Introduction
Artificial Intelligence (AI) has a variety of uses, ranging from identifying genetic codes for precision medical treatment to prosthetics and even robotics in surgery. AI can help to minimise human error and analyse patient data much quicker than a human could. However, in healthcare, a people-focused world- what are the uses for AI? What are the positive and negative impacts they have in this field? This article will examine the uses and issues in applying AI in healthcare.
AI in medical diagnosis
Research by the Multiple Sclerosis (MS) Society looked at MRIs from people with MS that were taken between 6 and 29 months (about 2 and a half years) apart1. Three different MRI reports were looked at. The routine MRI report (this is the report you normally get from your clinician when you have an MRI), a specific MRI report written by trained clinicians (following strict guidelines to increase objectivity) and information created by specialised AI programmes. The study discovered that AI was consistently better at finding new or growing lesions compared to the routine MRI report provided by the care team and the report written by a trained expert at the MRI lab. Detecting slowly growing lesions is crucial as it is an early indicator of disability progression, in particular when someone no longer has active relapses anymore - termed “smouldering” MS. This means in the future, MS can be detected in earlier stages to better prepare patients for the progression and allow them to receive appropriate treatment. X-rays, CT scans, and other diagnostic tools can also be significantly enhanced to improve the detection and diagnosis of many other diseases.
AI In treatment planning
AI algorithms can process large amounts of patient data such as genetic information, medical history and lifestyle to curate personalised treatment plans. Although a clinician can also formulate a patient-centred treatment plan, the difference is that AI can notice very subtle mutations that a clinician, especially in a time-pressured NHS environment may not notice.2 This improves health outcomes as more effective, preventative care can be given. In addition, some AI algorithms are constantly adapting, including the consideration of real-time responses from patients. However, some risks come along with using this technology. Technically, treatment plans formed by algorithms are personalised since they incorporate specific patient data and possibly have some form of shared decision-making, it still is not true to the core idea of personalised.
The core element of personalised care is that ‘people have choice and control over the way their care is planned and delivered’. It is based on ‘what matters’ to them and their individual strengths and needs.3 Although one can argue AI can achieve this goal, there is no individual-level relationship between the patient and the AI.5 This is because the plan is ultimately driven by big data analyses which detect a pattern from a population and generalise it to an individual patient. Furthermore, the AI tools used do not have extensive research supporting their use, for example, although IBM Watson’s decision support product has now been transferred ownership, there have not been enough randomised clinical trials (RCTs) to see if the algorithm is based on incomplete or biassed data. This makes the reliability of the AI used unreliable. That said, rather than using RCTs to review, it may be more efficient to implement learning systems where the data AI is based on is continuously updated, though again the data would have to be free of bias and would take a long time for the AI to become accurate and have reliable decision making. Last, deep learning algorithms and neural network-based algorithms, make AI predictions reliable if the initial data set is large. However, the link between the inputs (i.e., data) and the outputs (predictions) can be very hard to understand; this is called the ‘black box ‘problem. Thus, AI can be useful in data-based predictions and finding associations, when the outcome has a huge effect on human life it may not be worth the risk e.g., in identifying a drug target that, once modulated, causes a desired effect, finding the causal relationships is crucial.
Robotics in surgery
A famous example is the da Vinci Surgical System. This allows a surgeon to be seated away from the patient and operate with a much greater range of motion than any human hand could ever achieve. A single robotic arm is put through an incision smaller than three centimetres- leaving much less damage than if a surgeon were to perform without this equipment.6 Note that this is mainly recommended for minimally invasive surgery. However, it has been used for a range of specialist surgeries; a huge 75% of prostatectomies are performed robotically within the NHS.7
Dr Kathryn Oakland, Director of Robotics at HCA UK states patients undergoing minimally invasive surgery assisted by the da Vinci® can ‘return home faster, with less pain, less trauma and more mobility’. Although there is research to support this, it is very limited with a recent review showing just two studies which support the accuracy of the Da Vinci system, also revealed using the system increases the duration of surgery and does not have a higher accuracy than the freehand technique used in brain surgery.8 This indicates a large gap in the evidence base for the use of the system though it may have a lot of potential, it is up to clinicians/hospitals if they believe it is an investment worth making.
Smart prosthesis
An example of how machines can aid in improving a patient’s Health-Related Quality of Life (HRQL) is the use of lightweight smart prosthesis which can assist paralysed people in walking again. The Cyberdyne’s Hybrid Assistive Limb (HAL) exoskeleton uses sensors placed on the skin to detect electrical signals which indicate specific movement at the joint.9 This enables patients who are in rehabilitation (from spinal cord injuries or strokes) to move again. This can be life-changing for some as they can exercise independence and do activities they love again.
Ethical dilemmas
A very recent report revealed more than 60% of patients lack trust in AI in healthcare.10 This mistrust is caused by concerns over data privacy, potential biases, and the lack of transparency in AI decision-making processes. The General Data Protection Regulation (GDPR) is a privacy and security law passed by the European Union, following these regulations, all personal data and the activities of foreign communities and companies are processed by a union-based data processor or controller to protect information.11 Additionally, the US also has the Genetic Information Non-discrimination Act (GINA) which protects individuals against discrimination based on their personal genetic information (particularly in health insurance and employment) to encourage people to undergo genetic testing.12 Notice how neither of these data protection acts mentions the use of AI. Although there may be specific laws relating to AI, there is a lack of legal protection of patient data when AI is involved. Clinical data which AI uses can be hacked also some genetic testing/bioinformatics companies (illegal or unmonitored) sell customer data to pharmaceutical and biotechnology companies.
Conclusion
In conclusion, AI has the possibility to greatly improve patient outcomes due to its fast analysis of large data sets. However, there are many ethical and practical drawbacks which means the full potential of this technology cannot be unlocked. Furthermore, the research which supports the use of AI is limited. In the future, with improved regulation, research and understanding of AI, healthcare could be revolutionised.
References
- Multiple Sclerosis Society (2021) New research shows AI may help doctors interpret MRI scans more accurately. Available at: https://www.mssociety.org.uk/research/latest-research/latest-research-news-and-blogs/new-research-shows-ai-may-help-doctors-interpret-mri-scans-more-accurately#:~:text=What%20did%20the%20researchers%20find%3F&text=AI%20was%20also%20better%20at,that%20nerves%20are%20being%20damaged. (accessed 23rd May 2024)
- Rubio Laboratories (2020) The Role Of Artificial Intelligence In Personalized Medicine. Available at: https://www.laboratoriosrubio.com/en/ai-personalized-medicine/#:~:text=However%2C%20the%20true%20potential%20of,to%20create%20personalized%20treatment%20plans. (accessed 23rd May 2024)
- NHS england (2024) What is personalised care? Available at: https://www.england.nhs.uk/personalisedcare/what-is-personalised-care/#:~:text=Personalised%20care%20means%20people%20have,their%20individual%20strengths%20and%20needs. (accessed 23rd May 2024).
- Schork N. J. (2019). Artificial Intelligence and Personalized Medicine. Cancer treatment and research, 178, 265–283. https://doi.org/10.1007/978-3-030-16391-4_11. Available at: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7580505/ (accessed 22nd May 2024). ibid.
- Intuitive (2024). Da Vinci Robotic-Assisted Surgery. Available at: https://www.intuitive.com/en-us/patients/da-vinci-robotic-surgery/about-the-systems. (accessed 22nd May 2024).
- HCA Healthcare (2024). Robotic excellence with the da Vinci® surgical system. Available at: https://www.hcahealthcare.co.uk/about-hca-uk/robotics-and-technology-at-hca/the-da-vinci-surgical-system. (accessed 22nd May 2024).
- Lin, T., Xie, Q., Peng, T., Zhao, X., & Chen, D. (2023). The role of robotic surgery in neurological cases: A systematic review on brain and spine applications. Heliyon, 9(12), e22523. https://doi.org/10.1016/j.heliyon.2023.e22523. Available at: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10686875/#:~:text=It%20was%20reported%20that%20using,patient%20recovery%20times%20%5B40%5D. (accessed 22nd May 2024).
- Mohammed Yousef Shaheen. Applications of Artificial Intelligence (AI) in healthcare: A review. ScienceOpen Preprints. 2021. DOI: 10.14293/S2199-1006.1.SOR-.PPVRY8K.v1. Available at: https://www.scienceopen.com/hosted-document?doi=10.14293/S2199-1006.1.SOR-.PPVRY8K.v1. (accessed 22nd May 2024).
- Rajeev Ronanki (2024). Ethical AI in Healthcare: A Focus on Responsibility, Trust, and Safety. Forbes. Available at: https://www.forbes.com/sites/forbesbooksauthors/2024/01/04/ethical-ai-in-healthcare-a-focus-on-responsibility-trust-and-safety/?sh=68803ac5787f. (accessed 22nd May 2024).
- GDPR.EU (2018). What is GDPR, the EU’s new data protection law? Available at: https://gdpr.eu/what-is-gdpr/. (accessed 22nd May 2024).
- National Human Genome Research Institute (2024). GENETIC INFORMATION NONDISCRIMINATION ACT (GINA). Available at: https://www.genome.gov/genetics-glossary/Genetic-Information-Nondiscrimination-Act. (accessed 21st May 2024).