AI in Healthcare: Bridging the Gap Between Proof and Practice

By Dan Wilhelm and Roozbeh Sadeghian, Ph.D., Associate Professor and Program Lead of Analytics, Harrisburg University of Science and Technology

With machine learning (ML) and artificial intelligence (AI) in the spotlight, everything about how humans work, play, conduct business, and solve problems is changing.

Of course, we’ve been here before. For every sea-changing invention, there’s a new balance to find. We’ve made our peace with the abacus, but the internet still vexes us.

Where does this leave us with AI? And how can we invite this powerful force into sensitive industries where human contact, intuition, and privacy are critical?

I wanted to better understand the impact of ML and AI on healthcare, so I sat down with a scholar-practitioner to find out. For a start, he told me to stop thinking about this transition in terms of “years.”

Where Does Healthcare Stand When It Comes to AI?

Dr. Roozbeh Sadeghian is the Program Lead and Associate Professor of Analytics at Harrisburg University of Science and Technology in Pennsylvania. He knows about AI’s positive implications in healthcare because he’s actively advancing that frontier.

“There are barriers to rolling out AI and ML more widely,” said Dr. Sadeghian, “including pushback from the FDA about using AI in healthcare. I think it’s less about the soundness of the method and more about the technological novelty.”

Then, he said something that made me realize what kind of problem “AI in healthcare” really is.

It’s a communication problem.

“Researchers need to be especially vigilant when it comes to collecting case studies and proofs of concept, and then communicating those concepts to healthcare administrators, regulators, and stakeholders. It’s especially critical for computer and data scientists to be able to speak plainly about what their systems can do in the healthcare realm. The proof is there, but we need to make the case to decision-makers at hospitals, clinics, practices, and healthcare systems.”

When he’s not in the classroom, “making the case” is exactly how Dr. Sadeghian spends his time. He’s published multiple papers on using AI for patient care, including:

– Towards an Automatic Speech-Based Diagnostic Test for Alzheimer’s Disease
– A Deep-Learning Approach to Diagnose Skin Cancer Using Image Processing

In both cases, the goal was to use something familiar – smartphones and cameras – to create new, streamlined, noninvasive screening experiences for patients that support long-term health and can be deployed in a variety of settings.

This research is relevant because, now more than ever, we must learn to do more with less.

The US Department of Health and Human Services acknowledges massive underinvestment in America’s health system that has left 30 million individuals in underserved areas without reliable access to healthcare. Moreover, the US Government Accountability Office estimates that more than 100 rural hospitals closed between 2013 and 2020, resulting in patients traveling 20 more miles for inpatient care and 40 more miles for specialist services.

Tools are needed to address as much of the patient experience as possible before they enter the exam room.

“Using ML and AI could help keep sick or elderly people out of waiting rooms. We’re not going to replace doctors or nurses – we need them more than ever. Instead, we’re giving them advantages in efficiency, accuracy, and repeatability.”
What Does AI in Healthcare Actually Look Like?

Here are three other examples of AI and ML positively impacting the patient experience:

– Google DeepMind: These clinically validated algorithms analyze test results and vital signs to provide early detection for patients at risk of Acute Kidney Injury.
– IBM Watson: This system uses ML to study clinical trials, patient records, and medical literature to provide personalized cancer treatment options.
– PathAI: PathAI digitizes pathology slides into high-fidelity digital images and conducts advanced analysis. This assists with diagnostic processes and flags microscopic indicators of diseases.

Other avenues for AI to improve processes and outcomes in healthcare include automated transcription for nurses and EMRs, chatbots for patient engagement, remote monitoring and analysis, and better-designed clinical trials.

But we’re still up against the communication gap. Dr. Sadeghian says it’s often “hard to convince” stakeholders of the real-world value of these tools.

“We need to do everything in our power to lay out the proof. We can publish to share our findings with regulators, and we can come up with more accurate models to make the case. For speech recognition, regulators wanted to know our model would hold up. We needed lots of samples to show that it works, just as every other researcher needs to have.”

Access, Equity, and Inclusion in Science and Healthcare

I wanted to know what excites Dr. Sadeghian the most about AI and ML in healthcare right now. What’s happening that’s most consequential?

“The general awareness that I see coming together is probably the most important part. People used to see AI as something threatening … but we’re working in real-time to convince students, researchers, instructors, and companies. We need more people to communicate about these subjects and bridge the gap between cutting-edge AI research and the parties helping commercialize the technology.”

Will it happen soon?

piron guillaume U4FyCp3 KzY unsplash“I’m optimistic we’re getting to that point. As always, the Scientific Method is our best ally – but the proof isn’t enough by itself. We need people who speak that common language crossing the bridge between science and business.”

Access, equity, and inclusion must certainly be a part of that common language.

“We see huge social impacts from AI in healthcare – in the data we’ve collected regionally in Pennsylvania, for example,” Dr. Sadeghian added. “Many rural areas have insufficient access to medical procedures. AI will impact society through both safety and convenience. Everybody has smartphones now; why not have the doctor in your hand? A cultural shift is underway.”

AI can give a preliminary screening and keep people out of cities and congested areas, bringing access to more rural areas and saving office visits for people who need them. This also impacts transportation, walkability, and other aspects of civic planning – even pollution mitigation.

Inviting the black box of AI into healthcare isn’t some hazy dream. It’s happening today. Younger generations are the most scientifically engaged ever, though, which means consensus-building on tech policy could move faster going forward.

Politicians have noticed the social, cultural, and economic value of investing in science, technology, engineering, and mathematics education. Here in Pennsylvania, the governor is making it easier for universities to admit low-income students at no cost. Prioritizing access and equity in schools can reduce the bias creeping into algorithms, too; we need an abundance of perspectives to build truly smart and unbiased machine learning models.

Tomorrow, your iPhone could save your life. But first, the technologies inside that iPhone get dreamed up at a university. Then, people like Dr. Sadeghian figure out how to make them smart enough to assist nurses and doctors. Keep your eyes on our STEM universities, because we’re living through a technological revolution like few in history. There’s a brighter future ahead for healthcare being won in university halls and peer-reviewed journals.


Roozbeh Sadeghian, Ph.D., is the Associate Professor and Program Lead of Analytics at Harrisburg University of Science and Technology in Pennsylvania. He earned his BS and MS in Electrical Engineering at Isfahan University of Technology and Shiraz University, respectively, and earned his doctorate in Electrical Engineering at the State University of New York at Binghamton.

Dr. Sadeghian’s teaching interests include graduate courses such as machine learning, pattern recognition, analytics, big data, and signal processing. His research is founded at the nexus of science and engineering in clinical diagnosis, early notice, and assessments of healthcare disorders.


Dan Wilhelm earned his BA in Creative Writing at Susquehanna University. Since 2007, he’s worked as a copywriter, SEO and PR strategist, technical writer, lead editor, and process manager for teams of writers. He is currently the Content Manager for Harrisburg University of Science and Technology. Dan has ghostwritten for CEOs and founders, obtained the rank of Eagle Scout, and was once a national semifinalist in the Norman Mailer Writing Awards for narrative nonfiction.