Skip to main content

Over the last couple of years, we have intensified focus on applying artificial intelligence (AI) in many domains to address complex issues. Similarly, the acquisition of artificial intelligence in our healthcare has grown while dramatically shifting healthcare delivery’s face.

AI is being used in various settings such as clinical laboratories, research facilities, and hospitals. It approaches using machines in sensing and comprehending information like a human, which has paved the way for previously unrecognized or unavailable opportunities for health service industries and clinical practitioners.

Key examples include using artificial intelligence approaches to analyze unstructured information like videos, photos, physician notes enabling clinical judgment, and prognostic modeling for managing hospital resource/capacity allocation and patient flow.

However, there is still not a clear overall understanding of the implications of using AI, more so for clinicians.

Types of Relevant AI to Healthcare

Artificial Intelligence cannot be seen as a single technology but is instead a collection of technologies. The majority of these technologies have direct significance to the field of healthcare. But for particular tasks and processes, the support ranges widely.

Some specific Artificial Intelligence of great relevance to healthcare is outlined below:

Machine learning – deep learning and neural networks

Machine learning is a statistical method of fitting models to information and learning through training models having data. It is among the most prevalent forms of AI. According to a Deloitte survey conducted in 2018 on 1,100 US managers with companies that already adopted AI, 63% of organizations surveyed were using machine learning in their trades. It is a vast technique at the center of various AI approaches. There also exists a variety of versions to it.

Processing natural language

Since the 1950s, creating sense out of human language has been a target of AI researchers. NLP encompasses applications like text analysis, translation, speech recognition, among other language-related goals.

Two primary approaches to it exist. These are semantic and statistical NLP. Statistical NLP is found in machine learning and has promoted the recent growth in accuracy recognition. A large body or ‘corpus’ is required to learn from.

Rule-based professional systems

Professional systems found on collections of ‘if-then commands were the prevalent AI technology for the 1980s. They were extensively used commercially during that and later years.

They were broadly used in healthcare for ‘clinical decision support in the recent past and are still widely in use to date. For example, several electronic health record providers currently use their systems to design a set of rules.

Treatment and diagnosis applications

Disease treatment and diagnosis have been a point of interest for AI since around the 1970s when Stanford developed MYCIN to diagnose blood-borne bacterial infections.

This, along with other early rule-found systems, looked promising concerning accurate diagnosis and treatment of diseases. However, they were not embraced for clinical purposes. Moreover, they were not considerably better than human diagnosticians and were poorly integrated with medical record systems and clinician workflows.

Unifying Machine and Mind via Brain-Computer Interfaces

Communication via computers is not new. However, designing direct human mind and technology interfaces that do not require mice, monitors, and keyboards is a revolutionary research area with critical applications for patients.

Nervous system trauma and neurological diseases can deny some patients the ability to move, speak, and have meaningful interactions with their environments and other people. Brain-computer interfaces (BCIs) supported by AI have the power of restoring such fundamental experiences to people who thought they had lost them forever.

Leigh Hochberg, MD, Ph.D., Center for Neurotechnology and Neuro recovery at MGH director, previously mentioned that if he were to be in a neurology ICU, and a patient who had suddenly lost the ability to speak or move came in, the desire would be to restore the person’s communication ability by the next day.

He continued to say that through artificial intelligence and BCI, neural activities can be decoded together with the desired movement of a person’s hand. Hochberg added that we should have the ability to allow the person to communicate in a similar manner to the majority of people in the room by use of ubiquitous communication devices such as a phone or tablet.

Brain-computer interfaces have the potential of significantly boosting the quality of life for strokes, locked-in syndrome, or ALS patients. Five hundred thousand people worldwide who have had spinal cord injuries each year also stand to improve their quality of life.

Next-Generation Radiology Tools Development

Radiological images realized from MRI machines, X-rays, CT scanners provide a non-invasive human body’s inner workings visibility. However, many characteristic processes continue relying on physical tissue specimens acquired through biopsies, which are risky involving the probability of infection.

Experts predict the next radiology tools generation will be detailed and accurate enough to oust the necessity of tissue specimens through artificial intelligence in some cases.

Alexandra Golby, Brigham & Women’s Hospital (BWH) MD, Director of Image-Guided Neurosurgery, said they had the desire of unifying the interventional radiologist with the pathologist or surgeon with the diagnostic imaging team. This was coming together of various teams and alignment of goals, she said was a great challenge. A challenge that could be addressed by the use of AI.

Golby also mentioned that if we desire to obtain information currently derived from tissue specimens through imaging, then the ability to achieve very close registration will be required. This is in order to know a given pixel’s ground truth.

Conclusion

AI plays a crucial role in the present future healthcare offerings. Machine learning is a key factor in precision medicine development. It is widely acknowledged to be necessary as far as healthcare is concerned.

Although preliminary attempts at offering treatment recommendations and providing diagnosis have proven challenging, the expectation is that AI will eventually master the domain.

With the rapid AI advances for imaging analysis, most pathology and radiology images will likely be examined at some point by machines. Text and speech recognition is already being used for patient communication and capturing clinical notes, and their engagement is bound to increase.