Nowadays, Artificial Intelligence (AI) is much-discussed and attracts the interest of many people from different fields. Its developments are exciting and rapid, as can be seen by the new Wunderkind on the block, GPT-3.
Its technologies are being applied in an increasing number of industries. Despite this, it’s often not understood what AI is. To shed some light on the topic, we will explain what AI is and the different useful types for the healthcare field.
1) What actually is Artificial Intelligence?
To start from the beginning, as Deloitte stated in one report, there is not a single definition of AI as there are many ways to approach it. However, it is basically the science of designing ‘smart machines’ which have a set of key features, namely:
- Capable of processing a large amount of data in a very short time
- Ability to learn based on given information
- Deductive and inductive reasoning
- Problem solving and analyzing.
Further, just as a human can specialize in a specific field of interest, an AI system can be designed to handle a specific task. This type of system is defined as a Narrow AI. Other systems are referred to as General AI, or when the system ‘can learn about any problem and then solve it’ according to Deloitte.
2) AI in the healthcare
As we highlighted in the report “AI, Privacy, and Genomics: The Next Era of Drug Design”, AI is becoming more and more relevant in the healthcare field. Increasing complexity and the constant need for medical care have made this a primary sector of innovation.
‘Coronavirus has shown that the drug discovery and development cycle can be significantly accelerated in situations of urgency, has showcased the value of mining big databases for drug discovery, and highlighted the importance of data sharing and collaboration’.
However, this is not the only healthcare sector in which AI is taking hold, as it has multiple related technologies. Thanks to its algorithms and data-driven approach, AI tech can support human labour by accelerating and performing other key healthcare tasks.
There is a wide range of AI solutions already employed in healthcare, and over time, they are becoming increasingly integrated. Here below, we listed some of the most relevant AI technologies currently employed in the field and their related core use-cases.
- Machine learning (ML)
ML is a fairly common basis of AI systems and it ‘is a statistical technique for fitting models to data and to “learn” by training models with data’ as Davenport and Kalakota explain. There are different versions of ML. The traditional one is typically used in Precision Medicine, that as Tamás Török, Head of Business Development of Turbine, explained to us, allow us to ‘identify novel molecular targets to overcome the disease and precisely select patients for whom the therapies will work best’.
More complex versions are neural networks and deep learning. Those versions are used for ‘categorisation applications’, speech recognition or radiomics.
- Natural language processing (NLP)
This field contains several applications related to human languages, such as speech recognition or translation. ‘In healthcare, the dominant applications of NLP involve the creation, understanding and classification of clinical documentation and published research’, according to Davenport and Kalakota.
- Rule-based expert systems
Those systems were mainly used in the 1980s to support clinical decisions. However, they are now being replaced by ML and data-driven algorithms. They are based on ‘if-then’ rules related to a specific knowledge domain and constructed by human experts.
AI enabled the creation of physical and smart machines used to ‘perform pre-defined tasks like lifting, repositioning, welding or assembling objects […] and delivering supplies in hospitals’ Davenport and Kalakota continue.
- Robotic process automation (RPA)
This AI technology is likely the easiest one to program and use. It performs administrative tasks as a ‘semi-intelligent user’ of the information system.
AI solutions can be transformative for healthcare if their developments are driven by clinical needs and experts’ inputs. According to Mark-Jan Harte, CEO of Aidence, AI technologies and models can manage tasks previously delegated to humans, but they can’t necessarily replace human labourers. As he puts it:
‘An infamous 2016 quote argued that we should stop training radiologists as they will be replaced by autonomous machines, a statement that has raised concerns in healthcare. Not only did the prediction not come true, but there is a shortage of radiologists and their role is only becoming more important. Nor are AI systems anywhere near full autonomy’.
3) Areas for future developments
Over time, more and more tasks will be performed by computers, but the rise of ‘intelligent services’ — services provided by machines which have artificial intelligence capability — will be the real novelty. AI technologies are going to have a big role in it, as it will be more integrated into the existing workflow and healthcare system.
One area in which AI will impact the future of healthcare is drug discovery. Our latest report explains how drug R&D will benefit from AI solutions, leveraging its capability of processing a large amount of data. As Maxim Kholin, Co-Founder & Business Development Director of Gero,stated:
‘The data-driven approach should help establish the genetic determinants and molecular markers of the disease. Repurposed with the help of AI drugs could serve for prompt PoC studies and as starting points for de-novo drug discovery’.
Additionally, according to Davenport and Kalakota, AI will address the issues related to patient engagement and adherence, leading to better health outcomes. AI solutions might aid doctors to provide more personalised cures through specific applications that contact and alert patients with contents relevant to their treatment. While the use of AI technologies in this field may seem less revolutionary, it can bring a lot more efficiency.
Finally, AI has made rapid advances in imaging analysis. Thanks to its ability to learn based on given information, deep learning technologies are being applied to categorize and describe images and digital data. The application of AI in health sectors using digital data, such as radiology and pathology, might make medical care more affordable.
Overall, while our healthcare is improving, the need to improve services for patients and working conditions for healthcare providers leaves room for innovation. Due to the complexity of these systems, intelligent solutions are necessary. As innovation continues, we hope to see widespread adoption of those technologies into these sectors.
Thank you to everyone who contributed quotes. To learn more about AI in drug discovery, read our full report, “AI, Privacy, and Genomics: The Next Era of Drug Design”.