Salami medical article https://doi.org/10.1016/j.anclin.2019.04.007

uhpo34t

1. AI, big data and the protection of personal data in medical practice Emmanuel Salami, LL.M. [email protected] UNIVERSITY OF LAPLAND FACULTY OF LAW

2. Introduction EDUCATIONAL & PROFESSIONAL BACKGROUND LLB Law - University of Lagos. LLM IT and IP law - Leibniz Universität Hannover. Doctoral Candidate, IT/data protection and IP law - University of Lapland. Privacy Manager – HERE Technologies. RESEARCH INTERESTS The intersection between IT/data protection and IP law in the use of various emerging technologies. Data protection compliance. Intellectual property rights protection of AI systems . Data ethics .

3. Table of content What is AI? What is big data? Uses of AI in medical practice Data protection concerns and remediation actions in the use of AI in medicine Other recommendations Concluding analysis

4. What is AI and big data? AI is intelligent behavior (in computers) through the ability to achieve human-level performance in all cognitive tasks, sufficient to fool an interrogator. The key test of computer intelligence is when the results generated by the computer cannot be distinguished from those of its human counterparts – Turing. AI is “the study of agents that exist in an environment and perceive and act - Russel and Norvig . Strong AI vs Weak AI. AI uses big data sets to learn from its experiences (machine learning) Three Vs’ definition of big data where ‘Volume’ relates to massive datasets, ‘Velocity’ relates to real-time data and ‘Variety’ relates to different sources of data -Laney, Big data is difficult to process using traditional processing methods.

5. Uses of AI in medical practice ? Descriptions Disease diagnosis and prediction e.g – AI examines e chocardiograms and classifies heart conditions. AI also predicts the possibility of suffering conditions such and predicting who is more likely to have what disease. AI has correctly predicted 355 cases of heart attack better than human medical doctors. Telerobotic surgeries and robot-assisted surgeries – Surgeries are carried out with the aid of robotically controlled instruments through which coronary intervention can be undertaken by a doctor without any physical contact with the patient. (Operation Lindbergh, 2001). Genomics – AI is being used to make specific changes in the DNA sequence of human beings (gene editing) to remove certain traits therefrom and insert new traits. AI is cheaper. Drug discovery, development and repurposing – AI analyses available data and identifies good target proteins (drug discovery). Drug repurposing is the application of an approved drug for the treatment of a different disease by using AI to evaluate the approved data of drug molecules to meet new targets. AI accesses the sensitive personal data of patients in the process. Clinical trial – AI helps in the selection of candidates in determining the potency, side effects, etc. of new drugs.

6. Data protection concerns and remediation actions in the use of AI in medicine Concerns Description Remediation actions Lawfulness and transparency principle Necessity of processing , ‘Anonymised‘ genetic data , Data repurposing , Transparency becomes more difficult because of the above . Thorough anonymisation and review of anonymisation in genetic data. Assess and ensure strict compliance with Article 6(4) GDPR? Provision of proper information Decision making algorithms AI is determining people’s fate – who is healthy enough for certain jobs for instance but little or no explanation because of black boxes , opaque algorithms , human supervision? etc. Data protection by default and design to ensure AI is designed to provide adequate explanation , Human intervention must be established . Data security and data access Use of (personal) data in test environment , Data passes through more non- medical personel like information security experts . Where possible, avoid the use of personal data in test environment , Adequate technical and organizational measures , Restrict data access on a strictly need to- know -basis.

7. Data protection concerns and remediation actions in the use of AI in medicine Concerns Description Remediation actions Discrimination of data subjects This could be caused by the use of biased data , Non-standardisation in the development of algorithms which encourages the reflection of the biases of developers and engineers in AI. The US example where people of colour were erroneously and unlawfully exempted from medical benefits . The need for more standardizations and codes of conduct . Mittelstadt analogy on Doctors vs patients and AI developers vs end users , Algorithms (and non-personal data) should be accessible to researchers and stakeholders for auditing . Da ta Transfers Data transfer outside the EU. For instance - the first telerobotic surgery ‘ Operation Lindbergh’ was carried out on a patient in France while the medical team was located in the USA. Ensure compliance with data transfer rules .

8. Other recommendations Descriptions Accountability of AI developers in medicine : Mittelstadt analogy on doctors vs patients and AI developers vs end users , Hippocratic oath for AI developers : ” proactively understanding the ethical implications of technology for all stakeholders, telling the truth about the capabilities, advantages and disadvantages of a technology, and acting responsibly in situations which are morally challenging“ - Rothstein AI developers in medicine should be held to a standard as high as those of medical doctors. A sensitive and highly regulated endeavour like medicine cannot be a test ground for AI.

9. Concluding analysis What are the key take away points? AI should support and not replace doctors, AI in medical practice should be closely monitored, Stakeholder participation in medical AI regulation is key.

10.   THANKS FOR YOUR ATTENTION!

11. Feedback and reviewer comments