AI in Healthcare Industry

Artificial Intelligence is proving its prominence in every industry out there and the healthcare industry is no different. From patient care to Administrative processes AI has huge potential in the healthcare industry. There are many research studies suggesting that AI can perform as well as or better than humans at key healthcare tasks. We have seen robots performing surgeries or assisting doctors with more precision and flexibility. Algorithms are outperforming radiologists in detecting dangerous tumors and advising researchers on how to build cohorts for expensive clinical trials. However, I believe that it will be many years before AI replaces humans in broad medical process domains for a multitude of reasons. Let’s look at the potential for AI to automate aspects of care as well as some of the barriers to rapid AI implementation in healthcare.

Machine Learning:

The most common application of traditional machine learning in healthcare is precision medicine, which predicts which treatment protocols are likely to be successful on a patient based on various patient attributes and the treatment context. The vast majority of machine learning and precision medicine applications require a training dataset with known outcome variables (for example, disease onset); this is known as supervised learning.

The neural network is a more complex form of machine learning that has been available since the 1960s and has been well established in healthcare research for several decades and has been used for categorisation applications such as determining whether a patient will acquire a specific disease. It considers problems in terms of inputs, outputs, and variable weights or ‘features’ that link inputs to outputs. It has been compared to the way neurons process signals, but the analogy to brain function is weak.

Deep learning, or neural network models with many levels of features or variables that predict outcomes, is one of the most complex forms of machine learning. The faster processing of today’s graphics processing units and cloud architectures may reveal thousands of hidden features in such models. Recognizing potentially cancerous lesions in radiology images is a common application of deep learning in healthcare. Deep learning is increasingly being used in radiomics, or the detection of clinically relevant features in imaging data that go beyond what the human eye can see. Oncology-oriented image analysis frequently employs both radiomics and deep learning.

Natural Language processing:

Since the 1950s, AI researchers have sought to understand human language. NLP applications include speech recognition, text analysis, translation, and other language-related goals. There are two approaches: statistical NLP and semantic NLP. Statistical NLP is based on machine learning (particularly deep learning neural networks) and has contributed to a recent increase in recognition accuracy. It is necessary to have a large ‘corpus’ or body of language from which to learn.

The most common applications of NLP in healthcare involve the creation, comprehension, and classification of clinical documentation and published research. NLP systems can analyze unstructured clinical notes on patients, prepare reports (for example, on radiology exams), transcribe patient interactions, and perform conversational AI.

Rule-based expert systems:

Expert systems based on collections of ‘if-then’ rules were the dominant AI technology in the 1980s, and they were widely used commercially at the time. They were widely used in healthcare for ‘clinical decision support’ purposes over the last few decades and are still widely used today. Today, many electronic health record (EHR) providers include a set of rules with their systems.

Human experts and knowledge engineers are needed to build a set of rules in a specific knowledge domain for expert systems. They work well up to a point and are simple to grasp. However, when there are a large number of rules and the rules begin to conflict with each other, they tend to break down.

Diagnosis and treatment applications:

Many healthcare organizations are plagued by AI implementation issues. Although rule-based systems are widely used in EHR systems, including at the NHS, they lack the precision of more algorithmic systems based on machine learning. These rule-based clinical decision support systems are difficult to maintain as medical knowledge evolves, and they are frequently incapable of dealing with the explosion of data and knowledge based on genomic, proteomic, metabolic, and other omic-based’ approaches to care.

This situation is changing, but it is more prevalent in research labs and tech firms than in clinical practice. Almost every week, a research lab claims to have developed a method for using AI or big data to diagnose and treat diseases with equal or greater accuracy than human clinicians. Many of these discoveries are based on radiological image analysis, though others include retinal scanning and genomic-based precision medicine. Because these types of findings are based on statistically-based machine learning models, they usher in a new era of evidence- and probability-based medicine, which is generally regarded as positive but poses many challenges in medical ethics and patient/clinician relationships.

Patient engagement and adherence applications:

Patient engagement and adherence have long been regarded as the healthcare industry’s “last mile” problem — the final barrier between ineffective and good health outcomes. The more patients take an active role in their own health and care, the better the outcomes — utilization, financial outcomes, and member experience. Big data and artificial intelligence are increasingly being used to address these issues.

Providers and hospitals frequently use their clinical expertise to create a plan of care that they know will improve the health of a chronic or acute patient. However, this is often ineffective if the patient fails to make the necessary behavioral changes, such as losing weight, scheduling a follow-up visit, filling prescriptions, or adhering to a treatment plan. Noncompliance is a major issue when a patient does not follow a course of treatment or take the prescribed drugs as directed.

The future of AI in healthcare:

I believe that AI will play an important role in future healthcare offerings. It is the primary capability driving the development of precision medicine, which is widely acknowledged to be a much-needed advance in care. Although early efforts to provide diagnosis and treatment recommendations proved difficult, we anticipate that AI will eventually master that domain as well. Given the rapid advancements in artificial intelligence for imaging analysis, it appears likely that most radiology and pathology images will be examined by a machine at some point. Speech and text recognition are already used for tasks such as patient communication and clinical note capture, and their use will grow.