How AI Can Drive Innovation in Healthcare

For the best possible insights, we’ve combed through various scientific studies, examined plenty of statistics, and talked to experts in the field.

Hold on tight. This will be a captivating read.

Yet, six in ten Americans say they would be uncomfortable if their healthcare provider relied on AI to treat them.

Clearly, questions arise.

1. Is there anything to worry about, at least for the moment?
2. Can AI be truly helpful in treatment?
3. And how is it actually used today?

We’ll assess AI efficiency in healthcare using two criteria:

• Theory: research findings, AI prospects, and potential;
• Practice: real-world use cases, tangible benefits, and current challenges.

Theory

Researchers are now claiming that their AI-based approach can diagnose and treat diseases as accurately as a human clinician. Or even better.

Studies have shown AI’s ability to meet or exceed human experts’ performance in image-based diagnoses in radiology, dermatology, pathology, and cardiology.

For example, a study published in the UK used an AI system to diagnose breast cancer based on a large dataset of mammograms.

It showed that an AI-based mammogram interpretation system decreased false positive and negative results, which will potentially improve accuracy of breast cancer screening.

Another research used artificial intelligence to detect skin cancer.

It showed that AI technology had a 100% detection rate for melanoma and was 99.5% accurate in detecting all skin cancers.

Providers and payers for care also use ‘population health’ ML models to predict populations at risk of particular diseases or accidents.
These models can be effective, but they sometimes lack relevant data that adds predictive capability.

For instance, patient socio-economic status.

Practice

Indeed, automated classification of medical images proves to be useful in diagnostics and treatment.

In fact, it is the leading AI application today.

A recent review of AI/ ML-based medical devices approved in the USA and Europe from 2015 to 2020 found that 58% of devices in the USA and 53% in Europe were approved or CE marked for radiological use.

Let’s see how it works:

Imagine a computer tomography (CT) scan.

AI can find a very similar scan and tell the dynamics of changes based on these two pictures.

Now, instead of a few experts interpreting a scan, picture ten million radiologists doing so.

That’s basically how AI uses medical imaging to help reduce diagnostic mistakes by detecting anomalies a human might overlook.

What’s more, AI can learn from a wealth of data.

An AI tool that has learned all the ways a disease can present itself in an imaging study can quickly derive answers from new imaging studies.

And all this without getting tired.

A good example of a platform that uses AI is Nuance’s Precision Imaging Network. It interprets pictures and provides radiologists with recommendations.

We must say, recommendations is a keyword here.

Today, AI doesn’t produce an independent prognosis. No medical institutions fully delegate diagnostics to AI to avoid lawsuits.

Strictly speaking, it is used for more accurate measurements and highlighting spots potentially useful for the doctor.

There are more challenges to using AI in healthcare.

According to the scientific report, AI implementation issues are an uphill struggle for many healthcare organizations.

This situation is beginning to change, but AI is mostly present in research labs and in tech firms rather than in clinical practice.

Another challenge for AI adoption in healthcare is ensuring widespread use.

This requires regulatory approval, integration with EHR systems, standardization, clinician training, funding, and field updates.

Therefore, scientists claim that AI’s maturity will take longer, resulting in limited use within 5 years and more extensive use within 10.

Practice

This is the case where theory actually corresponds to the practice.

Here is how AI is used to optimize administrative tasks.

RPA (Robotic Process Automation) platforms are used for revenue cycle management, clinical recording, claims processing, medical record administration, and continuity care.

For example, they summarize discharge information and follow-up needs for post-acute care.They can also generate care summaries for referrals, and synthesize specialist notes for the primary-care physician team.

Machine learning is used in claims and payment administration for probabilistic data matching across databases.

As insurers have a duty to verify whether the millions of claims are correct, ML helps identify and correct issues faster and more effectively.

As a result, it saves all stakeholders (health insurers, governments and providers) a great deal of time, money and effort.

NLP systems are primarily used for clinical documentation and research.

Speech and text recognition are increasingly utilized in healthcare for tasks like patient communication and clinical note capture.

Practice

AI chatbots (e.g. Babylon and Ada) can identify symptoms and recommend further actions in community and primary care settings.

They can also be integrated with wearable devices like smartwatches to help patients and caregivers improve their behavior, sleep, and overall well-being.

For example, Emerald is a wireless sensor and ML platform for remote monitoring of sleep, breathing, and behavior.

It was founded by Massachusetts Institute of Technology faculty and researchers.

Yet, let’s remember that AI chatbots don’t provide you with a diagnosis.

Their suggestions only have a recommendation character, which they constantly remind us of.

Some healthcare organizations experiment with chatbots for patient interaction, mental health and wellness, and telehealth.

However, in a survey of 500 US users of the top five chatbots used in healthcare, patients expressed concern about revealing confidential information, discussing complex health conditions, and poor usability.

Common procedures performed using robots include gynecologic, prostate, head, and neck surgeries.

Ryan Cameron, vice president of technology and innovation at Children’s Hospital & Medical Center in Nebraska says:

“We do a high volume of laparoscopic surgeries using the da Vinci Surgical System. It’s hard for a person to achieve perfect stillness because of breathing and heart rate. The da Vinci system absorbs the surgeon’s natural body movements and makes sure everything happens with great precision.”

However, important decisions are still made by human surgeons.

Some more great uses of AI include 3D organ reconstruction for surgeons to plan operations and computer vision during surgeries to evaluate surgery control.

Practice

Over the past decade, synthetic biology has produced developments like CRISPR gene editing and some personalized cancer therapies.

Yet, the creation process for such sophisticated treatments is still incredibly costly and inefficient.

In the future, AI will enable us to handle far more systematic complexity and, consequently, help us transform the way we understand, discover, and affect biology.

This will be made possible by improved access to various data.

The efficiency of the drug discovery process is expected to improve by helping better predict early which agents are more likely to be effective, and anticipate adverse drug effects.

That is a critical point because these issues have stymied the development of effective drugs at a costly late stage in the development process.

As a result, we can expect more democratic access to advanced therapies at a lower cost.