremely fluently from one ownidea to another, information is critically deciphered and elaborated in convincing and perfect way withimpactful evidence that reflects highly insightful and unique own idea expression, all information is locatedin the appropriate section.Referencestyle must be precisely correct and citation arranged sequentially inmain text.Very Good: Report format is properly organized throughout including heading styles, fonts, and margins,figure/table/diagram are effectively interpreted and discussed, writing flows smoothly from one idea toanother, information is well explained in a logical, rationale and systematic way with comprehensiveevidence that reflects insightful own idea expression, all information is located in the appropriate section.Reference style must be correct and citation arranged sequentially in main text.
Appropriate: Report format is organized and proper, figure/table/diagram are properly interpreted andReference style must be correct andcitation arranged sequentially in main text.Needs Improvement: Report format is inconsistent and hard to read, figure/table/diagram are poorly
interpreted and discussed, sentence structure and/or word choice sometimes interfere with clarity inexplaining own idea, information is hard to follow as there is very little continuity, and many items are inthe wrong section. Reference style must be correct.Hard to Understand: Report format is inconsistent and hard to read, figure/table/diagram are not usedeffectively, sentence structure and word choice make reading and understanding difficult, sequence ofinformation is difficult to follow without much own idea expression, lack ofappropriate sections and manyitems are in the wrong section. Reference style might be incorrect.No Submission or Missing Section: No submission or missing section of the discussion in the report.