Biotech and Healthcare

The Importance of Bespoke AI Solutions in the Healthcare Sector





Biotech and Healthcare


The Importance of Bespoke AI Solutions in the Healthcare Sector

Published: 2024/05/16

7 min read

In 2021, the global AI in healthcare market was valued at just over $11 billion USD – by 2030 this will reach $188 billion USD, according to Statista. This transformative technology is creating significant opportunities, but how can the healthcare industry best integrate AI to deliver for patients and medical staff? This article will present ways to harness the power of AI in healthcare, from clinical trials and diagnostics to health services provided at medical centers and improving patient care at home.

It’s important to note that each of the application types mentioned in this blog has been explored in separate scientific publications and articles. The aim of this article is to showcase the vast possibilities and inspire you to push the boundaries of your own work in healthcare. With the right healthcare software development partner, you can achieve what was once considered impossible two decades ago. Let’s look at a few examples of how AI is transforming healthcare.

Enhancing clinical trials with AI

Clinical trials are the most expensive and fragile part of the path that new medicine takes towards the market. As a result, any new technology that can assist in optimizing and de-risking it as much as possible is being employed. AI algorithms are a prime example. Not so long ago, Matthew Hutson published a piece in Nature Index entitled How AI is being used to accelerate clinical trials. Before that, a Chinese medical academic group described examples in their scientific work “Harnessing artificial intelligence to improve clinical trial design”.

Perhaps the most exciting application in the clinical development of a drug is to predict what outcomes are expected from clinical trials. For example, researchers at the University of Illinois Urbana-Champaign developed an algorithm called HINT (Hierarchical Interaction Network), now offered by IQVIA. Jimeng Sun and his team curated and released benchmark datasets for clinical trial outcome predictions. Then, they built a hierarchical graph to capture the interactions between clinical trial components and simulate the trial process for the outcome prediction. In their publication, the group shows results of retrospective analysis and successful prediction of phase three clinical trials of drugs in a range of indications, including diabetes, asthma, heart failure, depression and liver cancer.

Another such algorithm is called SPOT (Sequential Predictive Modelling of Clinical Trial Outcome). In a publication describing it, available on a preprint server, the authors claim SPOT can surpass HINT. A further development in that direction might be observed, but the most important thing is that both HINT and SPOT demonstrate that such applications are possible.

Although analyzing and predicting outcomes of clinical trials both contribute to better clinical trial design by, for example, enabling the reduction of sample size, designing is not a task that has been directly handed over to AI.

How can AI help in patient recruitment?

Going further, the part of clinical trials to be the most explored with AI methods is actually patient recruitment, as summarized by Scott Askin from Novartis in his article “Artificial Intelligence Applied to clinical trials: opportunities and challenges”. The main opportunities for using AI in patient recruitment are matching potential participants with the most suitable clinical trials, automation of the trial recommendation process and acceleration of site initiation. All of the above are very important factors in how quickly a trial can be completed. Difficulty in recruiting patients can be a real blocker, and the limitation can come from two significant issues.

Firstly, some indications are rare enough that there are very few patients available. On the other hand, abundant indications with high unmet clinical needs and commercial potential are crowded with competing clinical trials, which limit the number of patients left available to recruit. To address this, claims they can increase the speed of clinical trial enrollment (CTE) three times. Their system uses AI to identify patients eligible for clinical trials at the time of the evaluation. Thanks to AI-assisted disease detection and real-time access to the hospital imaging system, it quickly notifies the clinical and research team when a new patient is identified. The software also allows for pre-screening of candidates eligible for the trial in the health system it accesses. A similar solution is offered by Opyl. The number of such solutions will grow because, to quote Urtė Fultinavičiūtė from Clinical Trials Arena, they are “benefiting all stakeholders”. Urtė rightfully points out that such a solution is highly sought after in oncology. The main reason for this situation is that effective treatment is still an unmet clinical need for many types of cancer. Hence, patients actively seek better chances with newly developed drugs.

Algorithms and identifying diseases

Although diagnostics. especially biomarker assessment, is crucial for patient recruitment in clinical trials the field itself is much broader, and so are artificial intelligence applications here. Previously, this article touched on biomarker identification in massive patient datasets. Once identified, measurable biological features of patient, such as specific gene expression level, presence of a mutation, etc., are combined with an advanced algorithm to create a diagnostic assay.

As of January 2024, according to UTHealth San Antonio researchers, there have been 691 FDA-approved artificial intelligence and machine learning-enabled medical devices. Certainly, not everything can be called AI and some of the algorithms, being classical machine learning, have been in development for a while. The Medical Futurist lists approved AI/ML-based algorithms in a useful table referencing the approval number, algorithm type, and device or algorithm name and its purpose. Some examples of approved applications worth exploring are the detection of diabetic retinopathy (Idx – Digital Diagnostics – LumineticsCore) or diagnosis of liver and lung cancer on CT and MRI (Arterys MICA now offered in Pixel from Tempus). Those solutions are designed to change point-of-care practice and support doctors in the diagnostic process.

Patient care

For a few decades now, the collection of multi-omics data underlying cancer and other diseases has provided a huge resource for analytical algorithm development. Companion diagnostic reports built on genomics tests are already available to support clinicians in assessing precision therapy success chances. Such solutions previously did not need AI. However, this is changing, as AI-solutions help them reach higher quality and precision levels. Over the last few years, researchers in academia and companies have applied and used the idea of the Digital Twin/Molecular Twin.

In January of 2024, Arsen Osipov and his colleague published their work in Nature Cancer entitled “The Molecular Twin artificial-intelligence platform integrates multi-omic data to predict outcomes for pancreatic adenocarcinoma patients”. The concept of digital twin has already been used successfully in other industries, while in the pharmaceutical industry, it is being explored to save costs and improve collaboration when implementing biological processes and building cGMP factories. Toni Manzano and William Whitford examine in depth how AI improves digital twins in biopharma manufacturing on the pages of BioProcess International.

Furthermore, predictive models are also used in the patient care sector to solve pressing yet mundane tasks, such as optimizing hospital bed occupancy and patient queues. One company, Terawe, showed how they repurpose a solution created for other industries to suit hospital bed management, while The National University Health System in Singapore’s Endeavour AI platform can predict the occupancy of hospital beds up to 2 weeks ahead of time.

Remote patient monitoring with AI

LAI is slowly making its way towards our homes in relation to healthcare. Wearable devices are readily available and enable near real-time data collection. With pattern recognition, anomaly detection and predictive analytics, AI solutions are becoming helpful tools to track patients and enable timely interventions –. leading to lower need for hospitalization and better outcomes. It is the future shown in many sci-fi works. Florida-based Healthsnap is one of the pioneers of virtual care management solutions and offers remote patient monitoring and chronic care management. The US company Tenovi is another such provider whose AI-supported services help patients with diabetes, heart failure, hypertension and chronic obstructive pulmonary disease (COPD).

Data security and privacy is the key

A critical issue that needs to be mentioned with any AI topic in healthcare relates to the sensitivity of patients’ data. Almost all the applications discussed involve handling personal, very sensitive data. Therefore, to implement algorithms using necessary data, companies and researchers need to follow compliance regulations (such as Health Insurance Portability and Accountability Act – HIPAA), anonymize and aggregate data whenever possible and receive a patient’s consent to their information prior to using it. It is also crucial to secure such systems from unauthorized access and leaks. With such high stakes, investing in the right partners to provide state-of-the-art security and deliver a solution’s quality is the foundation of success and stable operations.


Data security in genAI webinar

Empower your healtcare projects with AI

The application of AI in healthcare holds tremendous potential for improving outcomes for patients, healthcare providers and drug developers seeking approval of their innovative medicines in clinical trials. With continued advancements in AI algorithms and increased collaboration between biotech experts and AI researchers, the healthcare industry can expect further astounding developments that positively impact human health and well-being.

Clinical trials, diagnostics and patient care processes can already be aided by tools that can increase effectiveness, reduce costs and help an organization achieve more. There are many ways to use AI, but they often require additional work to fit a process or organization’s needs. Collaborating with AI specialists to develop a custom solution is a wise choice if you lack expertise. Contact our team using this form to create a bespoke AI platform to empower your role in the healthcare ecosystem.


How AI is being used to accelerate clinical trials (Nature Index): 

Harnessing artificial intelligence to improve clinical trial design (Nature: Communications Medicine): 

HINT: Hierarchical Interaction Network for Clinical Trial Outcome Prediction Insight Brief: 

HINT: Hierarchical interaction network for clinical- trial-outcome predictions (Cell Press Patterns): 

SPOT: Sequential Predictive Modeling of Clinical Trial Outcome with Meta-Learning: 

Artificial Intelligence Applied to clinical trials: opportunities and challenges (Health and Technology, Springer): 

Clinical Trials Arena “It’s a match! Connecting patients to clinical trials with AI” –  

FDA-Approved Artificial Intelligence and Machine Learning (AI/ML)-Enabled Medical Devices: An Updated Landscape: 

The Medical Futurist: 

Digital Diagnostics’s LumineticsCore: 

Tempus’ Pixel for lung nodules changes tracking in time: 

“The Molecular Twin artificial-intelligence platform integrates multi-omic data to predict outcomes for pancreatic adenocarcinoma patients” (Nature Cancer): 

Digital Twin for biopharma process: 

AI-Enabled Digital Twins in Biopharmaceutical Manufacturing 

Terawe 4SiteHospitalBed Management: 

NUHS press release: 



About the authorJacek Szmatka

Head of Digital Transformations and Software Consulting

An open-minded leader with over 20 years’ experience in the IT world, Jacek’s career has seen him evolve from a computer science graduate to software engineer to a co-founder and CTO of a tech start-up. Before joining Software Mind, Jacek was part of a team that developed a bioinformatics company and served as an executive board member. In his current role as Vice President of Business Development, Jacek helps leading American and European companies design and build innovative development teams from scratch. A true believer in the transformative power technology can have on our lives, Jacek maintains a keen interest in R & D, in particular with solutions that involve AI, IoT, life science and cloud technologies.

Subscribe to our newsletter

Sign up for our newsletter

Most popular posts