Artificial intelligence algorithms can improve the performance of radiologists, by improving the speed and accuracy of diagnosing their patients. The algorithms can interpret and give diagnoses on X-rays, CT scans and other images (Kim & Holzberger, 2019). However, the drawback is that AI only focuses on one question and hence, one answer. Each purpose would need its own algorithm, forcing developers to create thousands of algorithms (Kim & Holzberger, 2019). However, AI marketplaces gives access to a variety of AI models and hence, tries to solve this problem. Also, it asks for feedback to refine its algorithms. Moreover, Forbes (2019) claims that AI could help to solve a big problem in healthcare, namely the ‘iron triangle’. The triangle exist of the three factors access, affordability and effectiveness. AI can decrease costs, but also make treatment improvements and create a great accessibility. Forbes (2019) also sees a great future in the use of AI on robots that assist operations.
Even though prospects look great, we also need to discuss the other side. Of course, there is a large difference in using AI in for example buying stock and in using it to diagnose or operate a real human being. Is it ethical to use AI in healthcare? Whose fault will it be when someone passes away due to a fault in the algorithms? Also, AI gives an output it cannot further explain, and it can even rely on bias due to the data it has been provided with (Sanofi, 2019). AI algorthms can of course contain errors that lead to serious consequences (Keshinbora, 2019). So, how can one explain an outcome to a patient if the outcome is based on a very complex system that cannot explain itself? In order to do that, we need AI systems that can explain other AI systems, called XAI (Sanofi, 2019). I believe that AI is not ready for large healthcare decisions yet until it reaches full reliability and until its choices can be explained. What do you think?
Sources:
Forbes. (2019) AI and Healthcare: A Giant Opportunity. [Online] Available from: https://www.forbes.com/sites/insights-intelai/2019/02/11/ai-and-healthcare-a-giant-opportunity/#2fa19d5b4c68 .
Keskinbora, K. (2019) Medical ethics considerations on artificial intelligence. Journal of Clinical Neuroscience. 64(6), 277-282.
Kim, W. & Holzberger, K. (2019) What AI ”App Stores” Will Mean for Riadiology. Harvard Business Review. [Online] Available from: https://hbr.org/2019/06/what-ai-app-stores-will-mean-for-radiology .
Sanofi (2019). The Ethics of AI in Healthcare. [Online] Available from: https://www.sanofi.com/en/about-us/our-stories/the-ethics-of-ai-in-healthcare .
I totally agree for the big health care decisions. However, I do think AI systems can already make a major contribution by for instance, preventive interventions, which can be achieved by predicitve diagnostics. By preventing a problem instead of solving a problem, AI does not have to be involved in a big decisions, but can have a big impact in the healthcare.
Great topic Loes! I think the healthcare industry should first be optimized fully for AI is used. Nowadays, patient dossiers are handled so inefficiently still. Robotic Process Automation (RPA) carries great potential to streamline processes, like administrative processes around patient dossiers. A lot of error in patient analysis is still made because of this inefficient use of patient data from dossiers. If these errors and activities are firstly optimized and coherent the next step will be AI. RPA in combination with AI is able to cause a lot of disruption in the healthcare industry!
Hi Loes,
A very interesting post! Unlike the business world, where failure in the worst case scenario will lead to bankruptcy, failures in the medical world could cost human lives. So regarding the trustworthiness of A.I. in Healthcare (radiology), I would argue that radiologists need to see A.I. as a helping tool for diagnosing diseases, instead of a tool that diagnoses diseases itself. A radiologist is always responsible in my opinion. Considering the fact that A.I. learns from deep learning, there will always be a risk of a ‘bias’ in the data (in case of unusual data points), which results in a decrease in accuracy of the analysis or diagnosis. Therefore, I agree with your opinion that A.I. currently is not ready to be solely responsible for decisions made in Healthcare.
Nevertheless, I think that A.I. could enhance the speed of diagnoses on the short term. I found this interesting company called Zebra Medical Vision, which is working on implementing A.I. for mammography, bone health and triage (deciding the order of treatment of patients).This is their website: https://www.zebra-med.com/solutions/mammography/ (in case you’re interested). Zebra’s founder argues that their A.I. mammography will solve the large amount of false diagnosis and, consequently, will save a lot of unnecessary healthcare costs.
Kind regards,
Rowan
Hi Loes,
A very interesting article indeed. AI definitely has the potential to diagnose human I noted future. There are already applications that can determine if one has skin cancer based on images of the coercing area of the body. However, I cannot help but point that AI in healthcare has enormous benefits outside of the diagnosis area. For instances hospital bed management is a known problem in the healthcare industry. Having an AI that can optimise the placement of patients will help achieve efficiency use of space and resources. Which may help reduce costs. Perhaps the area that AI should target in the medical field is optimisation of processes rather than diagnostics at first?
Let me know your thoughts.
Kind regards,
Roma