Artificial intelligence and machine learning can be used to interpret medical test results. Robots studying holographic hearts could cease to be merely a concept. (Sergii Iaremenko/Science Photo Library)
The popular travel aggregator Expedia is able to search three million hotels in a second to deliver a selection of personalised results, thanks to the wonders of artificial intelligence (AI) and machine-learning technologies.
The emergence of supercomputing, however, also has profound implications for the advancement of technological interventions in healthcare, assisting medical professionals with early detection, diagnosis and treatment.
The next-generation supercomputer MareNostrum4 is used for scientific research. Powered by Lenovo’s ThinkSystem servers and located at the Barcelona Supercomputing Centre (BSC) in Spain, it can process a mind-boggling 13 677-trillion operations a second.
Scientists at the BSC were able to analyse a sample of 20 000 human genes with a six-billion sequence combination and, from the 120-trillion data points, were able to isolate chronic lymphocytic leukaemia mutations down to just four genes. Other research carried out at the BSC was to improve the accuracy of retinal screening processes by using AI to enable early detection of retinal diseases.
Rick Koopman, an AI technical leader at Lenovo, explains that AI could play a huge role in healthcare segments that rely on the interpretation and processing of images such as X-rays, MRIs, or CT scans.
“An image is nothing more than a number of layers of ones and zeros on top of each other. Computing capabilities are extremely capable of number-crunching, that’s why anything that is image-related is of huge importance for AI in healthcare,” says Koopman.
If you go to see a specialist, there are long waiting times because they have to analyse everything beforehand, and find what’s not supposed to be there, in order to talk a patient through potential treatment plans, he says. “What about patients who don’t feel well but don’t have a brain tumour?”
“If you teach an algorithm to recognise content on images that shouldn’t be there, then it is able to recognise it, pull it out, and qualify it, and if you provide enough metadata, it will qualify what it is and prioritise it for the doctor, and even suggest a treatment plan.”
According to Pricewaterhouse Coopers, technologies such as IBM’s Watson for Healthcare are helping organisations to apply cognitive technology to unlock health data and power diagnosis, by reviewing and storing existing medical data, symptoms and case studies of treatments faster than any human.
Google’s DeepMind Health is working with clinicians, researchers and patients to solve real-world problems by combining machine learning and neuroscience to build algorithms into neural networks that mimic the brain.
Apart from the time and cost savings, Koopman says nothing is going to change for the specialist. “They have a crucial role in our healthcare system. AI is meant to augment their capabilities and give them reliable tools with a predictable outcome, and improve the quality of service. The robots are not going to take over.”
Koopman admits that a mis-diagnosis could occur, but it would depend on the data given and not on the algorithm. “It depends on the information that we have been feeding the algorithm during the learning phase, so if you train an algorithm with 10 images, the quality of your outcome is not going to be as good as if you feed it with a million images, and with metadata.”
Although Lenovo is working with a number of academic and research institutes, such as the BSC, to optimise and improve areas within the healthcare industries, none of the innovations can be implemented yet as regulations have not been put into place by European law.
The company’s technology in the high-performance computing space earned recognition in February 2016 (with the University of Birmingham) for its role in tracking and identifying the origin of the Zika virus.
AI and cognitive computing are expected to generate savings of more than $150-billion for the healthcare industry by 2025, and the market is expected to grow to $6.16-billion by 2022, according to market research company Frost & Sullivan.
The research firm says the uptake of AI in healthcare has been slow because of strategic and technological challenges and, to date, only 15% to 20% of end users have actively used AI to effect real change in the way healthcare is delivered.
But it says the status quo is going to change dramatically in the next three to five years because the democratisation of AI is now being made possible by big IT companies such as IBM Watson Health, Microsoft, Google, Philips, GE Healthcare, Amazon and Salesforce, which are offering cost-effective infrastructure support to modular and specialty-specific vendors.
Koustav Chatterjee, an industry analyst at Frost & Sullivan, says: “To be successful, healthcare IT providers need to devise AI-based business models that fetch real benefits in the form of tangible return on investment to end users.
“One must realise that patient-generated data, which AI platforms interpret, has multiple utilities for diverse healthcare stakeholders. Fully informed consent from patients, coupled with 100% compliance with stringent data usage regulation, has to be ensured to remain relevant in the market.”