Experts Share: How Can Diagnostic AI Be Integrated Into Existing Medtech Systems?

Medtech companies are already using diagnostic AI to make devices smarter and workflows faster. But making sure these tools work well within existing systems is not just a simple process. It takes planning, the right people, and tech that’s designed to fit the real pace of hospitals, clinics, and labs.

It’s already being used in everything from helping with paperwork for clinical trials to deciding how sales reps talk to different hospitals.

According to McKinsey’s 2024 survey of 40 senior medtech leaders, 2/3 of their companies had already started using generative AI tools. About 20% were going beyond trial runs and using them across teams.

Some of the fastest results came in research and development. Teams there deal with piles of documents, from early study design to packaging labels.

With AI handling the first drafts of these materials, teams were able to finish tasks more quickly and focus more on actual product work. In a few cases, labelling tasks became 30% more efficient.

Sales teams are also using diagnostic AI inside their existing dashboards. These tools help teams spot buying patterns, recommend talking points, and even pull together emails and presentations. Instead of trying to guess what a surgeon or procurement officer might care about, reps get suggestions that match that buyer’s actual interests.

 

Can AI Be Brought Into Daily Operations As Well?

 

A lot of the time, stock sits in hospitals under consignment or is loaned out to clinics, which makes tracking harder. AI tools are helping companies forecast demand and prevent supply problems.

In procurement, AI has been used to speed up invoice matching and review contract terms. Around 20% of companies surveyed are either using or planning to use AI to flag contract issues or spot saving opportunities. These changes have already helped some firms cut costs by up to four percent.

But the companies that saw the best results did not try to do everything at once. They picked one part of their business such as marketing or procurement, and started there. When they were ready, they added AI into another area. Those who tried to launch in too many places too quickly usually struggled to get the results they wanted.

 

What Do UK Rules Say About All This?

 

The Medicines and Healthcare products Regulatory Agency is in charge of keeping medical AI safe in the UK. It treats any software used to support medical decisions as a regulated device, and that includes diagnostic AI.

Their Software Group works across the organisation to check that these tools are safe and meet real clinical needs. That means reviewing files before a device can go on the market and keeping an eye on what happens after it’s in use. They are also updating how they classify and regulate these tools to reflect the way AI actually works, including systems that keep learning over time.

In 2022, MHRA published a change roadmap. It lays out how they want to update rules for AI tools used in health. They’re working with groups like the FDA in the United States and Health Canada to make sure their rules don’t clash with other countries. The idea is to set clear expectations for transparency, training, and risk handling… especially for systems that keep updating how they work.

 

What’s Getting In The Way?

 

Even though the interest in AI is high, not everyone is getting it right. One of the main problems is a lack of focus. McKinsey found that 31% of AI users in medtech weren’t sure how to decide which use cases to start with. In many cases, teams worked in silos, making it hard to scale what worked.

On top of that, more than half of respondents pointed to tech problems, like privacy concerns or scattered data systems, as a big hurdle. Many organisations also said they didn’t have enough people with the right technical skills. Even where companies had the right tools, they often failed to get staff to use them, usually because the tools were too generic or felt like extra work.

 

How Do Experts Think AI Can Be Integrated To Existing MedTech?

 

Our Experts:

 

• Dr Antonio Espingardiro, IEEE Member, Software And Robotics Expert
• Dr. Loina Prifti, Leadership Strategist & Executive Digital Transformation Consultant, Nerisia
• Marc Fernandez, Chief Strategy Officer, Neurologyca

 

Dr Antonio Espingardiro, IEEE Member, Software And Robotics Expert

 

 

“There has always been demand for scalability and quality within healthcare. With the global population living longer than ever before, and many citizens living with health conditions that require regular medical checks, leaders are looking to technology to ensure equity in care delivery and fully democratise healthcare. Artificial intelligence (AI) can analyse vast quantities of information, and when coupled with machine learning, it can search through records and infer patterns or anomalies in data, that would otherwise take decades for us to analyse as human beings.

“AI is already in use, through the application of machine learning (ML) to enable medical professionals to better understand certain health conditions. For example, through AI programmes, scientists can explore the huge volume of data that makes up the DNA in the blood of cancer patients. With this insight, they can detect specific mutations that will eventually be used to classify those most at risk, or identify those that may perhaps require further consultation. In the future, AI could potentially screen patients on a much greater scale, and even assist with overall disease management. In time, medical professionals will eventually be able to improve or better coordinate healthcare plans for long term treatment and infer timely and reliable data throughout diagnosis.

“We are just starting to see the beginning of a new era, where machine learning could bring substantial value and transform the traditional role of the doctor. With the increased adoption of AI and robotics, we will soon be able to deliver the scalability that the healthcare sector needs and establish more proactive care delivery. We may even be able to solve of some of the biggest challenges and issues of our time.”

 

 

Marc Fernandez, Chief Strategy Officer, Neurologyca

 

 

Can diagnostic AI connect to existing hospital systems?
“Yes, diagnostic AI can readily connect to existing hospital systems. Most hospitals utilize electronic health record systems (EHRs), which are designed for compatibility with other technologies. These systems house patient histories, lab results, imaging scans, prescriptions, and clinician notes. Diagnostic AI can integrate seamlessly, pulling in relevant patient data, analyzing it in context, and returning findings directly into the clinician’s workflow.

“For example, AI might review a patient’s recent lab results and scan reports alongside historical data to flag early signs of disease progression. Because these data are already digital and structured, AI tools can process them quickly and offer suggestions in real time.

“Imaging systems and lab equipment can also be linked. When an MRI or blood test result is uploaded, AI can immediately scan for abnormalities. What makes this useful is not just speed, but AI’s exceptional ability in pattern recognition, often detecting subtle signals easily missed by the human eye.

What are the integration challenges?
“While the foundation is in place, it does take work to implement these tools effectively.

For hospital tech teams, the main challenges include:
• Making sure different systems communicate correctly. Hospitals often use a mix of older and newer technologies, so data consistency is a key issue.
• Ensuring sufficient computing power and cloud access to support AI workloads. Many hospitals are upgrading their infrastructure to support advanced tools like this.
• Meeting strict privacy and data protection standards. Patient data must remain secure and compliant with laws like HIPAA.
• Supporting adoption by clinicians. Even the best AI system is only useful if doctors and nurses understand how to use it and trust its recommendations.

For AI vendors, including companies like Neurologyca and Microsoft, the priorities should be:
• Making AI systems flexible enough to work across a variety of hospital environments.
• Ensuring transparency in how AI arrives at its conclusions. (Clinicians need to see not just the result, but the reasoning behind it.)
• Validating performance through clinical trials and regulatory approval. Diagnostic AI must meet high safety and reliability standards.
• Addressing potential bias in training data to avoid uneven performance across different patient populations.

“The importance of Human Context
At Neurologyca, we specialise in what we call Human Context AI. Instead of just analysing lab results or images, our system can look at the behavioural and emotional signals that often go unspoken.

These include changes in speech, facial expressions, and interaction patterns. Use cases include:
• Detecting early signs of stroke by analyzing micro-facial expressions during tele-health appointments.
• Helping clinicians better understand neurodivergent patients, whose emotional cues may differ from neurotypical norms.
• Evaluating how a patient is truly responding to a mental health treatment by analysing emotional signals, rather than relying solely on verbal feedback.

“These insights are especially valuable for conditions where timing and nuance matter. If AI can pick up on small changes before they become serious issues, it can support more proactive care.

“Ultimately, the goal is to give clinicians better tools that work in the background and help them focus more on the human side of care helping them make smarter decisions, earlier interventions, and a deeper understanding of each individual patient.”

 

Dr Loina Prifti, Leadership Strategist & Executive Digital Transformation Consultant, Nerisia

 

 

““Honey, I might have cancer” – probably, there’s no Millennial that has not been down the rabbit hole of googling their symptoms and got some terrible diagnosis. I remember peeing At some point we have all been there and after the initial shock, most of us have developed a deep distrust in diagnoses through a computer.
However, I must say that I am convinced that the no doctor could ever beat AI with regards to clinical diagnostic and clinical therapy.

“If we go one step back, and try to analyse the cognitive internalised process a doctor goes through while diagnosing and assigning therapy for a patient. They gather all the symptoms and based on their experience, meaning cases they have already treated, or cases they have read about, they derive what the cause might be. So if you are coughing, have chest pain and they can listen that you are not breathing normally, you probably have bronchitis let’s say. And if you have bronchitis, they would assign you either antibiotic A or antibiotic B. Now an AI system can include and analyse all symptoms ever happened to any woman being and define a very specific diagnosis.

“It can also analyse every possible treatment ever applied and suggest the best possible treatment for the patient at hand. A good trained system would make no mistakes, ever. It would be better at clinical diagnostic as any doctor. The main challenge remains the data to train this system, however most hospitals are required by law to keep records of their patients, by providing the data for the mentioned purpose.

“Now as with any technology the biggest challenge remains the human factor. People resist to change, and people distrust anything new and unknown. Even penicillin was distrusted at first although it was a turning point in human medicine worldwide. So although the technology is already able to offer world class results, there is a long way to get to AI diagnostic yet.”