iDoctor

By collecting and analysing data, treatments and technologies can be developed much faster, knowledge about rare diseases can be more easily spread, and epidemics can be detected and fought faster. However, all these new computeraided possibilities still require the processing, control and evaluation by doctors, pharmacists or other scientists. This naturally limits the speed of the evaluation processes. In addition, the new methods have created a flood of information that one person alone can no longer handle. This was partly compensated by the specialisation of doctors. The general practitioner, who used to be able to do almost everything, can only pass on to specialists.

In addition to today’s advanced computer technology, there are new possibilities of producing micro-components, such as sensors or cameras, cheaply and in large quantities. This has made the internet of things (IoT) possible – every device, room and wearable can now generate data, and upload it to the cloud. All of these IoT components are networked together, and can communicate and interact with each other. Self-learning algorithms record and analyse data from the worldwide web in real time, draw conclusions and make decisions. Intelligent systems become possible.

Define AI

AI is a machine’s ability to make decisions and perform tasks that simulate human intelligence and behaviour, such as visual perception, speech recognition, decision-making and translation between languages. A more elaborate definition characterises AI as a system’s ability to correctly interpret external data, to learn from such data, and to use those learnings to achieve specific goals and tasks through flexible adaptation.

What distinguishes AI technology from traditional technologies in healthcare is the ability to gain information, process it and give a well-defined output to the end user. AI does this through machine-learning algorithms, which can recognise patterns in behaviour and create their own logic. It is not surprising that, besides medical institutions, large IT companies such as IBM and Google have developed AI algorithms for healthcare.

History

Research in the 1960s and 1970s produced the first problem-solving program, or expert system, known as Dendral, designed for applications in organic chemistry. It provided the basis for a subsequent system, MYCIN, a knowledge-based consultation programme for infectious disease diagnosis, considered one of the most significant early uses of AI in medicine. However, MYCIN and other systems did not achieve routine use by practitioners.

The 1980s and 1990s brought the microcomputer and new levels of network connectivity. During this time, there was recognition by researchers and developers that AI systems in healthcare must be designed to accommodate the absence of perfect data. Approaches involving fuzzy logic and artificial neural networks were applied to intelligent computing systems in healthcare. Several new advanced technologies were developed in this period. For example, there were improvements in computing power, resulting in faster data collection and data processing. Increased volume and availability of health-related data from personal and healthcarerelated devices occurred.

In addition, there was dramatic growth of genomic sequencing databases, widespread implementation of electronic health-record systems, improvements in natural language processing and computer vision as well as massively enhanced precision in robot-assisted surgery. All these developments led to a number of useful applications for AI within health treatment.

The power

Professor Holger Hanssle from the University Hospital of Heidelberg, was one of the first to publish a study that showed in direct comparison who made better diagnoses, humans or machines. Specifically, which can better recognise the dangerous black skin cancer. “AI was significantly better than the average performance of the doctors,” reported Hanssle. “Only 13 of the 58 dermatologists involved were able to beat the algorithm. And they were the world’s top experts.”

For this experiment, Hanssle and his colleagues had trained an artificial neural network with 100,000 photos showing black skin cancer or harmless birthmarks. In addition, they gave the AI the correct diagnosis. Thereafter, the neural network was superior to an average dermatologist – but only for this specific question.

Applications

Correctly diagnosing diseases takes years of medical training. Even then, diagnostics is often an arduous, time-consuming process. In many fields, the demand for experts far exceeds the available supply. This puts doctors under strain and often delays life-saving patient diagnostics.

Machine-learning has recently made huge advances in automatically diagnosing diseases, making diagnostics cheaper and more accessible. Machine learning algorithms can learn to see patterns similarly to the way doctors see them. Examples where the advantages of AI has been shown already include detecting lung cancer or strokes based on CT scans, assessing the risk of sudden cardiac death or other heart diseases based on electrocardiograms and cardiac MRI images, classifying skin lesions in skin images and finding indicators of diabetic retinopathy in eye images. In comparison with a human doctor, the algorithm can draw conclusions in a fraction of a second, and it can be reproduced inexpensively all over the world.

AI will also be an important part of telemedicine. The ability to monitor patients 24/7 using AI allows sending information to physicians when possible disease activity occurs. A wearable device allows constant monitoring of a patient and also notices changes that may be less distinguishable by humans.

Different patients respond to drugs and treatment schedules differently, so personalised treatment has enormous potential to increase patients’ lifespans. But it’s hard to identify which factors should inform the choice of treatment. Machine learning can automate this complicated statistical work – and help discover which characteristics indicate that a patient will have a particular response to a particular treatment. The resulting outcome predictions make it much easier for doctors to design the right treatment plan.

The medicine of the future will also be determined by the creation of artificial organs. With 3D printing, more components of the human body will be produced. There are already adapted implants – serious facial injuries can be treated with digital diagnostics, reprocessing of models and the production of individually manufactured implants. Even more is possible with so-called bioprinting, which can produce ear pinnae, bladder or skin; one day it will be kidney, liver and heart.

Implications

The use of AI is predicted to decrease medical costs as there will be more accuracy in diagnosis and better predictions in the treatment plan as well as more prevention of disease.

Virtual nursing assistants are predicted to become more common, and these will use AI to answer patients’ questions and help reduce unnecessary hospital visits. They will be useful as they are available 24/7, and are already able to give rough wellness checks with the use of AI and voice.

Already, there is evidence that the use of chatbots leads to positive outcomes in the field of mental health. Other future uses for AI include braincomputer interfaces (BCI), which are predicted to help those with problems in moving, speaking or with a spinal cord injury. The BCIs will use AI to help these patients move and communicate by decoding neural signals.

Ethics

The use of AI in medicine also raises social, legal and ethical issues. Specifically important is data protection, but also responsibility and transparency. Therefore, the German Government promotes research projects on ethical, legal and social aspects of digitisation, big data and AI in health.

Is everything that is possible, ethically acceptable? Everything that can be digitised is digitised – should this paradigm also apply to medicine in world 4.0? What can the medicine of the future afford? Where are the limits? Is technical progress really compulsory and unstoppable in all areas of life?

At the moment, AI systems are used as support for physicians. Ultimately, the last decision lies with the doctor. But does it always have to be this way? If the self-learning, ever-improving ‘Dr AI’ creates its diagnoses and therapies so well, does it make human doctors unnecessary?

Data protection

AI systems could be of great help in medicine, diagnosing, preventing and prolonging lives. But one issue is that AI algorithms require huge amounts of data to learn. But medical records, genetic information and electronic patient records can contain sensitive personal information.

That’s why researchers at Stanford Medical School, University of California, Berkeley, its spin-off Oasis Labs and scientists at ETH Zurich are working to secure the learning processes so that the data cannot be leaked or otherwise misappropriated. In Stanford, the data always stays within the Oasis cloud. Outsiders can inject algorithms and get results without the data itself ever leaving the system.

To ensure the protection and respect of values such as freedom, privacy, sovereignty, charity, justice, solidarity and responsibility, even under big-data conditions, the German Ethics Council recommended a design and regulation concept based on selfdetermination and data sovereignty.

According to the Gartner hype cycle, AI is currently experiencing a hype. Many articles praise it as either a solution to all the problems in medicine or as the beginning of a time where the machines take over power.

But we are facing a phase of disillusionment. ‘Dr Watson fails’ headlined Der Spiegel in an issue on the use of AI in medicine. It is to be expected that the media will report the tragic consequences of wrong decisions of the AI oversized and scandalising.

But over time, the use of AI will be just as normal and indispensable as the use of electric power. We cannot and do not want medical staff to do things that computers can handle better and faster.