AI in healthcare is making rapid strides. Artificial Intelligence (AI) has come a long way since the turn of the Millennium and accelerated exponentially in the last decade. Several factors have augmented the progress. Computing power has increased to help develop more complex solutions, and the cloud has ensured those solutions can be scaled. There are also vast amounts of data available as people adopt digital channels.
Consumers mainly see AI in the Internet of Things (IoT) devices like smartphones, wearables, or voice-activated systems. However, industries that have previously been resistant to technology are taking advantage of the trust gained from those applications and embedding it in their processes. The healthcare sector is one that has always found barriers to innovation but is slowly starting to catch up.
The Internet of Medical Things (IoMT) is unlocking the colossal amount of potential that has always existed with AI in healthcare. In this article, we will look at the applications of AI in the industry and the fantastic benefits it holds.
Applications of AI in Healthcare
Accenture forecast that the healthcare AI market will be worth £6.6 billion by 2021 with a 40% CAGR. Within the same report, 83% of executives said they saw a vital competitive advantage from adopting AI in the sector. Although below, we talk about some of those applications, it is essential to note that there are significant challenges when attempting to deploy AI in an industry that handles vast amounts of personal data, like healthcare.
Gaining Trust
The public is happy to let AI tell them what to buy on Amazon, but there is some natural reluctance in allowing a machine to advise on their health. People would still instead seek help from a human doctor than trust a computer-based one. When people hear about AI causing Uber cars to accidents, regardless of the context, it rings alarm bells.
AI and machine learning are complex to understand, and there needs to be a process of gaining trust before full acceptance.
Clinical Jobs
A survey on AI for automation found that 21% of healthcare workers were concerned about their job due to machines “taking over.” For healthcare institutions looking to invest in AI tools, it needs to be made clear that they are designed to augment human professionals and not replace them. For example, AI that can scan millions of rows of data helps a doctor make a faster decision so they can spend time caring for the patient, not being an analyst.
Data
Medical data is deeply personal information. People do not want to share this data willingly, meaning healthcare services must find ways to present the social benefits of doing so. For example, if we’re going to create AI for detecting cancer, there needs to be a massive amount of data ingested into the platform to ensure it is accurate enough to make the right decisions.
In July 2018, Japanese researchers ran an experiment that used AI to detect stomach cancer in patients. It started by using only 100 images of those with symptoms and 100 of those without to help learn patterns associated with the condition. In 0.004 seconds, the AI detected early signs at an 80% accuracy. This was a considerable breakthrough given such early alerts are very tough to detect for doctors, often misconstrued as inflammation.
However, to build confidence in a solution, there needs to be thousands or millions of data points available for the algorithms to use. Sourcing that data through regulation and data privacy acts like GDPR and CCPA is a real challenge and one that needs to be addressed from the outset of any tools being produced.
How AI is being used in healthcare today
There are several applications of AI in the healthcare industry today. The tools are slowly gaining public trust and showing how such methods can be hugely advantageous. Here are some of the primary use cases that are already out in the real world.
Virtual Assistants
Sensely has developed a virtual nursing assistant called Molly. The smart machines can handle many of the repetitive tasks currently associated with nursing, such as monitoring vitals, symptoms, and care plans. For example, instead of a nurse needing to walk through wards to do patient observations, alerts can be sent directly to them when there is a probably through Molly.
Bluetooth-connected monitoring sensors send attributes like weight and blood pressure straight to monitors. They can pick up on signs of heart failure and other pulmonary conditions, amongst other things. When patients go home, the same devices can be used and alerts sent to the healthcare provider for a response if needed.
Medical Imaging
A medical facility in Texas has generated more than half a million medical images to augment the fight against cancer. The project leveraged computer vision. This is an AI technique that scans images and interprets them by converting the points it sees into data.
The technology takes potentially cancerous images, processes them through a series of complex algorithms (deep learning models), and finds indicators of cancer that might have otherwise gone unnoticed by healthcare professionals.
As we’ve already said, early case cancer symptoms are tough to spot, and thorough training on thousands of images, AI can teach itself to develop an accuracy beyond that of humans. As doctors can see signs of cancer in seconds, they can move straight on to treatment plans, creating a far better chance of recovery.
Dementia detection
According to studies, dementia costs around $500 billion worldwide per annum. Early detection of the onset of dementia could amount to significant cost savings later in life, both for healthcare providers and patients.
Cognetivity Neurosciences have created a platform technology for early, rapid, and easy detection of dementia. Participants take a cognitive assessment where they are shown a variety of images. The participants respond as quickly and accurately as possible during the test.
Once all the answers are received, the system ingests them into AI algorithms that compare the response to a vast dataset. The framework can cluster the respondent in terms of their performance and offer an accurate view of the early signs of dementia.
Managing beds
IBM Watson has worked with Glintt to create a platform known as WISEWARD. The sophisticated technology can predict at what point a patient will be discharged and their bed freed up. Hospitals can then have the insight required to allocate the right bed, to the right patient at the right time.
Using variables like gender, ward, and surgery type, WISEWARD is capable of predicting availability as far as a week in advance, using existing datasets to collate information.
There are clear benefits for hospitals here, but patients also get a far improved experience during their stay.
Fraud Reduction
The National Health Care Anti-Fraud Association estimates that healthcare fraud costs the US about $68 billion annually.
AI can be used to spot patterns in data and find anomalies. The behaviors could be signs of suspicious behavior, as they deviate from the norm. Such tools require a lot of data to be processed from past claims, budgets, and medical records to be useful, but the potential is there to resolve cases quickly or negate them altogether.
If the healthcare industry good even half the cost of fraud, there is $34 billion available to be put to better use, such as with patient treatment.
Repetitive jobs
The healthcare industry is full of documentation and administrative tasks. Clinicians must spend a lot of time reviewing case notes, studying records, and analyzing data to come up with the best form of treatment for a patient.
AI techniques like machine learning and natural language processing (NLP) can analyze documents in real-time. The data gathered can be compared to existing datasets and provide recommendations to the doctor on symptoms and/or how to best care for the patient.
Remote surgery
As computing power improves and connectivity gets stronger with 5G, it presents an opportunity for remote AI solutions. One of the most ground-breaking is remote surgery.
The application of 5G technology in healthcare would ensure that data transfer in real-time. There is no lag in receiving information at the other end, making it possible for surgeons to operate robots in remote locations.
The human surgeon in hospital A can guide the robotic surgeon in hospital B to complete a procedure without being physically present. Surgery requires a tremendous amount of precision and, as robots don’t get tired or need a break, there aren’t concerns about keeping a steady hand during lengthy procedures.
In time, combining robotics with computer vision, it may be able to perform surgery independently without any human involvement.
The future of AI in healthcare
This article has only scratched the surface of the possibilities of AI in healthcare. In an industry that has enormous amounts of data available, with time and experience, there is a possibility that we start seeing better-optimized care and treatment plans for patients. The ability to find even the tiniest details in images and medical data can perfectly augment the work of the healthcare professional and allow them to focus on helping people.
But before AI in healthcare is prevalent, mitigating the bias in artificial intelligence is a key challenge.
If the challenges that we laid out at the start of this article can be resolved, healthcare providers and institutions will be operating very differently by the end of the next decade.