AI in Healthcare – Separating the Hype from Reality
Artificial Intelligence or AI has emerged as the one technology that has the potential to completely change the world as we know it. This technology has already made its presence felt in other industries such as manufacturing, logistics, retail, etc. Now AI is trying to enter the healthcare segment to give it a complete overhaul and make it more efficient. We have already witnessed the rise of robotic-assisted surgeries. AI also has the potential to improve doctor-patient interactions and enhance hospital efficiencies with the use of self-learning algorithms that have the capability to perform clinical and administrative healthcare functions. AI is clearly emerging as a self-running engine for healthcare with the potential to create annual savings amounting to USD$150 billion by 2026 in the US healthcare ecosystem.
The current state of AI in healthcare
AI is a collection of technologies that enable machines to comprehend, learn, and act to augment a human activity. It is because of these capabilities that the AI healthcare market is expected to grow at an annual compound rate of 40% and cross USD $6.6 billion by 2021. While there are several use cases of AI in the healthcare market, like every other new technology, the implementation of AI in the healthcare scenario isn’t that simple or straightforward. While the images of machines replacing doctors might be doing the rounds of the internet, the reality of AI in action in healthcare is quite a far cry from that.
- Presently, most AI applications in healthcare have been utilized to improve clinician workflows and other administrative aspects such as faster claims processing…and AI definitely does have the potential to enable hospitals and the entire healthcare segment do more with fewer resources.
- It has immense capability to provide analytical support that can aid preventive medicine, and provide diagnosis support for doctors. H
However, the healthcare sector is quite diverse and encompassed many areas such as medical, pharma, healthcare insurance, patient care etc. providing immense opportunities to the technology.
At this given point in time, AI is mainly being used for it analytical and automation capabilities. So, today, for example, hospitals utilizing AI algorithms will automatically assess if a surgical room is properly stocked with the right inventory of supplies and will send an intimation to the right authority if it is not so. We have also heard of AI algorithms being capable of assessing the needs of the surgeon in the operation theatre and sending instructions to a robot to pick up the necessary supplies before a surgery commences. AI is being used to track clinical data and improve care delivery as well. However, the question that needs to be answered first is if we are ready to completely rely on the assessing powers of computers and be able to trust them entirely?
The challenges of AI
Despite all the progress that AI has made in healthcare, there is still a mountain of challenges that this progressive technology needs to scale. The first one being that of data.
Today we are surrounded by a sea of data. However, it is the quality of that data that is questionable. In order to make AI work, healthcare providers have to get legal access to dependable data to avoid false positives and improve diagnosis efficiency. Considering the healthcare segment is still fragmented and there is limited interoperability between hospital systems, inadequate measures to keep data current and relevant etc. the quality of the data in use becomes further questionable. Added to that is the problem of finding the right data in time, structuring and normalizing datasets in light speed and implementing fast-paced analytics to make this data work. Unless the data mining and processing are done right, leveraging AI to improve outcomes seems almost like a distant dream. It also should be highlighted that AI doesn’t just need huge volumes of data but also needs the understanding of the context in which it will be applied and who will use this. This understanding is quite nascent and immature in the healthcare segment.
This brings us to the regulatory challenge that AI has to scale. Regulations around AI are almost non-existent which makes it hard to design technical solutions that can be integrated smoothly into patient care and clinician practice.
We also have to consider the medical limitations of present-day AI, which inadvertently circle back to data as well. For example, in the case of image recognition and the use of machine learning and AI for radiology, there stands a risk of feeding the AI system with underlying bias along with the thousands of images since it also takes into consideration the subjective assumptions of the teams working on the algorithm.
AI, that is in use today, also has some technological challenges to mitigate. Presently AI that is in use is an amalgamation of several machine learning methods that help a system achieve intelligence in certain fields. So, we come across instances of intelligent programs defeating humans in certain tasks, such as that of the IBM supercomputer winning at a game of chess. However, unlike the human mind, these programs are not capable, at least at present, to create something like art, for example. At most these programs presently capably recognize patterns from blocks of text using natural language processing, or understand images or videos leveraging computer vision (something that is used extensively now in the field of medical imaging).
AI definitely possesses immense potential to give a complete makeover to the healthcare industry. However, for that, along with advancements in the field of AI technology, we have to overcome the challenge of disparate systems and introduce an environment of interconnectivity and interoperability between healthcare departments to produce data that is dependable and reliable. While the march of the machines has definitely begun, we are still a considerable time away from the age where machines are capable of abstracting concepts and enabling knowledge transfer within domains.
Curtis Langlotz, Professor of Radiology and Biomedical Informatics at Stanford University, very aptly compared the use of AI in healthcare with that of the autopilot in aviation. Just like how technology did not replace the pilots, it will be the same in the healthcare sector. Keeping the future in mind, it can be safe to say that while AI will definitely not replace the doctor, it does have the potential to replace doctors and healthcare systems that don’t use AI.