Artificial intelligence shows potential for filling these gaps
These gaps have been growing for decades, and the promise of technology filling them is imminent with digital health becoming widespread. Authors of this essay argue that AI might not only fill the human resource gaps, but also raise ethical questions we need to assess today.
As Nick Bostrom describes in his book Superintelligence, AI is divided broadly into three stages: artificial narrow intelligence (ANI), artificial general intelligence (AGI) and artificial superintelligence (ASI) [6]. In the next decade, ANI has the highest chance of being used in the medical practice for analyzing large datasets, finding new correlations and generally supporting caregivers’ jobs.
An obvious first step is clearing up the definitions around AI to stop its misuse in medical communication. Here we attempt at providing short definitions for the most common expressions.
Artificial narrow intelligence
It is good at performing a single task, such as playing chess, poker or Go, making purchase suggestions, online searches, sales predictions and weather forecasts.
Artificial general intelligence
It can understand and reason its environment similarly as a human being would do, therefore it’s also known as human-level AI.
Artificial superintelligence
According to Nick Bostrom, it’s smarter than the best humans in every field from scientific creativity to general wisdom and social skills.
Supercomputers
A supercomputer is a computer with a high level of computing performance used for resource-intensive tasks, such as machine and deep learning.
Machine learning
Machine learning is one of the many subsets of AI that refers to creating programs based on data as opposed to programming rules. A software that learns from large sets of relevant data (e.g. feeding it with a lot of radiology images and letting it discover recurring patterns).
Deep learning
It is a specialized subset of machine learning that uses neural networks, an artificial replication of the structure and functionality of the brain. It’s efficient at various tasks such as image recognition, natural language processing and translation. The performance of deep learning algorithms continues to improve as datasets grow significantly which means the bigger the dataset, the better the outcome and efficiency improves.
Various companies and organizations have already demonstrated how AI can contribute to improve the quality of care and/or decreasing costs [7].
Deepmind Health launched a cooperation with the Moorfields Eye Hospital NHS Foundation Trust to improve eye treatment by mining one million anonymized eye scans with the related medical records. IBM launched Watson Oncology to provide clinicians with evidence-based treatment options and an advanced ability to analyze the meaning and context of structured and unstructured data in clinical notes and reports.
In the Netherlands, Zorgprisma Publiek helps caregivers and hospitals avoid unnecessary hospitalizations of patients by analyzing the digital invoices obtained from insurance companies with IBM Watson in the cloud.
In radiology, the Medical Sieve project aims at building the next-generation “cognitive assistant” with analytical, reasoning capabilities and a range of clinical knowledge. Such an assistant would be able to analyze radiology images to detect medical issues. In genomics, Deep Genomics helps identify linkages to diseases in large data sets of genetic information and medical records.
In pharmaceutical research, Atomwise uses supercomputers to find new therapies speeding up clinical trials that take sometimes more than a decade and cost billions of dollars. As an example, Atomwise found two drugs predicted by the company’s AI technology which may significantly reduce Ebola infectivity in less than a day of research, instead of years.
Deep learning algorithms have demonstrated to be able to help the diagnosis of conditions in cardiology, dermatology and oncology [8, 9].
Arterys already received FDA clearance for its AI-assisted cardiac imaging system in 2017. AI supported messaging apps and voice controlled chatbots can also help take off the burden on medical professionals regarding easily diagnosable health concerns or quickly solvable health management issues. Safedrugbot is a chat messaging service that offers assistant-like support to health professionals who need appropriate information about the use of drugs during breastfeeding.
AI-based services could facilitate more accurate diagnoses, administration, decision-making, big data analytics, post-graduate education, among others. However, we need to emphasize that practicing medicine is not a linear process. Every single element and parameter cannot be translated into a programming language. Moreover, there is no clinical trial or peer-reviewed data about the data points that contribute to a medical decision. It’s clear that AI is not the ultimate solution for all the challenges healthcare faces today. Although, in many areas, its use is inevitable and advantageous in supporting caregivers’ job.
However, a tight framework from regulatory agencies would further stop companies from providing false hope for patients claiming more than what they can deliver and prove. Moreover, the FDA has assembled a team of computer scientists and engineers to help oversee and anticipate future developments in AI-driven medical software [10]. These are encouraging steps forward, but the range of ethical, legal and social implications of using AI in healthcare are even beyond the scope of what we can deal with today.