How TIMELY is revolutionizing the healthcare sector: An interview with Dimitris Gatsios, COO of Capemed
Dimitris Gatsios works at Capemed, a leading company in the Greek medical technology sector. In his position as COO, he plays a key role in the development of innovative solutions that revolutionize healthcare, particularly in the fields of radiology, laboratory information systems and Parkinson's and cardiac rehabilitation. We spoke to him about his work on the TIMELY project for the latest post on the SEMDATEX blog.
About
Dimitris Gatsios is the COO of Capemed, a forward-thinking company in the Greek medical technology sector specializing in radiology and laboratory information systems as well as pioneering solutions for Parkinson's and cardiac rehabilitation. With a strong background in research and a commitment to the advancement of healthcare, Dimitris combines operational acumen with a visionary approach. He is focused on developing innovative products that address critical healthcare needs to improve patient outcomes and optimize medical workflows. The Greek company Capemed is one of the consortium partners in the TIMELY project, alongside SEMDATEX and others.
SEMDATEX: The AI in the TIMELY project continuously analyzes data from thousands of CAD patients. Can you explain in more detail what kind of data goes into the AI models and how this data is analyzed?
The AI in TIMELY continuously processes a rich array of data collected from patients with coronary artery disease (CAD) to support remote cardiac rehabilitation (CR) tailored to individual needs. Key data sources include patients’ EHR, physiological measurements from wearable devices, such as heart rate, blood pressure, ECG readings, and physical activity metrics, all collected via a one-key blood pressure monitor, self-applied Holter-ECG patch, and activity tracker. Patients also input self-reported data through an app, including daily health updates, symptoms, and feedback on their progress and goals.
Once collected, using machine learning algorithms, TIMELY applies predictive analytics, which can analyze this data to detect trends and assess risks, like the probability of secondary cardiac events or potential mortality and the predicted effectiveness of the prescribed cardiac rehabilitation programs. TIMELY’s AI models forecast risks like atrial fibrillation by analyzing ECG rhythms and pulse wave patterns. This early detection enables targeted interventions, supporting the case managers and clinicians to adjust treatment plans proactively. Behavioral data, such as adherence to exercise and health behavior goals, is also integrated into the models, facilitating personalized exercise prescriptions and motivational feedback.
How do TIMELY's AI algorithms differ from conventional heart monitoring methods? Are there any particular technical challenges that you had to overcome when developing the AI models?
TIMELY’s AI algorithms offer a more proactive and personalized approach than conventional heart monitoring methods by continuously analyzing a wide range of patient data in real time. Unlike traditional monitoring, which often relies on periodic in-clinic assessments or simple threshold-based alerts, TIMELY’s AI models provide advanced predictive insights by processing complex, multi-dimensional data from wearable devices, self-reported information, and electronic health records. These algorithms can detect nuanced patterns in heart rate variability, ECG data, and pulse wave analysis (PWA) that indicate potential cardiac risks, such as atrial fibrillation, even before symptoms manifest. Additionally, TIMELY’s AI supports individualized treatment by suggesting to case managers and clinicians how to adjust exercise and recovery plans in response to a patient’s real-time condition, helping to keep rehabilitation goals aligned with current health.
Developing these AI models came with several technical challenges. First was building models that were both accurate and explainable. Clinicians need to understand AI-driven recommendations to trust them in making critical decisions. TIMELY tackled this by developing “explainable” machine learning models, which use transparent, interpretable methodologies that clarify the reasoning behind predictions.
Another challenge was handling continuous data streams which required developing robust, real-time processing capabilities. This necessitated both technical optimization for efficiency and ethical considerations, such as ensuring data security and patient privacy, which TIMELY addressed through encrypted data exchanges and strict compliance with privacy regulations. These technical solutions collectively allow TIMELY to provide comprehensive, patient-centered care while advancing beyond the limitations of conventional heart monitoring.
Finally, ensuring the algorithms can accurately process data from various sources—often with differences in format and quality—requires substantial effort. This challenge is addressed by adopting Fast Healthcare Interoperability Resources (FHIR) standards for data compatibility, allowing seamless integration of data from wearable devices, patient apps, and EHRs.
About TIMELY
The TIMELY project is an EU-funded initiative focusing on innovative monitoring of patients with coronary heart disease. At its heart is an advanced, AI-powered eHealth platform developed in collaboration with leading European institutions. The platform uses artificial intelligence to detect risks at an early stage, provide personalized treatment recommendations and thus prevent complications. TIMELY integrates data from smart devices and patient records to provide precise, predictive healthcare solutions. With the aim of revolutionizing the care of CAD patients across Europe, the project is setting new standards in the field of preventive medicine.
The platform suggests personalized therapy adjustments based on the AI. What does that mean?
TIMELY’s AI and the resulting decision support system (DSS) is integral to cardiac rehabilitation programs, enabling clinicians to make evidence-based, individualized recommendations. These AI-driven systems can provide insights on factors such as the patient’s likelihood of adhering to the CR program, optimizing intervention strategies for enhanced engagement and outcomes. Behavioral Change Techniques (BCTs), including contextual nudges and feedback on goal achievement, are applied to support patient motivation and self-management. Altogether, the AI processes enable TIMELY to improve patient outcomes, reduce on-site visit needs, and provide a sustainable, patient-centered approach to long-term CR.
How does TIMELY ensure that the recommendations provided by AI can be seamlessly integrated into clinical practice? What role does interoperability with existing hospital information systems play?
TIMELY ensures that AI-driven recommendations are seamlessly integrated into clinical practice by focusing on interoperability, user-friendly interfaces, and evidence-based decision support. The platform’s design prioritizes alignment with clinical workflows, allowing healthcare providers to efficiently incorporate TIMELY’s recommendations into patient care. Key to this integration is TIMELY’s compliance with Fast Healthcare Interoperability Resources (FHIR) standards, which enable seamless data exchange with existing hospital information systems (HIS) and electronic health records (EHRs). This standardization ensures that TIMELY’s data inputs and outputs are compatible with various systems across hospitals, avoiding duplicate records and facilitating a holistic view of each patient’s health.
Interoperability is essential for clinicians to access a complete, up-to-date dataset, which includes both TIMELY’s remote monitoring insights and data from in-hospital tests or past clinical encounters. This integration reduces the risk of information gaps and allows TIMELY’s AI recommendations to be directly actionable within the broader context of a patient’s medical history. For example, TIMELY can retrieve baseline data from a hospital’s EHR to tailor AI-driven recommendations, such as personalized exercise prescriptions or monitoring alerts, that align with each patient’s current health condition.
Additionally, TIMELY provides an intuitive dashboard and explainable AI models that present recommendations in clear, clinically relevant terms. This transparency allows clinicians to quickly assess the rationale behind each recommendation, enhancing trust and enabling timely, informed decisions. By integrating with HIS, TIMELY’s decision support systems (DSS) function as a seamless extension of existing clinical tools, making it easier for healthcare providers to adopt AI-enhanced workflows that improve patient outcomes while maintaining efficiency and continuity in care. This interoperability thus plays a critical role in ensuring that TIMELY’s innovative capabilities are both practical and beneficial in real-world clinical settings.
A key feature of TIMELY is personalized healthcare, a very popular topic at the moment. How does the AI adapt recommendations to the individual needs of patients and how is it ensured that these recommendations are evidence-based and practical?
The AI models analyze the patients’ real time data to detect individual patterns and trends, allowing TIMELY to adjust recommendations dynamically based on a patient’s current health and response to therapy.
For example, TIMELY’s algorithms can support case managers modify exercise plans in response to changes in heart rate variability or recent symptoms reported by the patient. If the data suggests elevated cardiovascular risk, such as arrhythmia detected through ECG or high blood pressure spikes, the AI can recommend more frequent monitoring, modified physical activity, or even trigger alerts for clinical intervention. Similarly, TIMELY’s predictive models use data on cardiopulmonary metrics to adjust rehabilitation goals, tailoring the intensity and frequency of exercises to support safe, effective cardiac recovery.
To ensure recommendations are evidence-based, TIMELY’s AI integrates clinical guidelines and best practices from cardiology and cardiac rehabilitation, aligning with standards for managing ischemic heart disease (IHD). These guidelines inform the foundational logic of the decision support systems (DSS), ensuring that all automated suggestions adhere to established medical protocols. Additionally, explainable AI models make it possible for clinicians to see the data and reasoning behind each recommendation, which builds trust and allows healthcare providers to validate or adjust AI-driven guidance in a clinically appropriate way.
Practicality is achieved through a user-friendly interface that translates AI insights into clear, actionable steps for both patients and clinicians. By combining real-world data with clinical guidelines, TIMELY ensures that each recommendation is not only tailored but also medically sound, relevant, and easy to implement in everyday patient care.
To what extent can the AI in TIMELY enable predictive measures for the secondary prevention of heart disease?
One key component of TIMELY’s predictive capability is its machine learning algorithms, which are trained on large datasets to recognize patterns associated with cardiovascular risks. For instance, TIMELY’s AI can identify an increased risk of atrial fibrillation (AF) by analyzing pulse wave and ECG data, allowing for preemptive steps that mitigate further complications associated with AF, such as stroke. This predictive analysis also extends to assessing a patient’s overall risk of mortality and the likelihood of secondary events, such as recurrent myocardial infarctions. Clinicians can use these insights to adjust treatment plans or recommend lifestyle changes proactively, preventing the progression of the disease.
TIMELY’s predictive models are further supported by explainable AI frameworks that provide transparent risk assessments and reasoning for clinicians, making it easier to incorporate these insights into clinical decisions. By continuously learning from new data, TIMELY’s AI can also adapt predictions over time, fine-tuning its risk assessments based on each patient’s evolving health profile. This dynamic, predictive capability enables TIMELY to play a proactive role in secondary prevention, offering timely, personalized interventions that support long-term heart health and reduce the likelihood of critical cardiovascular events.
Does TIMELY have the potential to be used in other areas of monitoring, such as the prevention of disease? How flexible is the AI platform and how might it be further developed to cover new indications?
TIMELY’s AI platform indeed has significant potential to be adapted for primary prevention and other areas of disease monitoring, thanks to its flexible, modular design. Currently tailored for secondary prevention in ischemic heart disease (IHD), the platform could be expanded to address primary prevention by focusing on detecting early risk factors and lifestyle indicators associated with developing cardiovascular conditions. For primary prevention, TIMELY’s AI could monitor baseline metrics like blood pressure, physical activity, heart rate variability, and even behavioral patterns, identifying trends that may indicate an increased risk of disease onset. By flagging these early risk factors, the platform could support preventive interventions aimed at mitigating disease before it manifests.
The flexibility of TIMELY’s AI lies in its foundational architecture, built on Fast Healthcare Interoperability Resources (FHIR) standards and designed with modularity in mind. This adaptability allows TIMELY to incorporate additional health metrics and integrate new data sources relevant to other diseases, such as glucose monitoring for diabetes or respiratory rate tracking for chronic obstructive pulmonary disease (COPD). By adjusting its machine learning algorithms and decision support systems (DSS) to accommodate these new data types, TIMELY’s platform could evolve into a more comprehensive preventive care tool that addresses multiple chronic conditions.
Further development to cover new indications would likely involve training the AI models on expanded datasets that reflect the specifics of other diseases, enabling predictive capabilities tailored to diverse conditions. This might include adding specialized wearable devices, such as continuous glucose monitors, or implementing algorithms that detect risk factors for metabolic syndrome or hypertension.
TIMELY’s modular design and evidence-based approach position it as a promising tool for expanding into primary prevention and broader health monitoring, making it highly adaptable for supporting preventive healthcare across a range of chronic disease areas.
Please explain how Machine Learning is integrated and how it works in the platform?
In TIMELY, machine learning (ML) models are integrated as microservices, creating a flexible, scalable architecture that enhances both performance and modularity. By deploying each ML model as an independent microservice, TIMELY allows for seamless integration and continuous operation of multiple predictive algorithms without interrupting the platform's other functions.
Each ML microservice operates independently, handling specific tasks such as analyzing ECG data for arrhythmias, assessing heart rate variability, or predicting patient adherence to the rehabilitation program. These microservices are loosely coupled and communicate with the main TIMELY platform through APIs (Application Programming Interfaces), allowing data to flow securely and efficiently between the AI models and the rest of the system. For instance, when new data from a wearable device is uploaded, TIMELY’s platform automatically routes it to the relevant microservices for analysis, which then return processed insights or risk scores back to the clinician’s dashboard or patient app.
The use of microservices for ML models also enhances TIMELY’s scalability and adaptability. As each model is a separate microservice, new ML models or upgraded algorithms can be added without affecting the existing system, facilitating the continuous development and testing of improved predictive tools. This modular structure allows TIMELY to respond quickly to advancements in AI or to incorporate models targeting new health metrics, such as blood glucose monitoring, for expanded disease management.
Moreover, the microservice architecture supports efficient load balancing and resource management, crucial for handling the continuous data streams from patient devices. It ensures that each ML model runs optimally, scaling up or down based on demand. This approach also simplifies deployment and maintenance, enabling TIMELY to deliver reliable, real-time AI insights across its eHealth platform, with minimal downtime or disruption to end-users.