Medical Digital Twins for Personalized Chronic Care

Medical Digital Twins for Personalized Chronic Care | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Medical Digital Twins for Personalized Chronic Care Michael P. Papazoglou, Bernd J. Krämer, Mira Raheem, Amal Elgammal This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5826330/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract This work introduces the concept of Patient Medical Digital Twins (PMDTs) to simulate treatment outcomes, optimize drug dosages, and deliver personalized chronic care. The PMDT model, supported by an interconnected ecosystem, is validated iteratively by medical institutions to ensure its efficacy and applicability. At its core, the PMDT leverages expressive knowledge structures to capture a patient’s psychosomatic, cognitive, biometric, and genetic data, creating a comprehensive personal digital footprint. This enables medical professionals to run simulations predicting health issues over time and to proactively implement personalized preventive interventions. The PMDT ecosystem integrates big data analytics, continuous monitoring, cognitive simulation, and AI technologies. By connecting stakeholders across the care continuum, it provides deeper insights into a patient’s medical history and supports informed, shared decision-making. Validated in a pilot study through an EU-funded healthcare initiative, the PMDT demonstrates its transformative potential at the intersection of Big Data and AI, positioning itself as a critical tool for advancing personalized preventive care. Healthcare Data Integration Interoperability Medical Digital Twin Knowledge Representation and Processing Federated Analytics Medical Intelligence Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 1 Introduction Digital Twin (DT) technology is an emerging field that enables the creation of digital representations of real-world entities, which can be either physical or perceived (e.g., processes), and tracks the state of these entities during their lifetime. DT can perform bi-directional automated data flow between a physical object and its digital representation. Changes experienced by the physical object can be reflected in its digital model and the insights gained from the digital model can support decisions to optimize the physical object. The integration of the Digital Twin paradigm into healthcare has the potential to revolutionize care processes, create patient-specific digital models of human organs and individual cells, improve patient experience, reduce operational costs, and enhance the overall quality of personalized and preventative care [1]. One of the areas that can benefit from the use of DT technology in healthcare is chronic disease management. Here, a DT can be used to address the wellness, prevention, and ongoing management of focused chronic disease conditions. According to the World Health Organization, chronic diseases, such as cardiovascular disease, diabetes, and Alzheimer’s, account for nearly three-quarters of all deaths worldwide [2]. Yet many of these chronic diseases are preventable, as they are linked to risk factors such as poor diet, obesity, and lifestyle choices. The DT paradigm is a powerful tool for personalized chronic disease prevention, enabling early symptom detection and the implementation of preventive measures and targeted interventions tailored to individual patients. This approach has the potential to proactively address identified risk factors, potentially reversing or preventing the onset of chronic diseases. One of the biggest challenges in applying DT technology to the healthcare ecosystem is that caregivers currently lack a holistic view of their patient needs, often 'swivel-chairing' between multiple systems to manage basic interactions such as scheduling treatments and engaging in patient outreach. Patient and clinician access to patient-generated health data to monitor specific chronic conditions is rarely linked to patient longitudinal patient data. This results in fragmented care, due to poor communication and fragmented sources of patient data that are not yet widely integrated. This makes it much more difficult to align and coordinate chronic care across care teams and associated settings. A real challenge in providing optimal care for chronic conditions is the need to transition from an episodic to a continuous monitoring cycle of the patient’s physiological data and the need to develop collaborative relationships with patients. This can be achieved by sustaining vigilant monitoring of health indicators to optimize treatment and an operational shift that gives patients greater control and personalization over how they manage their condition. DTs can help alleviate these problems by modeling an individual’s genetic makeup, physiological characteristics, and lifestyle habits by acting as a critical harbinger of a “health-tech” approach toward prevention-focused and outcome-oriented healthcare. The DT can achieve this by collecting, collating, and analyzing masses of an individual’s health data to help medical professionals to better predict a patient’s (long-term) response, including side effects, recommend a personalized treatment plan, provide therapy guidance, and prevent deterioration. We call such a DT a Patient Medical Digital Twin (PMDT). The data that will provide the information used to represent an individual’s PMDT will be gathered from a variety of health data sources, including medical sources, electronic health records (EHRs), vital signs, a patient’s medical history (diagnosis and prescriptions), medical and clinical data, symptoms, medical tests, medications, physiological and psychosocial data, nutritional data, patient-generated data, etc. These data are used by different entities in the care continuum and are presented in different formats. A PMDT can bring together and codify such rich data. It can also continuously pull in real-time sensor (IoT) data to create accurate snapshots of the physical patient. This information can be integrated with historical data and predictive analytics to alert healthcare professionals to potential problems and suggest solutions. Our contribution can be summarized as follows: the development of a PMDT - a faithful digital (in vitro) model of the patient - as the linchpin in the transition from one-treatment-fits-all to smart personalized healthcare. This digital model acts as a rich knowledge representation framework for decision support, encompassing a complete patient profile. It achieves this by (a) recognizing a patient’s current state, (b) predicting potential developments, and (c) facilitating interaction with caregivers and healthcare professionals to collaboratively develop a patient-tailored treatment plan, thereby advancing the principles of precision medicine. The remainder of this article is structured as follows: a literature review is presented in Section 2. Section 3 introduces the pilot application, which is used throughout this article. The conceptual framework for the PMDT is discussed in Section 4, followed by the PMDT and its processing environment in Section 5. The implementation and validation of both are sketched in Section 6. The paper concludes in Section 7 by highlighting ongoing and future work. 2 Literature Review DT-based medicine is still in its infancy and has been described in the literature as the nexus of four components - data science, software engineering, representation of expert knowledge, AI and machine learning, - which culminates in tools that support effective clinical decision-making. The study in [1] systematically reviewed emerging technologies for data modelling and analytics for sustainable data-centric healthcare that marked DT technology as a recent inclusion in the healthcare industry, which has the power to revolutionize the healthcare paradigm. This study reviewed eight DT related studies [3–10], and the findings revealed that each study has a limited analytical focus with no studies addressing the integration of heterogeneous and diverse data sources. This study concluded that none of the examined data modelling and analytics technologies for sustainable healthcare operates in isolation. It became apparent that for sustained viability, technological solutions must offer seamless and proficient capabilities for data capture, storage, and analytics. These highlight the contributions of the work presented in this article. Various initiatives have emerged from building DTs of organs, such as the heart, to observe how they might respond to various interventions, thus minimizing the risk of early human trials and accelerating the availability of the treatments to patients. Simulated organs could change how medicine works, making it hyper-personal and less invasive. For example, Hewlett Packard Enterprise deployed its supercomputer to create digital models of the brain for research purposes, while Siemens Healthineers has a DT model to simulate the use of cardiac resynchronization therapy. Dassault released the Living Heart, a realistic model that accounts for electricity, mechanics, and blood flow. The software can turn a 2-D scan from an individual human into a personalized full-dimensional model of his or her heart and run hypotheticals. Research in [11] proposes a DT for the human head to detect carotid stenosis severity from a video of a human face with the help of a coupled blood flow and head vibration model. This DT model represents an attempt to link a video of a patient face to the percentage of carotid occlusion. Other existing prominent examples of digital twins in healthcare include “the artificial pancreas” [12], and pediatric cardiac digital twins [13]. The study in [14] propose TWIN-GPT, a digital twin approach form clinical trials by utilizing Large Language Models (LLMs). All these examples focus on just one single aspect of the human body due to its extreme complexity. Commercial medical device manufacturers are also increasingly using DTs to model both devices and patients to better design devices for people with specific conditions [15]. Further companies are using CT scans and MRI images to create three-dimensional computational models of individual patients that will help clinicians decide on and prepare for surgeries or other procedures. Other research initiatives provide the opportunity to explore predicted disease trajectories for a patient and have the potential to inform the patient’s current state using data derived from that individual to avoid providing a generalized prediction. In [16] the authors present a DT model to simulate and understand the progression of valvular disease associated with the mitral valve. This DT model integrates cardiac electrophysiology with hemodynamic modeling giving a broader understanding of the effect of disease progression on various parameters like ejection fraction, cardiac output, blood pressure, etc., to assess the severity of mitral valve disorders. Several studies have also focused on assisting understanding or management of a target condition or class of conditions with DTs. Research in [17] employs machine learning techniques applied to a DT model based on a variational autoencoder for simulating the clinical trajectories of patients who went on to experience an ischemic stroke. Research presented in [18] claims that precision cardiology will be delivered in a synergetic fashion that combines induction, by using statistical models learnt from data, and deduction, through mechanistic modelling and simulation integrating multiscale knowledge and data, which comprise the two foundation pillars of the DT. Authors in [19] presented an AI-generated DTs approach to support the rapid assessment of in silico intervention strategies. The medical Digital Twin in this paper represents the convergence of Big Data, Knowledge Representation technology, software engineering techniques, and AI endeavors. Through a twinning process, it comprehensively addresses an individual's physiological, historical, behavioral, and biological conditions. The medical Digital Twin in this paper represents a transformative leap from one-size-fits-all treatments to smart, personalized healthcare. The PMDT achieves this by: Recognizing the patient’s current health state, enabling accurate assessments. Predicting potential health developments, allowing for proactive interventions. Facilitating collaboration between caregivers and healthcare professionals, ensuring the creation of a tailored treatment plan that aligns with the principles of precision medicine. This innovative approach centers on creating a faithful digital (in vitro) replica of the patient, serving as a comprehensive knowledge framework for decision support. Unlike traditional methods, the PMDT provides a holistic view of the patient's health, identifying potential disease symptoms, tracking transitions, and signaling impending deterioration. The overarching objective is to create a personalized, life-long in silico replica of a patient, capable of predicting the individual's responses to medical conditions. By integrating these capabilities, the PMDT not only enhances decision-making but also empowers a proactive, patient-centered approach to healthcare, delivering improved outcomes and advancing the vision of precision medicine. This model is designed to facilitate personalized therapy simulation, selection, and outcome prediction. 3 Pilot Study The pilot study presented in this section is an international, observational multicenter cohort study that was conducted in France, The Netherlands, Portugal and Spain and included adult patients treated with CAR T-cell or immune checkpoint inhibitors therapy 1 [20]. The pilot was conducted in the context of the EU H2020 QUALITOP project 2 - which partially funds this research. The main objective of the QUALITOP project is to develop and implement an IT-based European immunotherapy platform that utilizes big data analytics and artificial intelligence to collect and aggregate efficiently and effectively real-world Quality of Life (QoL) data, monitor patients’ health status and manage patients with personalized and preventive focus, to improve their QoL. Immunotherapy is considered one of the complicated treatments protocols as it causes toxicities or side effects, known as Immune-Related Adverse Events (IRAEs), that are challenging to predict because they are not caused by mechanisms involved in other treatment types, such as chemotherapy and radiation [20]. In the cohort pilot study, adult patients are monitored in their real-life, which overcomes the known limitations of Randomized Clinical Trails (RCTs) [20], where patients’ profiles and behaviours may be far different from those seen in RCTs. Relevant historical real-world databases and medical administrative registries were also identified. The study included patients recruited specifically for QUALITOP. Patients’ clinical health status and QoL are monitored up until 18 months post treatment initiation. Psycho-social wellbeing is also monitored with a tailored QUALITOP questionnaire at baseline and 3, 6, 12 and 18 months post treatment initiation. To identify advanced analytical needs of medical institutions across these four EU countries, we have followed a highly iterative/agile software development methodology that started with a structured iterative Requirements Engineering (RE) approach to identify analytical query patterns of recurring analytical requirements, and then incrementally modeled and formalized identified analytical patterns. As will be illustrated in the next discussion, a virtual data lake [6] and the PMDTs models enabled the implementation of these analytical query patterns by utilizing a federated data analytics approach, which is an innovative decentralized machine learning paradigm [21] that facilitates collaborative model training across multiple parties without the necessity of centralizing sensitive data in compliance with regulatory bodies such as GDPR 3 and HIPAA 4 . To improve understanding, we assume that: “ a medical professional in Amsterdam is interested in predicting the QoL of his female melanoma patient, who is 40-years-old, classified with TNM stage "T3AN2Cm0" and undergoing treatment with "Pembrolizumab: 200 mg" at a frequency of "Q3W ”. This example is an instantiation of the textual analytical query pattern “Predict QoL of patient with ”. By using the graphical web interface shown in Fig. 1 , the medical professional may specify the cancer-type and the specific-characteristics of her cancer patient and submit the analytical query request. In the following Sections, this example is used as a running scenario showing the workflow of its execution based on the novel digital twinning approach presented in this article. [1] The QUALITOP cohort study has been registered as an observational study at www.clinicaltrials.gov. Trial registration number NCT05626764 [2] Monitoring multidimensional aspects of QUAlity of Life after cancer ImmunoTherapy - an Open smart digital Platform for personalized prevention and patient management: https://cordis.europa.eu/project/id/875171/reporting [3] General Data Protection Regulation: https://gdpr.eu [4] Health Insurance Portability and Accountability Act: https://www.ncbi.nlm.nih.gov/books/NBK500019/ 4 Digital Twin Conceptualization Scheme A PMDT environment would comprise dispersed patient data sources, sensors, analytics, and visualization. Patient information is often distributed among various healthcare providers who have treated the individual for diverse symptoms and conditions throughout their lifetime. The imperative lies in the seamless integration of this information to offer a comprehensive and unified view of the patient's health history. Otherwise, misdiagnoses, inappropriate medications, duplicate tests, medical-legal issues, and other problems are inevitable. Patient-Generated Health Data (PGHD) involves sensor data from wearable sensors and data generated from handheld devices. These may be used for a variety of purposes including recommendations, alerts, self-assessment, daily activity tracking, time of medications, current nutritional status, and more. The use of PGHD can empower patients and caregivers to manage their health and to collaborate with clinicians via shared decision-making that considers patients’ concerns and preferences. AI-based analytics tools will be used to process patient data to derive insights about current medical conditions as well as make predictions about emerging conditions. This leads to the possibility of targeted chronic disease interventions at an early stage (the notion that prevention is better than cure). It will also be possible to factor in wider aspects of health, such as diet and exercise in chronic disease treatment. Finally, visualization typically collates diverse medical data coming from various sources into a single unified view, allowing users – clinicians, non-specialist staff, and patients – to piece everything together and get a complete overview of the patient’s situation, which unveils insight into patterns and correlations to help medical staff interpret data analytics results faster, recognize trends, and make better decisions as well as aid patients in interpreting and contextualizing their health information. Figure 2 presents a streamlined conceptual digital health ecosystem and technology map, integrating the four previously mentioned components to facilitate the utilization of digital twins for medical applications. At the foundational level depicted in Fig. 2 , the Data Source Level encompasses diverse data inputs from healthcare providers within the PMDT ecosystem. As stated in the pilot study in section 3, these encompasses QUALITOP’s participating hospitals databases in the four EU countries in France, The Netherlands, Portugal and Spain. Here, patient health records from Electronic Health Records (EHRs) seamlessly integrate with provider-generated data such as medical visit records and PGHD like wellness and fitness information. This unified, longitudinal record offers a comprehensive view of a patient's medical history, contributing to the generation of refined health data. The resultant high-quality health data serves as a basis for informed decisions related to personalized early risk prediction, prevention, and guidance, facilitated by the utilization of AI-tools depicted in Fig. 2 . The Transformation Level in Fig. 2 normalizes captured patient-centric meta-data in a standardized format to ensure wide applicability and interoperability. Patient data generated by different systems will be matched, reconciled, harmonized, and semantically enhanced employing standard vocabularies and standards for the exchange of medical data such as the HL7 FHIR. A common set of FHIR profiles ( https://www.hl7.org/fhir/ ) here specify that data must be coded according to meaningful use terminologies, including RxNorm for medications, LOINC for observations, and SNOMED CT for health problems. To address the challenges of heterogeneity and interoperability, Fig. 2 illustrates that each healthcare provider or hospital within the PMDT ecosystem contributes its unique contextual embedding model. This ensures that no patient-level information persists within the acquired representations effectively reconciling issues related to heterogeneity and promoting interoperability. Then, health care organizations can federate and share their own local models and subsequently their own wealth of information without violating patient privacy. This federation of local data models operates as a surrogate for a unifying model, embodied in the form of PMDT knowledge structures. These structures amalgamate information sourced from a diverse array of medical providers and information outlets, creating a cohesive representation that transcends individual data sources. Referring to the pilot study in Section 3, Fig. 3 (a) illustrates a simplified view of the Hospital in Spain’ (one of the partners in the QUALITOP project) meta-data and Fig. 3 (b) presents fictional examples of corresponding data (real data cannot be presented in this article in compliance with the Data Transfer Agreements (DTAs) signed with respective hospitals). Due to space limitations, similar meta-data tables of other participating hospitals in France, the Netherlands and Portugal are not shown here. Figure 4 shows a simplified snippet of the PMDT semantic knowledge model meta-data mapping of the Hospital in Spain. The AI and Big Data Technologies Level: Leveraging Big Data technology offers the potential to fuse multi-scale clinical, biomedical, contextual, and behavioral data about each patient. Big Data technology in conjunction with trustworthy AI technologies helps provide decision support tools to facilitate optimized patient-centered, evidence-based decisions and AI-assisted services. At the remote databases' end, data is transmitted using a 'push' mechanism, diverging from a centralized 'pull' approach. This affords remote databases the autonomy to determine the frequency of data streaming and exercise control over the information to be transmitted, thereby preserving privacy. Integrated and structured data is seamlessly unified into the Patient Medical Digital Twin (PMDT), enabling AI tools to deliver personalized preventive measures and interventions. The PMDT Level employs advanced knowledge representation through abstract knowledge structures, or blueprints, to provide a comprehensive patient profile (details in Section 5). The interaction between AI, Big Data, and PMDT levels supports diverse applications, emphasizing personalized chronic disease management within a prevention-focused, outcome-driven healthcare model. For example, in the ecosystem depicted in Fig. 2 , this interaction can recommend an optimal treatment plan for a 10-year-old patient hospitalized within 28 days post-CAR-T cell infusion. The system considers treatment efficacy while aiming to minimize immune-related adverse events (IR-AEs), predicting their occurrence and assessing the impact on the patient’s quality of life (QoL) as discussed in Section 3. This approach enhances the overall healthcare experience. At the Stakeholder View Level, clinicians can access harmonized PMDT data, select personalized risk models (e.g., risk of hospitalization) from a model library, and train and validate these models using available patient risk factors, as illustrated in Fig. 2 . 5 The Patient Medical Digital Twin Figure 5 illustrates the PMDT proposed in this paper. The PMDT is shown to represent virtual cancer patients receiving Immunotherapy [22]. To this end the PMDT includes information about patient data and medical capability, and it also encapsulates policies and rules, safety procedures, treatment schedule information, past treatment performance data, and guidelines on how to improve a patient’s health. The concept of PMDT is essentially a patient-care-centric logical construct. 5.1 Medical Digital Twin Definition The Digital Twin definition illustrated in Fig. 5 is shown to encompass seven interconnected processable abstract knowledge types, referred to as blueprints, which encapsulate and convey medical knowledge: Medical Stakeholder Blueprint : this type represents and stores health professional capability data regarding which stakeholder is available for treatment, e.g., physicians, nurses, social-carers, nutritionists, etc., their skills, capabilities and capacity, responsibilities, and available health facilities Patient Blueprint : this abstract type represents and stores patient definition data including patient profile and characteristics, medication, treatment history, patient-generated data, health status data, nutrition and lifestyle data, physiological and psychosocial data. Disease Blueprint : this type represents and stores patient disease profile data such as, disease type, characteristics, diagnosis, symptoms, adverse events, tests, QoL impact, etc. For cancer patients it includes pathological data, genetic tests, tumor marker tests, symptoms, cancer subtype, concomitant Cancer, co-morbid conditions, and so on. Treatment Blueprint : this type captures current and historical treatment information, including treatment types and medication statements. Medical Pathways Blueprint : this type represents in a graphical form and stores treatment schedule data regarding what to cure and how to improve patient QoL. It includes personalized medical plans, agreed goals and actions, medical schedule, treatment sequences and transitions, treatment flow diagrams, personnel from the Medical Stakeholder Blueprint, diagnosis, clinical results, procedures, drug treatment, and so on. This abstract type uses Business Process Modeling and Notation 2.0. BPMN, an open executable standard that uses graphic symbols to represent a process, participants, choices, and flow. Figure 6 presents the BPMN model of the Management of the Cytokine Release Syndrome (CRS) as one of the common IR-AEs based on best practices recommendations in [22] on the management of CAR T-cell therapy. Figure 7 illustrates a simplified sub-process for managing Grade 1 CRS presented in Fig. 6 . Medical Safety Blueprint : this type represents and stores safety definition data. It encapsulates policies and rules, safety procedures, clinical practice guidelines, privacy regulations, and data security policies and practices. Treatment Performance Blueprint : this type represents and stores patient treatment performance data such as treatment performance indicators and attribute, e.g., patient readmission rate, workforce utilization rate, timeliness of care, improvement in costs, etc. The information encapsulated within the seven PMDT knowledge type definitions can be instantiated to portray a comprehensive representation of a patient. This instantiation aids clinicians and care teams in obtaining a holistic view of the patient's health, enabling a deeper understanding of how various contributing factors may influence health outcomes and overall QoL. Blockchain technology can be used in conjunction with PMDT contained information to help health information exchanges (HIEs) by relieving security concerns. To secure information exchange, each time caregivers provide patients with a PMDT service, they update their patients’ health data on a blockchain-enabled HIE. Each blockchain member has a private key, which is secure, and a public key that acts as a visible identifier. Because of these permission layers, patients can limit data access and share only the relevant parts of their medical records with their caregivers or other clinicians [23]. This represents a current ongoing extension to the work presented in this article 5.2 Processing the Medical Digital Twin Once a PMDT is defined and instantiated, it will be available for processing. An operational PMDT blends and cross-correlates information from the blueprint types in Fig. 5 . Simple processing can take the form of: Viewing and querying blueprints (by linking and cross-correlating information from diverse PMDT blueprints). Querying is classified into: Descriptive/retrieval querying : is a statistical interpretation used to analyze historical data to identify patterns and relationships, which usually answers the question of “What has happened?”. Example-1 : Referring to the pilot study in Section 3, “a policy maker might be interested in understanding the correlation between ‘Age’ (demographic data) and ‘Annual income’ (socio-economic data) of cancer patients’ information distributed among the four participating hospitals in Europe”. Figure 8 presents a screenshot of the platform implementation (cf. Section 9) showing the output of this query by displaying the quantitative correlation coefficient of these two factors and a dot chart/plot of the results. Analytical querying : uses homogenized data from the PMDT to predict future trends and events and forecast potential scenarios that can help drive strategic decisions. Example-2 : Analytical querying provides an answer to the question “What might happen in the future?”. An example of an analytical/predictive query is the predication of QoL of the 40-years-old female melanoma patient presented in Section 3. The output of this predictive query is an interactive bar chart showing the probability of each predicted adverse event for this particular patient as shown in Fig. 1 . Extending blueprint definitions by composing multiple blueprints of the same type, typically organized into a larger, composite blueprint, for instance, by adding more disease blueprints for elderly individuals diagnosed by melanoma with type 2 diabetes. The blueprint composition operator exhibits a closure property: each operation takes one or more compatible blueprints and returns a higher-level blueprint. Compatible blueprints are n-wise combinations of the same type of blueprints (e.g., combinations of disease-to-disease or treatment-to-treatment blueprints). This allows for blueprint composition using the merge operation that interconnects a set of blueprints end-to-end by describing all of their characteristics of a personalized patient replica. Adding (e.g., adding stamped periodic Patient Generated Health Data (PGHD) from sensors) and revising information contained in blueprints. For example, a diabetic melanoma patient using wearable technologies for continuous glucose monitoring (CGM) systems that continuously track glucose levels. Time stamped PGHDs are can be added to the PMDT replica if they do not already exist. Conventional complex PMDT processing operations include: Monitoring patient medication adherence : Sensor data may be effectively used in monitoring missed doses. Example-3 : A monitoring patient medication adherence example could be used whena wearable insulin pumps automatically delivers a dose of insulin when it is needed. Tracking of adverse events and drug efficacy : Sensor data are useful to close-track any adverse events and generate alerts for patients and caregivers. Similarly, any health improvements and drug-related benefits may be noted. Time point-based patient monitoring and vitals tracking : Remote patient monitoring provides care teams with the tools they need to remotely track the health of their patients at home measuring patient vital signs. Example-4 : Time point-based patient monitoring and vitals tracking allows to assign patients to specific care protocols and interventions that are tailored to condition or acuity level in order to establish better care and prevent serious adverse events. Time Series Analysis : Patient-generated data can be used for time series/trend analysis. This can be achieved with the help of various statistical analysis tools to predict and interpret results. Concurrent medications and drug interaction monitoring : Example-5 : Capturing sensor data and merging with concomitant medications and patient history data can be an effective way to prevent drug-related adverse events. More advanced processing abilities that rely exceedingly on AI and Machine Learning support include: Developing care plans and recommendations : Using the information contained in the PMDT, clinicians can improve their interactions with their patients and work jointly to develop care plans and recommendations, such as the one depicted in Fig. 6 . Example-6 : Referring to the pilot study in Section 3, a healthcare professional might be interested in developing a treatment/care plan for the hospitalization of a 10-year-old patient in the first 28 days after CAR-T infusion. Running in silico tests, comparisons, and simulations, and study performance issues : A synchronized linking of PMDT type instances is a time-stamped copy of a PMDT containing data and functions specific for an individual. These can be processed using a simple language to run in silico tests, comparisons, and simulations, and to study performance issues all with the goal of generating valuable insights that mirror the current status of an individual patient, predict the patient’s behavior and future states. Example-7 : for the 40-years-old female melanoma patient example presented in Section 3, it may be desirable to compare the performance of the treatment after three months of receiving Pembrolizumab 200 and contrast with her historical information. Answering specific disease questions : For example, PMDT instances can be used to answer questions such as ‘What is the expected QoL of a particular melanoma patient after one year of receiving immunotherapy treatment?’. Also, to predict the IR-AEs of a specific cancer patient with specific characteristics, like the analytical example presented in Section 3. Creating disease assessment reports : Example-8 : For instance, the creation of a melanoma assessment report that summarizes the impact of drinking and smoking habits on the diagnosis, clinical management, and the outcomes of combining multiple cancer treatment therapies. Using the AI tools in Fig. 2 , the PMDT offers a risk-free playground to explore innovations and test limitless “what-if” scenarios. Predictive AI techniques can be used to analyze each instantiation of the PMDT model to understand how ongoing lifestyle changes are impacting a specific patient’s health and suggest new tweaks that could speed up those improvements and reverse the effects of long-term conditions. 6 Implementation and validation The implementation 5 and validation of the PMDT digital health ecosystem presented in this article (cf. Figure 1 ) is a challenging yet requisite effort to ascertain the validity and soundness of the proposed approach. The following two subsections present the implementation and evaluation & validation efforts, respectively. 6.1 Implementation Figure 9 presents a standard UML component diagram showing the overall system architecture and its logical components. The identified and implemented components interact with each other via Application Programming Interfaces (APIs) using the Python programming language. As shown in Fig. 9 , the implementation comprises of four main components: (i) Portal component, (ii) Data Homogenization component, (iii) Federated Query Processing component, and (iv) Distributed Hospitals Database System components. In the following discussion, each component is discussed by showing its interactions with other components, which is followed by highlighting the iterative validation efforts conducted in the context of the QUALITOP project. Data Homogenization Component This component represents the backbone of the whole system. It consists of PatientMedicalDigitalTwinKB , which implements the detailed conceptual model of the PMDT (the detailed conceptual model of the PMDT is outside the scope of this article) in the Ontology Web Language 6 (OWL) standard using protégé 7 framework. The OWL implementation provides semantic support to the PMDT models, which is a vital requirement to surmount not only the syntactic and schematic data heterogeneity problem, but also the semantic heterogeneity of diverse medical data sources [24]. This brings intelligence to healthcare as discussed in Section 4. MetaDataManagement , which requests and imports meta-data from different involved HospitalDataBaseSystem(s). Such technical meta-data includes information about tables, columns, indexes, constraints, and any other relevant details in the disparate databases and is associated with the relevant blueprint entities in the unifying knowledge model so that they can be discovered in their respective databases. It also requests the “ MetaDataMapping ” component to carry out the meta-data mapping of the local databases against the PMDT schema, which acts as a global view over the local database systems (cf. Figure 3 and Fig. 4 ). This leads to simple query processing strategies [24]. The cross-schema meta-data mapping is then stored and maintained by the “ MetaDataManagement ” component. Portal Component : The PMDT digital health ecosystem is designed and implemented as a cross, private platform using Humhub 8 and by using the Python programming language. Thus, the platform is built to give intended users the tools to make communication and collaboration easy and successful. The portal component incorporates and implements a wide range of analytical (AI-powered) queries as described in Section 5. This component comprises of three sub-components: (i) PortalInterface , which includes all the GUIs between the user and the system to submit descriptive/analytical queries and display and manipulate the results. For example, the predication of QoL analytical query example presented in Section 3, named as query 1 thereafter, (ii) QueryTranslator : once query 1 is submitted via the GUI as shown in Fig. 1 , this component will then translate it into a SPARQL 9 query, which is the standard query language and protocol of OWL (the PMDT implementation), and (iii) Visualization component : once the “ FederatedQueryProcessing ” component processes the query as demonstrated below, the result will be visualized and presented on the portal’s integrated dashboards using appropriate graphs, charts, tables and advanced visualization capabilities (cf. Figure 1 ). Hybrid Federated Data Mesh Query Processing Component the “ QueryDecomposition ” sub-component receives the SPQRQL representation of query 1 from the “ QueryTranslator ” of the “ Portal ”. the “ QueryDecomposition ” sub-component then decomposes or replicates the query into a number of sub-queries, corresponding to the physical location of the queried data. That’s, for example Hospital-1 (e.g., in France), Hospital-2 (e.g., in Spain),…, Hospital-n. This is done by consulting the “ DataHomogenization” component for metadata mapping. For example, query 1 will be replicated by the the “ QueryDecomposition ” sub-component into sub-query 1 and sub-query 2 , assuming that Melanoma data exists in Hospotal-1 and Hospital-2. The “ QueryTranslator ” sub-component then translates each sub-query into the query languages accepted by the local database system, e.g., Structured Query Language (SQL). The “ RoutingComponent ” is then responsible for requesting local database systems of hospitals to process each respective SQL sub-query locally, and receives the sub-queries results. This is followed by the “ ResultAggregation ” sub-component aggregating and performing any further processing on the sub-queries results, and then forwarding the aggregated results to the “ Visualization ” sub-component of the “ Portal ” component. The “ ResultAggregation ” also sends the aggregated query result to the “ DataHomogenization ” component to be saved and maintained. The Federated Query Processing Component is implemented in Python using the Flask framework 10 for web integration. Its primary function is to decompose the incoming SPARQL query into multiple sub-queries that correspond to the physical locations of the queried data. Distributed Hospitals Database System Components The implementation architecture in Fig. 9 is designed as a federation of domain-oriented data products. Each data product is owned and managed by a specific medical domain. These data products expose standardized interfaces for other domain-specific data sources to access and use their data. This ensures the autonomy of the participating hospitals and preserves the privacy of medical data. Standardized APIs are implemented in Python based on JSON 11 . The use of APIs ensures effective data retrieval and enhances the interoperability of the system across different hospital databases. 6.2 Validation and Evaluation PMDTs represent the nerve center and the enabler of the digital health ecosystem presented in the article. PMDTs are based on a rigour formal foundation of Description Logic [25] through its design and implementation in OWL. Furthermore, the building blocks of the proposed PMDT-based health ecosystem, including the PMDT models, the platform, its collaboration features and analytical capabilities, were continuously evaluated, refined and improved by using it in its applications (see Section 5) and by discussing it with medical partners and patients’ representatives in the QUALITOP project. These efforts embody an agile/iterative knowledge engineering and system development approach for the development of the PMDT models and the collaborative analytical platform, respectively. Based on the analysis of the pilot study presented in Section 3, and the heavy involvement of QUALITOP medical partners in the agile/iterative analysis, design and development of the PMDT digital health ecosystem, through regular online and physical meetings spanning the lifetime of the QUALITOP project, medical partners and patients’ representatives: validated the applicability and efficacy of the PMDT knowledge models conceptually and technically, which went through continuous cycles of refinement and improvement. iteratively identified and validated the descriptive/analytical query capabilities presented in this article. experimented and validated the digital health platform as a user-friendly collaborative healthcare ecosystem. Usability is a major issue in today’s systems development, which measures how easily a user can accomplish their goals when using a service. This is achieved through research methodologies of customer satisfaction, which is attested by the medical partners. Medical partners and patients’ representatives appreciated the collaborative and advanced analytical capabilities of the smart digital platform, which are enabled by the formal PMDT as the nerve center of the ecosystem presented in Fig. 2 . Specifically, a team member said that “[…] AI-generated solutions have immense potential when developed with humility and a clear focus on the end-users—be it patients or healthcare professionals. Ensuring that these tools address well-defined healthcare challenges is critical to their success. […]. I am truly impressed by the outstanding work your team and consortium have accomplished. It’s essential that you leverage this success to make the tool accessible to both patients and doctors. […]”. However, the following points were raised to further improve the developed solutions. Some are ongoing work: Piloting the platform with real patients in a controlled environment to gather feedback and make necessary adjustments: discussions started with the participating hospitals in Europe to initiate this experiment. Ensuring that Data Protection Officer (DPO) and regulatory experts at participating medical institutions/hospitals assess the legal framework applicable to the project before its launch. Compliance with privacy laws, such as GDPR, is essential for building patient trust. Cited from one of the patients’ representatives: “[…] Patients care deeply about whether an ethics committee has reviewed the project or if a privacy expert has confirmed its GDPR compliance […]”: communications have started with participating hospitals in this regard. Moreover, as discussed in this article, the proposed PMDT ecosystem is privacy-preserving by nature, as the incorporated analytical capabilities illustrated in Section ‎5.2 are implemented following a federated data analytics approach [26]. Furthermore, the platform incorporates security and privacy measures at different levels including: (i) a multi-factor authentication method using Time-based one-time passwords or email verification, as well as manually verifying the identity and the role of the person registering to the platform, which is necessary in such a highly regulated domain to ensure the identity and role of the person signing up to the platform (ii) role-based access Control, and (iii) and input validation. Additionally, a GDPR notice is incorporated in all the GUIs of the platform. Figure 11 shows a screenshot of the GDPR privacy notice in a “Melanoma Patients” support group page, a collaborative feature supported by the platform. Patient-to-patient (P2P) engagement shown in Fig. 11 encourages controlled interaction among patients, fostering a supportive community that promotes sharing of experience and emotional support. To control P2P interactions, “Private chats” is disabled in this collaboration mode, furthermore, Natural Language Processing (NLP) techniques will be employed to ensure that the interactions between patients is only for sharing experiences and for emotional support, and no medical advice are given in such communications (details of the collaborative capabilities of the platform is outside the scope of this article) [5] Video demonstrations of the platform and its analytical module are available at: https://drive.google.com/drive/folders/1BmFfh2C1yEB5PJeCX0sdgxDdpydsEgKg?usp=sharing [6] OWL: www.w3.org/OWL/ [7] https://protege.stanford.edu/ [8] HumHub: https://www.humhub.com/en [9] SPARQL 1.1 Query Language: https://www.w3.org/TR/sparql11-query/ [10] Flask framework: https://flask.palletsprojects.com/ [11] JavaScript Object Notation: https://www.json.org/json-en.html 7 Conclusions and Future Work This paper proposed a DT as a full personalized in silico replica of a patient. This is a virtual model (data plus algorithms) with special features not found in traditional models and simulations, one that dynamically pairs the physical (patient and treatment) and digital worlds, and AI in order to support improved collaboration between patients and healthcare professionals to achieve disease prevention goals and understanding of when one is in danger of developing a disease or their condition deteriorates. By making highly resolved digital representations of individuals available, the DT concept enables precise personalized health protection and medical care. The PMDT is fully implemented in OWL as part of a collaborative platform and was continuously validated by medical partners and patients’ representatives to ensure it applicability, efficacy and usability. It is crucial to work against cancer inequalities . Tools have to be produced for the benefit of the patients, equally accessible for all, no matter the country of residence, the economical and social status of the patient, the beliefs, the health care unit he/she received treatment: theoretically the developed technological solutions presented in this article satisfy this requirement, as the PMDT, the platform, its collaborative features, and analytical capabilities are extensible in nature, because they are built on rigor formal models of the PMDTs and by following a structured agile system development approach, which enforces the loose coupling design principle as well as the open-closed-principle (software systems must be open for extensions, closed for modifications). Future work directions include: (i) the utilization of Blockchain technology with PMDT containing information to help health information exchanges (HIEs) by relieving security concerns, (ii) the incorporation of virtual and augmented reality technologies, and (iii) experimenting the proposed approach with colorectal cancer screening as part of the ONCOSCREEN 12 project. [12] ONCOSCREEN project: https://cordis.europa.eu/project/id/101097036 Declarations Funding This research is partially funded by the EC Horizon 2020 project QUALITOP, under contract number H2020 - SC1-DTH-01-2019 – 875171. Availability of Data and Materials QUALITOP cohort study has been registered as an observational study at www.clinicaltrials.gov. Trial registration number NCT05626764. Statements and Declarations Ethical Approval Not applicable Competing Interests The authors declare no financial or non-financial competing interests. Acknowledgements We wish to thank Hospital Clinic Barcelona (IDIBAPS), Hospices Civils de Lyon (HCL), University Medical Center Groningen (UMCG), and Instituto Português de Oncologia, Lisboa (IPOL) for providing the essentials of the pilot study in this paper, and for the continuous evaluation and validation of the reported results. Special thanks goes to Dr Roxana Albu, Chief Scientific Officer, Association of European Cancer Leagues (ECL), and Dr Menia Koukougianni, Fellow of the European Patients' Academy, for their insightful feedback in evaluating and validating the work presented in this paper. References Zahid A, Poulsen JK, Sharma R, Wingreen SC (2021) A systematic review of emerging information technologies for sustainable data-centric health-care. Int J Med Informatics 149:104420. https://doi.org/10.1016/j.ijmedinf.2021.104420 World Health Organization. Noncommunicable diseases (2020) April https://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases Jain P, Agarwal A, Behara RS, Baechle C (2019) HPCC based framework for COPD readmission risk analysis. J Big Data 6:1–13 Lopez-Perez L, Hernández L, Ottaviano M, Martinelli E, Poli T, Licitra L et al BD2Decide: Big Data and Models for Personalized Head and Neck Cancer Decision Support. 2019 IEEE 32nd International Symposium on Computer-Based Medical Systems (CBMS)2019. pp. 67 – 8 Akbar W, Wu WP, Faheem M, Saleem S, Javed A, Saleem MA Predictive Analytics Model Based on Multiclass Classification for Asthma Severity by Using Random Forest Algorithm. 2020 International Conference on Electrical, Communication, and Computer Engineering (ICECCE)2020. pp. 1–4 Jing S, Hou K, Yan J, Ho Z-P (2020) An application of big data mining and fuzzy prediction of total knee replacement surgery patients attributes --Fuzzy TKR patients attributes prediction. J NonLinear Convex Anal 21(8):1689–1702 Satti FA, Ali T, Hussain J, Khan WA, Khattak AM, Lee S (2020) Ubiquitous Health Profile (UHPr): a big data curation platform for supporting health data interoperability. Computing 102(11):2409–2444. 10.1007/s00607-020-00837-2 Sivaparthipan CB, Muthu BA, Manogaran G, Maram B, Sundarasekar R, Krishnamoorthy S et al (2019) Innovative and efficient method of robotics for helping the Parkinson's disease patient using IoT in big data analytics. Trans Emerg Telecommunications Technol. ;31(12) Ali F, El-Sappagh S, Islam SMR, Ali A, Attique M, Imran M et al (2021) An intelligent healthcare monitoring framework using wearable sensors and social networking data. Future Gener Comput Syst 114:23–43 Sun K, Yang D (2021) Microprocess Microsyst 81:103793. https://doi.org/10.1016/j.micpro.2020.103793 . Human health big data evaluation based on FPGA processor and big data decision algorithm Chakshu NK, Carson J, Sazonov I, Nithiarasu P (2019) A semi-active human digital twin model for detecting severity of carotid stenoses from head vibration-A coupled computational mechanics and computer vision method. Int J Numer Methods Biomed Eng 35(5):e3180. 10.1002/cnm.3180 Kovatchev B (2019) A Century of Diabetes Technology: Signals, Models, and Artificial Pancreas Control. Trends Endocrinol Metabolism 30(7):432–444. https://doi.org/10.1016/j.tem.2019.04.008 Grande Gutierrez N, Mathew M, McCrindle BW, Tran JS, Kahn AM, Burns JC et al (2019) Hemodynamic variables in aneurysms are associated with thrombotic risk in children with Kawasaki disease. Int J Cardiol 281:15–21. 10.1016/j.ijcard.2019.01.092 Wang Y, Fu T, Xu Y, Ma Z, Xu H, Du B et al (2024) TWIN-GPT: Digital Twins for Clinical Trials via Large Language Model. ACM Trans Multimedia Comput Commun Appl. 10.1145/3674838 Smarter devices (2020) better patient care. MIT Technology Review InsightsJuly Mazumder O, Roy D, Bhattacharya S, Sinha A, Pal A Synthetic PPG generation from haemodynamic model with baroreflex autoregulation: a Digital twin of cardiovascular system. 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC)2019. pp. 5024-9 Allen A, Siefkas A, Pellegrini E, Burdick H, Barnes G, Calvert J et al (2021) A Digital Twins Machine Learning Model for Forecasting Disease Progression in Stroke Patients. Appl Sci 11(12):5576 Corral-Acero J, Margara F, Marciniak M, Rodero C, Loncaric F, Feng Y et al (2020) The 'Digital Twin' to enable the vision of precision cardiology. Eur Heart J 41(48):4556–4564. 10.1093/eurheartj/ehaa159 Łukaniszyn M, Majka Ł, Grochowicz B, Mikołajewski D, Kawala-Sterniuk A (2024) Digital Twins Generated by Artificial Intelligence in Personalized Healthcare. Appl Sci 14(20):9404 Vinke PC, Combalia M, de Bock GH, Leyrat C, Spanjaart AM, Dalle S et al (2023) Monitoring multidimensional aspects of quality of life after cancer immunotherapy: protocol for the international multicentre, observational QUALITOP cohort study. BMJ open 13(4):e069090. 10.1136/bmjopen-2022-069090 Rahman A, Hossain MS, Muhammad G, Kundu D, Debnath T, Rahman M et al (2022) Federated learning-based AI approaches in smart healthcare: concepts, taxonomies, challenges and open issues. Cluster Comput 26(4):2271–2311. 10.1007/s10586-022-03658-4 Hayden PJ, Roddie C, Bader P, Basak GW, Bonig H, Bonini C et al (2022) Management of adults and children receiving CAR T-cell therapy: 2021 best practice recommendations of the European Society for Blood and Marrow Transplantation (EBMT) and the Joint Accreditation Committee of ISCT and EBMT (JACIE) and the European Haematology Association (EHA). Annals oncology: official J Eur Soc Med Oncol 33(3):259–275. 10.1016/j.annonc.2021.12.003 Gordon R, Perlman M (2018) Shukla. The hospital of the future: How digital technologies can change hospitals globally. Deloitte Center for Health Solutions Cruz IF, Xiao H (2005) The role of ontologies in data integration. Int J Eng Intell Syst Electr Eng Commun 13:245–252 Markus K, Frantisek S, Ian HA (2012) Description Logic Primer. eprint arXiv:1201.4089. Cornell University Aledhari M, Razzak R, Parizi RM, Saeed F (2020) Federated Learning: A Survey on Enabling Technologies, Protocols, and Applications. IEEE Access 8:140699–140725. 10.1109/ACCESS.2020.3013541 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5826330","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":401959076,"identity":"b12d9a90-85bd-4888-b1c2-2eff4205bfff","order_by":0,"name":"Michael P. Papazoglou","email":"","orcid":"","institution":"Scientific Academy for Service Technology e.V. (ServTech)","correspondingAuthor":false,"prefix":"","firstName":"Michael","middleName":"P.","lastName":"Papazoglou","suffix":""},{"id":401959077,"identity":"4b173356-78c1-49a2-81e9-71bbaa599598","order_by":1,"name":"Bernd J. Krämer","email":"","orcid":"","institution":"Scientific Academy for Service Technology e.V. (ServTech)","correspondingAuthor":false,"prefix":"","firstName":"Bernd","middleName":"J.","lastName":"Krämer","suffix":""},{"id":401959078,"identity":"cbefcc32-1537-4574-bc32-dfba74195afd","order_by":2,"name":"Mira Raheem","email":"","orcid":"","institution":"Cairo University","correspondingAuthor":false,"prefix":"","firstName":"Mira","middleName":"","lastName":"Raheem","suffix":""},{"id":401959079,"identity":"d6a53add-5781-4cad-8ca7-1cc9e157b4a6","order_by":3,"name":"Amal Elgammal","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABD0lEQVRIie3OMUvDQBTA8RcCmQJdHxTbr3ClIC2G61e54yBdGnDM0CEgxE234lD0KyhC55NAF/sBCm+Ju0MmcbI+WyiCXsVN8P7LXcL97h2Az/c3CyzAYLetc4CIV/zJ2P0Ztfo10eVuPUhas6q2kKO8nD3qRl8n+gLCB4ohkYWDIKXCwgrNFWX3qBcpD4rMSQypcREgJWxQogHKbplUTOLjdgyVcYkujRsbvKHp8pRXPd8wab0w2TiJoAlPKVAKyhaoC/sxJWJipYv0aHJq1RJVj8lALU2/DKP+cC6McpEOje/qZpqMOvywdTOVRzfnZ0/r51yOXGQbX6iL/Ve4ffDnP9/39c7DU3w+n+8/9Q5sq1leWTVBoQAAAABJRU5ErkJggg==","orcid":"","institution":"Cairo University","correspondingAuthor":true,"prefix":"","firstName":"Amal","middleName":"","lastName":"Elgammal","suffix":""}],"badges":[],"createdAt":"2025-01-14 10:38:15","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5826330/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5826330/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":73943276,"identity":"cc64db9d-7a04-47e4-8546-0f7dfb0aaeeb","added_by":"auto","created_at":"2025-01-16 08:21:29","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":88917,"visible":true,"origin":"","legend":"\u003cp\u003eDigital Health Platform Analytical Software Module\u003c/p\u003e","description":"","filename":"image1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-5826330/v1/3cec0645f82ccbfc79b2df2d.jpeg"},{"id":73942928,"identity":"585bd836-992f-49b4-bce4-6bdd5cbd4c13","added_by":"auto","created_at":"2025-01-16 08:13:28","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":331656,"visible":true,"origin":"","legend":"\u003cp\u003eConceptual architecture of the digital health ecosystem and technology map\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-5826330/v1/a43bcb51145666c2bfe7e5fb.png"},{"id":73942929,"identity":"7970acbd-7829-465b-a0d2-4a4a3517ba6c","added_by":"auto","created_at":"2025-01-16 08:13:28","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":364776,"visible":true,"origin":"","legend":"\u003cp\u003e(a) Hospital in Spain Meta Data; (b) Fictional data in the hospital in Spain\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-5826330/v1/7698580d6cf2bf70c4b41bd3.png"},{"id":73942940,"identity":"47ffd6ea-4714-41a3-b578-aecd690cdfd8","added_by":"auto","created_at":"2025-01-16 08:13:29","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":214438,"visible":true,"origin":"","legend":"\u003cp\u003eA snippet of PMDT semantic knowledge model mapping of the Hospital in Spain\u003c/p\u003e","description":"","filename":"image4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-5826330/v1/3c1f3ce08be5ef8b2a9483ed.jpeg"},{"id":73942937,"identity":"52d5fdf8-284a-48fa-99f2-56a27d7edd11","added_by":"auto","created_at":"2025-01-16 08:13:29","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":252630,"visible":true,"origin":"","legend":"\u003cp\u003eThe extensible PMDT model and its abstract types\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-5826330/v1/500cea71407ccea5077f1060.png"},{"id":73942935,"identity":"8194ab7b-1840-4ac4-a162-c3c2b20e4e19","added_by":"auto","created_at":"2025-01-16 08:13:28","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":84815,"visible":true,"origin":"","legend":"\u003cp\u003eManagement of Cytokine release syndrome\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-5826330/v1/5b2e204cde41758e0312afa2.png"},{"id":73942941,"identity":"69d21836-f7d5-42e0-a130-d7e3f6117f80","added_by":"auto","created_at":"2025-01-16 08:13:29","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":94548,"visible":true,"origin":"","legend":"\u003cp\u003eManage Grade 1 CRS BPMN model\u003c/p\u003e","description":"","filename":"image7.png","url":"https://assets-eu.researchsquare.com/files/rs-5826330/v1/433e278acf750cdc1dbf9f88.png"},{"id":73943281,"identity":"05f5bc6a-b81c-4739-afc9-4ac983abccd1","added_by":"auto","created_at":"2025-01-16 08:21:30","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":35099,"visible":true,"origin":"","legend":"\u003cp\u003eA descriptive query result showing the relationship between 'Age' and 'Annual Income' information processed in a federated fashion\u003c/p\u003e","description":"","filename":"image8.png","url":"https://assets-eu.researchsquare.com/files/rs-5826330/v1/4aa11a592013e5e4541d5ecb.png"},{"id":73942961,"identity":"a9f8030a-29b5-45ef-a5e2-dabd67a412db","added_by":"auto","created_at":"2025-01-16 08:13:29","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":91022,"visible":true,"origin":"","legend":"\u003cp\u003eComponent diagram of the PMDT digital health ecosystem implementation\u003c/p\u003e","description":"","filename":"image9.png","url":"https://assets-eu.researchsquare.com/files/rs-5826330/v1/acf0a0f19c93802bc4d82f9d.png"},{"id":73942962,"identity":"808bd32b-ea4f-4138-8e41-e054f3ae851c","added_by":"auto","created_at":"2025-01-16 08:13:29","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":261082,"visible":true,"origin":"","legend":"\u003cp\u003eScreenshot of PMDT implementation as an ontology in OWL 2.0 using protégé\u003c/p\u003e","description":"","filename":"image10.png","url":"https://assets-eu.researchsquare.com/files/rs-5826330/v1/41c901bb63d4a9f88a86a7d5.png"},{"id":73943274,"identity":"47841f86-a9a8-4b84-857a-4d260006ad3a","added_by":"auto","created_at":"2025-01-16 08:21:29","extension":"jpeg","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":198787,"visible":true,"origin":"","legend":"\u003cp\u003eGDPR notice incorporated in the portal implementation\u003c/p\u003e","description":"","filename":"image11.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-5826330/v1/aab139fdc350e4e5529bd762.jpeg"},{"id":74305839,"identity":"16e9d8fc-6db1-46af-a579-3faae2929a76","added_by":"auto","created_at":"2025-01-21 00:01:18","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2674023,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5826330/v1/6de26c9f-b412-447b-be75-548358fed40f.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Medical Digital Twins for Personalized Chronic Care","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eDigital Twin (DT) technology is an emerging field that enables the creation of digital representations of real-world entities, which can be either physical or perceived (e.g., processes), and tracks the state of these entities during their lifetime. DT can perform bi-directional automated data flow between a physical object and its digital representation. Changes experienced by the physical object can be reflected in its digital model and the insights gained from the digital model can support decisions to optimize the physical object. The integration of the Digital Twin paradigm into healthcare has the potential to revolutionize care processes, create patient-specific digital models of human organs and individual cells, improve patient experience, reduce operational costs, and enhance the overall quality of personalized and preventative care [1].\u003c/p\u003e \u003cp\u003eOne of the areas that can benefit from the use of DT technology in healthcare is chronic disease management. Here, a DT can be used to address the wellness, prevention, and ongoing management of focused chronic disease conditions. According to the World Health Organization, chronic diseases, such as cardiovascular disease, diabetes, and Alzheimer\u0026rsquo;s, account for nearly three-quarters of all deaths worldwide [2]. Yet many of these chronic diseases are preventable, as they are linked to risk factors such as poor diet, obesity, and lifestyle choices. The DT paradigm is a powerful tool for personalized chronic disease prevention, enabling early symptom detection and the implementation of preventive measures and targeted interventions tailored to individual patients. This approach has the potential to proactively address identified risk factors, potentially reversing or preventing the onset of chronic diseases.\u003c/p\u003e \u003cp\u003eOne of the biggest challenges in applying DT technology to the healthcare ecosystem is that caregivers currently lack a holistic view of their patient needs, often 'swivel-chairing' between multiple systems to manage basic interactions such as scheduling treatments and engaging in patient outreach. Patient and clinician access to patient-generated health data to monitor specific chronic conditions is rarely linked to patient longitudinal patient data. This results in fragmented care, due to poor communication and fragmented sources of patient data that are not yet widely integrated. This makes it much more difficult to align and coordinate chronic care across care teams and associated settings.\u003c/p\u003e \u003cp\u003eA real challenge in providing optimal care for chronic conditions is the need to transition from an episodic to a continuous monitoring cycle of the patient\u0026rsquo;s physiological data and the need to develop collaborative relationships with patients. This can be achieved by sustaining vigilant monitoring of health indicators to optimize treatment and an operational shift that gives patients greater control and personalization over how they manage their condition.\u003c/p\u003e \u003cp\u003eDTs can help alleviate these problems by modeling an individual\u0026rsquo;s genetic makeup, physiological characteristics, and lifestyle habits by acting as a critical harbinger of a \u0026ldquo;health-tech\u0026rdquo; approach toward prevention-focused and outcome-oriented healthcare. The DT can achieve this by collecting, collating, and analyzing masses of an individual\u0026rsquo;s health data to help medical professionals to better predict a patient\u0026rsquo;s (long-term) response, including side effects, recommend a personalized treatment plan, provide therapy guidance, and prevent deterioration. We call such a DT a Patient Medical Digital Twin (PMDT).\u003c/p\u003e \u003cp\u003eThe data that will provide the information used to represent an individual\u0026rsquo;s PMDT will be gathered from a variety of health data sources, including medical sources, electronic health records (EHRs), vital signs, a patient\u0026rsquo;s medical history (diagnosis and prescriptions), medical and clinical data, symptoms, medical tests, medications, physiological and psychosocial data, nutritional data, patient-generated data, etc. These data are used by different entities in the care continuum and are presented in different formats. A PMDT can bring together and codify such rich data. It can also continuously pull in real-time sensor (IoT) data to create accurate snapshots of the physical patient. This information can be integrated with historical data and predictive analytics to alert healthcare professionals to potential problems and suggest solutions.\u003c/p\u003e \u003cp\u003eOur contribution can be summarized as follows: the development of a PMDT - a faithful digital (in vitro) model of the patient - as the linchpin in the transition from one-treatment-fits-all to smart personalized healthcare. This digital model acts as a rich knowledge representation framework for decision support, encompassing a complete patient profile. It achieves this by (a) recognizing a patient\u0026rsquo;s current state, (b) predicting potential developments, and (c) facilitating interaction with caregivers and healthcare professionals to collaboratively develop a patient-tailored treatment plan, thereby advancing the principles of precision medicine.\u003c/p\u003e \u003cp\u003eThe remainder of this article is structured as follows: a literature review is presented in Section 2. Section 3 introduces the pilot application, which is used throughout this article. The conceptual framework for the PMDT is discussed in Section 4, followed by the PMDT and its processing environment in Section 5. The implementation and validation of both are sketched in Section 6. The paper concludes in Section 7 by highlighting ongoing and future work.\u003c/p\u003e"},{"header":"2 Literature Review","content":"\u003cp\u003eDT-based medicine is still in its infancy and has been described in the literature as the nexus of four components - data science, software engineering, representation of expert knowledge, AI and machine learning, - which culminates in tools that support effective clinical decision-making.\u003c/p\u003e \u003cp\u003eThe study in [1] systematically reviewed emerging technologies for data modelling and analytics for sustainable data-centric healthcare that marked DT technology as a recent inclusion in the healthcare industry, which has the power to revolutionize the healthcare paradigm. This study reviewed eight DT related studies [3\u0026ndash;10], and the findings revealed that each study has a limited analytical focus with no studies addressing the integration of heterogeneous and diverse data sources. This study concluded that none of the examined data modelling and analytics technologies for sustainable healthcare operates in isolation. It became apparent that for sustained viability, technological solutions must offer seamless and proficient capabilities for data capture, storage, and analytics. These highlight the contributions of the work presented in this article.\u003c/p\u003e \u003cp\u003eVarious initiatives have emerged from building DTs of organs, such as the heart, to observe how they might respond to various interventions, thus minimizing the risk of early human trials and accelerating the availability of the treatments to patients. Simulated organs could change how medicine works, making it hyper-personal and less invasive. For example, Hewlett Packard Enterprise deployed its supercomputer to create digital models of the brain for research purposes, while Siemens Healthineers has a DT model to simulate the use of cardiac resynchronization therapy. Dassault released the Living Heart, a realistic model that accounts for electricity, mechanics, and blood flow. The software can turn a 2-D scan from an individual human into a personalized full-dimensional model of his or her heart and run hypotheticals. Research in [11] proposes a DT for the human head to detect carotid stenosis severity from a video of a human face with the help of a coupled blood flow and head vibration model. This DT model represents an attempt to link a video of a patient face to the percentage of carotid occlusion. Other existing prominent examples of digital twins in healthcare include \u0026ldquo;the artificial pancreas\u0026rdquo; [12], and pediatric cardiac digital twins [13]. The study in [14] propose TWIN-GPT, a digital twin approach form clinical trials by utilizing Large Language Models (LLMs). All these examples focus on just one single aspect of the human body due to its extreme complexity.\u003c/p\u003e \u003cp\u003eCommercial medical device manufacturers are also increasingly using DTs to model both devices and patients to better design devices for people with specific conditions [15]. Further companies are using CT scans and MRI images to create three-dimensional computational models of individual patients that will help clinicians decide on and prepare for surgeries or other procedures.\u003c/p\u003e \u003cp\u003eOther research initiatives provide the opportunity to explore predicted disease trajectories for a patient and have the potential to inform the patient\u0026rsquo;s current state using data derived from that individual to avoid providing a generalized prediction. In [16] the authors present a DT model to simulate and understand the progression of valvular disease associated with the mitral valve. This DT model integrates cardiac electrophysiology with hemodynamic modeling giving a broader understanding of the effect of disease progression on various parameters like ejection fraction, cardiac output, blood pressure, etc., to assess the severity of mitral valve disorders.\u003c/p\u003e \u003cp\u003eSeveral studies have also focused on assisting understanding or management of a target condition or class of conditions with DTs. Research in [17] employs machine learning techniques applied to a DT model based on a variational autoencoder for simulating the clinical trajectories of patients who went on to experience an ischemic stroke. Research presented in [18] claims that precision cardiology will be delivered in a synergetic fashion that combines induction, by using statistical models learnt from data, and deduction, through mechanistic modelling and simulation integrating multiscale knowledge and data, which comprise the two foundation pillars of the DT. Authors in [19] presented an AI-generated DTs approach to support the rapid assessment of in silico intervention strategies.\u003c/p\u003e \u003cp\u003eThe medical Digital Twin in this paper represents the convergence of Big Data, Knowledge Representation technology, software engineering techniques, and AI endeavors. Through a twinning process, it comprehensively addresses an individual's physiological, historical, behavioral, and biological conditions. The medical Digital Twin in this paper represents a transformative leap from one-size-fits-all treatments to smart, personalized healthcare. The PMDT achieves this by:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eRecognizing the patient\u0026rsquo;s current health state, enabling accurate assessments.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003ePredicting potential health developments, allowing for proactive interventions.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eFacilitating collaboration between caregivers and healthcare professionals, ensuring the creation of a tailored treatment plan that aligns with the principles of precision medicine.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eThis innovative approach centers on creating a faithful digital (in vitro) replica of the patient, serving as a comprehensive knowledge framework for decision support. Unlike traditional methods, the PMDT provides a holistic view of the patient's health, identifying potential disease symptoms, tracking transitions, and signaling impending deterioration. The overarching objective is to create a personalized, life-long in silico replica of a patient, capable of predicting the individual's responses to medical conditions.\u003c/p\u003e \u003cp\u003eBy integrating these capabilities, the PMDT not only enhances decision-making but also empowers a proactive, patient-centered approach to healthcare, delivering improved outcomes and advancing the vision of precision medicine. This model is designed to facilitate personalized therapy simulation, selection, and outcome prediction.\u003c/p\u003e"},{"header":"3 Pilot Study","content":"\u003cp\u003eThe pilot study presented in this section is an international, observational multicenter cohort study that was conducted in France, The Netherlands, Portugal and Spain and included adult patients treated with CAR T-cell or immune checkpoint inhibitors therapy\u003csup\u003e1\u003c/sup\u003e [20]. The pilot was conducted in the context of the EU H2020 QUALITOP project\u003csup\u003e2\u003c/sup\u003e - which partially funds this research.\u003c/p\u003e\n\u003cp\u003eThe main objective of the QUALITOP project is to develop and implement an IT-based European immunotherapy platform that utilizes big data analytics and artificial intelligence to collect and aggregate efficiently and effectively real-world Quality of Life (QoL) data, monitor patients\u0026rsquo; health status and manage patients with personalized and preventive focus, to improve their QoL. Immunotherapy is considered one of the complicated treatments protocols as it causes toxicities or side effects, known as Immune-Related Adverse Events (IRAEs), that are challenging to predict because they are not caused by mechanisms involved in other treatment types, such as chemotherapy and radiation [20].\u003c/p\u003e\n\u003cp\u003eIn the cohort pilot study, adult patients are monitored in their real-life, which overcomes the known limitations of Randomized Clinical Trails (RCTs) [20], where patients\u0026rsquo; profiles and behaviours may be far different from those seen in RCTs. Relevant historical real-world databases and medical administrative registries were also identified. The study included patients recruited specifically for QUALITOP. Patients\u0026rsquo; clinical health status and QoL are monitored up until 18 months post treatment initiation. Psycho-social wellbeing is also monitored with a tailored QUALITOP questionnaire at baseline and 3, 6, 12 and 18 months post treatment initiation.\u003c/p\u003e\n\u003cp\u003eTo identify advanced analytical needs of medical institutions across these four EU countries, we have followed a highly iterative/agile software development methodology that started with a structured iterative Requirements Engineering (RE) approach to identify analytical query patterns of recurring analytical requirements, and then incrementally modeled and formalized identified analytical patterns.\u003c/p\u003e\n\u003cp\u003eAs will be illustrated in the next discussion, a virtual data lake [6] and the PMDTs models enabled the implementation of these analytical query patterns by utilizing a federated data analytics approach, which is an innovative decentralized machine learning paradigm [21] that facilitates collaborative model training across multiple parties without the necessity of centralizing sensitive data in compliance with regulatory bodies such as GDPR\u003csup\u003e3\u003c/sup\u003e and HIPAA\u003csup\u003e4\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eTo improve understanding, we assume that: \u0026ldquo;\u003cem\u003ea medical professional in Amsterdam is interested in predicting the QoL of his female melanoma patient, who is 40-years-old, classified with TNM stage \u0026quot;T3AN2Cm0\u0026quot; and undergoing treatment with \u0026quot;Pembrolizumab: 200 mg\u0026quot; at a frequency of \u0026quot;Q3W\u003c/em\u003e\u0026rdquo;. This example is an instantiation of the textual analytical query pattern \u0026ldquo;Predict QoL of \u0026lt;\u0026thinsp;cancer-type\u0026thinsp;\u0026gt;\u0026thinsp;patient with \u0026lt;\u0026thinsp;specific-characteristics\u0026gt;\u0026rdquo;. By using the graphical web interface shown in Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e, the medical professional may specify the cancer-type and the specific-characteristics of her cancer patient and submit the analytical query request. In the following Sections, this example is used as a running scenario showing the workflow of its execution based on the novel digital twinning approach presented in this article.\u003c/p\u003e\n\u003cp\u003e[1] The QUALITOP cohort study has been registered as an observational study at www.clinicaltrials.gov. Trial registration number NCT05626764\u003c/p\u003e\n\u003cp\u003e[2] Monitoring multidimensional aspects of QUAlity of Life after cancer ImmunoTherapy - an Open smart digital Platform for personalized prevention and patient management: https://cordis.europa.eu/project/id/875171/reporting\u003c/p\u003e\n\u003cp\u003e[3] General Data Protection Regulation: https://gdpr.eu\u003c/p\u003e\n\u003cp\u003e[4] Health Insurance Portability and Accountability Act: https://www.ncbi.nlm.nih.gov/books/NBK500019/\u003c/p\u003e"},{"header":"4 Digital Twin Conceptualization Scheme","content":"\u003cp\u003eA PMDT environment would comprise dispersed patient data sources, sensors, analytics, and visualization. Patient information is often distributed among various healthcare providers who have treated the individual for diverse symptoms and conditions throughout their lifetime. The imperative lies in the seamless integration of this information to offer a comprehensive and unified view of the patient's health history. Otherwise, misdiagnoses, inappropriate medications, duplicate tests, medical-legal issues, and other problems are inevitable. Patient-Generated Health Data (PGHD) involves sensor data from wearable sensors and data generated from handheld devices. These may be used for a variety of purposes including recommendations, alerts, self-assessment, daily activity tracking, time of medications, current nutritional status, and more. The use of PGHD can empower patients and caregivers to manage their health and to collaborate with clinicians via shared decision-making that considers patients\u0026rsquo; concerns and preferences.\u003c/p\u003e \u003cp\u003eAI-based analytics tools will be used to process patient data to derive insights about current medical conditions as well as make predictions about emerging conditions. This leads to the possibility of targeted chronic disease interventions at an early stage (the notion that prevention is better than cure). It will also be possible to factor in wider aspects of health, such as diet and exercise in chronic disease treatment. Finally, visualization typically collates diverse medical data coming from various sources into a single unified view, allowing users \u0026ndash; clinicians, non-specialist staff, and patients \u0026ndash; to piece everything together and get a complete overview of the patient\u0026rsquo;s situation, which unveils insight into patterns and correlations to help medical staff interpret data analytics results faster, recognize trends, and make better decisions as well as aid patients in interpreting and contextualizing their health information.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e presents a streamlined conceptual digital health ecosystem and technology map, integrating the four previously mentioned components to facilitate the utilization of digital twins for medical applications. At the foundational level depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, the Data Source Level encompasses diverse data inputs from healthcare providers within the PMDT ecosystem. As stated in the pilot study in section 3, these encompasses QUALITOP\u0026rsquo;s participating hospitals databases in the four EU countries in France, The Netherlands, Portugal and Spain.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eHere, patient health records from Electronic Health Records (EHRs) seamlessly integrate with provider-generated data such as medical visit records and PGHD like wellness and fitness information. This unified, longitudinal record offers a comprehensive view of a patient's medical history, contributing to the generation of refined health data. The resultant high-quality health data serves as a basis for informed decisions related to personalized early risk prediction, prevention, and guidance, facilitated by the utilization of AI-tools depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eThe Transformation Level in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e normalizes captured patient-centric meta-data in a standardized format to ensure wide applicability and interoperability. Patient data generated by different systems will be matched, reconciled, harmonized, and semantically enhanced employing standard vocabularies and standards for the exchange of medical data such as the HL7 FHIR. A common set of FHIR profiles (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.hl7.org/fhir/\u003c/span\u003e\u003cspan address=\"https://www.hl7.org/fhir/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) here specify that data must be coded according to meaningful use terminologies, including RxNorm for medications, LOINC for observations, and SNOMED CT for health problems.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo address the challenges of heterogeneity and interoperability, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e illustrates that each healthcare provider or hospital within the PMDT ecosystem contributes its unique contextual embedding model. This ensures that no patient-level information persists within the acquired representations effectively reconciling issues related to heterogeneity and promoting interoperability. Then, health care organizations can federate and share their own local models and subsequently their own wealth of information without violating patient privacy. This federation of local data models operates as a surrogate for a unifying model, embodied in the form of PMDT knowledge structures. These structures amalgamate information sourced from a diverse array of medical providers and information outlets, creating a cohesive representation that transcends individual data sources.\u003c/p\u003e \u003cp\u003eReferring to the pilot study in Section 3, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e (a) illustrates a simplified view of the Hospital in Spain\u0026rsquo; (one of the partners in the QUALITOP project) meta-data and Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e (b) presents fictional examples of corresponding data (real data cannot be presented in this article in compliance with the Data Transfer Agreements (DTAs) signed with respective hospitals). Due to space limitations, similar meta-data tables of other participating hospitals in France, the Netherlands and Portugal are not shown here. Figure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e shows a simplified snippet of the PMDT semantic knowledge model meta-data mapping of the Hospital in Spain.\u003c/p\u003e \u003cp\u003eThe AI and Big Data Technologies Level: Leveraging Big Data technology offers the potential to fuse multi-scale clinical, biomedical, contextual, and behavioral data about each patient. Big Data technology in conjunction with trustworthy AI technologies helps provide decision support tools to facilitate optimized patient-centered, evidence-based decisions and AI-assisted services. At the remote databases' end, data is transmitted using a 'push' mechanism, diverging from a centralized 'pull' approach. This affords remote databases the autonomy to determine the frequency of data streaming and exercise control over the information to be transmitted, thereby preserving privacy. Integrated and structured data is seamlessly unified into the Patient Medical Digital Twin (PMDT), enabling AI tools to deliver personalized preventive measures and interventions. The PMDT Level employs advanced knowledge representation through abstract knowledge structures, or blueprints, to provide a comprehensive patient profile (details in Section 5).\u003c/p\u003e \u003cp\u003eThe interaction between AI, Big Data, and PMDT levels supports diverse applications, emphasizing personalized chronic disease management within a prevention-focused, outcome-driven healthcare model. For example, in the ecosystem depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, this interaction can recommend an optimal treatment plan for a 10-year-old patient hospitalized within 28 days post-CAR-T cell infusion. The system considers treatment efficacy while aiming to minimize immune-related adverse events (IR-AEs), predicting their occurrence and assessing the impact on the patient\u0026rsquo;s quality of life (QoL) as discussed in Section 3. This approach enhances the overall healthcare experience.\u003c/p\u003e \u003cp\u003eAt the Stakeholder View Level, clinicians can access harmonized PMDT data, select personalized risk models (e.g., risk of hospitalization) from a model library, and train and validate these models using available patient risk factors, as illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"5 The Patient Medical Digital Twin","content":"\u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e illustrates the PMDT proposed in this paper. The PMDT is shown to represent virtual cancer patients receiving Immunotherapy [22]. To this end the PMDT includes information about patient data and medical capability, and it also encapsulates policies and rules, safety procedures, treatment schedule information, past treatment performance data, and guidelines on how to improve a patient\u0026rsquo;s health. The concept of PMDT is essentially a patient-care-centric logical construct.\u003c/p\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e5.1 Medical Digital Twin Definition\u003c/h2\u003e \u003cp\u003eThe Digital Twin definition illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e is shown to encompass seven interconnected processable abstract knowledge types, referred to as blueprints, which encapsulate and convey medical knowledge:\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eMedical Stakeholder Blueprint\u003c/em\u003e: this type represents and stores health professional capability data regarding which stakeholder is available for treatment, e.g., physicians, nurses, social-carers, nutritionists, etc., their skills, capabilities and capacity, responsibilities, and available health facilities\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003ePatient Blueprint\u003c/em\u003e: this abstract type represents and stores patient definition data including patient profile and characteristics, medication, treatment history, patient-generated data, health status data, nutrition and lifestyle data, physiological and psychosocial data.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eDisease Blueprint\u003c/em\u003e: this type represents and stores patient disease profile data such as, disease type, characteristics, diagnosis, symptoms, adverse events, tests, QoL impact, etc. For cancer patients it includes pathological data, genetic tests, tumor marker tests, symptoms, cancer subtype, concomitant Cancer, co-morbid conditions, and so on.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eTreatment Blueprint\u003c/em\u003e: this type captures current and historical treatment information, including treatment types and medication statements.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eMedical Pathways Blueprint\u003c/em\u003e: this type represents in a graphical form and stores treatment schedule data regarding what to cure and how to improve patient QoL. It includes personalized medical plans, agreed goals and actions, medical schedule, treatment sequences and transitions, treatment flow diagrams, personnel from the Medical Stakeholder Blueprint, diagnosis, clinical results, procedures, drug treatment, and so on. This abstract type uses Business Process Modeling and Notation 2.0. BPMN, an open executable standard that uses graphic symbols to represent a process, participants, choices, and flow. Figure\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e presents the BPMN model of the Management of the Cytokine Release Syndrome (CRS) as one of the common IR-AEs based on best practices recommendations in [22] on the management of CAR T-cell therapy. Figure\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e illustrates a simplified sub-process for managing Grade 1 CRS presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eMedical Safety Blueprint\u003c/em\u003e: this type represents and stores safety definition data. It encapsulates policies and rules, safety procedures, clinical practice guidelines, privacy regulations, and data security policies and practices.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eTreatment Performance Blueprint\u003c/em\u003e: this type represents and stores patient treatment performance data such as treatment performance indicators and attribute, e.g., patient readmission rate, workforce utilization rate, timeliness of care, improvement in costs, etc.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe information encapsulated within the seven PMDT knowledge type definitions can be instantiated to portray a comprehensive representation of a patient. This instantiation aids clinicians and care teams in obtaining a holistic view of the patient's health, enabling a deeper understanding of how various contributing factors may influence health outcomes and overall QoL.\u003c/p\u003e \u003cp\u003eBlockchain technology can be used in conjunction with PMDT contained information to help health information exchanges (HIEs) by relieving security concerns. To secure information exchange, each time caregivers provide patients with a PMDT service, they update their patients\u0026rsquo; health data on a blockchain-enabled HIE. Each blockchain member has a private key, which is secure, and a public key that acts as a visible identifier. Because of these permission layers, patients can limit data access and share only the relevant parts of their medical records with their caregivers or other clinicians [23]. This represents a current ongoing extension to the work presented in this article\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e5.2 Processing the Medical Digital Twin\u003c/h2\u003e \u003cp\u003eOnce a PMDT is defined and instantiated, it will be available for processing. An operational PMDT blends and cross-correlates information from the blueprint types in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e. Simple processing can take the form of:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eViewing and querying blueprints\u003c/em\u003e (by linking and cross-correlating information from diverse PMDT blueprints). Querying is classified into:\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003e \u003cem\u003eDescriptive/retrieval querying\u003c/em\u003e: is a statistical interpretation used to analyze historical data to identify patterns and relationships, which usually answers the question of \u0026ldquo;What has happened?\u0026rdquo;. \u003cb\u003eExample-1\u003c/b\u003e: Referring to the pilot study in Section 3, \u0026ldquo;a policy maker might be interested in understanding the correlation between \u0026lsquo;Age\u0026rsquo; (demographic data) and \u0026lsquo;Annual income\u0026rsquo; (socio-economic data) of cancer patients\u0026rsquo; information distributed among the four participating hospitals in Europe\u0026rdquo;. Figure\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e presents a screenshot of the platform implementation (cf. Section 9) showing the output of this query by displaying the quantitative correlation coefficient of these two factors and a dot chart/plot of the results.\u003c/p\u003e \u003cp\u003e \u003cem\u003eAnalytical querying\u003c/em\u003e: uses homogenized data from the PMDT to predict future trends and events and forecast potential scenarios that can help drive strategic decisions. \u003cb\u003eExample-2\u003c/b\u003e: Analytical querying provides an answer to the question \u0026ldquo;What might happen in the future?\u0026rdquo;. An example of an analytical/predictive query is the predication of QoL of the 40-years-old female melanoma patient presented in Section 3. The output of this predictive query is an interactive bar chart showing the probability of each predicted adverse event for this particular patient as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eExtending blueprint definitions\u003c/em\u003e by composing multiple blueprints of the same type, typically organized into a larger, composite blueprint, for instance, by adding more disease blueprints for elderly individuals diagnosed by melanoma with type 2 diabetes. The blueprint composition operator exhibits a closure property: each operation takes one or more compatible blueprints and returns a higher-level blueprint. Compatible blueprints are n-wise combinations of the same type of blueprints (e.g., combinations of disease-to-disease or treatment-to-treatment blueprints). This allows for blueprint composition using the merge operation that interconnects a set of blueprints end-to-end by describing all of their characteristics of a personalized patient replica.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eAdding\u003c/em\u003e (e.g., adding stamped periodic Patient Generated Health Data (PGHD) from sensors) and \u003cem\u003erevising\u003c/em\u003e information contained in blueprints. For example, a diabetic melanoma patient using wearable technologies for continuous glucose monitoring (CGM) systems that continuously track glucose levels. Time stamped PGHDs are can be added to the PMDT replica if they do not already exist.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eConventional complex PMDT processing operations include:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eMonitoring patient medication adherence\u003c/em\u003e: Sensor data may be effectively used in monitoring missed doses. \u003cb\u003eExample-3\u003c/b\u003e: A monitoring patient medication adherence example could be used whena wearable insulin pumps automatically delivers a dose of insulin when it is needed.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eTracking of adverse events and drug efficacy\u003c/em\u003e: Sensor data are useful to close-track any adverse events and generate alerts for patients and caregivers. Similarly, any health improvements and drug-related benefits may be noted.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eTime point-based patient monitoring and vitals tracking\u003c/em\u003e: Remote patient monitoring provides care teams with the tools they need to remotely track the health of their patients at home measuring patient vital signs. \u003cb\u003eExample-4\u003c/b\u003e: Time point-based patient monitoring and vitals tracking allows to assign patients to specific care protocols and interventions that are tailored to condition or acuity level in order to establish better care and prevent serious adverse events.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eTime Series Analysis\u003c/em\u003e: Patient-generated data can be used for time series/trend analysis. This can be achieved with the help of various statistical analysis tools to predict and interpret results.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eConcurrent medications and drug interaction monitoring\u003c/em\u003e: \u003cb\u003eExample-5\u003c/b\u003e: Capturing sensor data and merging with concomitant medications and patient history data can be an effective way to prevent drug-related adverse events.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eMore advanced processing abilities that rely exceedingly on AI and Machine Learning support include:\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eDeveloping care plans and recommendations\u003c/em\u003e: Using the information contained in the PMDT, clinicians can improve their interactions with their patients and work jointly to develop care plans and recommendations, such as the one depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e. \u003cb\u003eExample-6\u003c/b\u003e: Referring to the pilot study in Section 3, a healthcare professional might be interested in developing a treatment/care plan for the hospitalization of a 10-year-old patient in the first 28 days after CAR-T infusion.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eRunning in silico tests, comparisons, and simulations, and study performance issues\u003c/em\u003e: A synchronized linking of PMDT type instances is a time-stamped copy of a PMDT containing data and functions specific for an individual. These can be processed using a simple language to run in silico tests, comparisons, and simulations, and to study performance issues all with the goal of generating valuable insights that mirror the current status of an individual patient, predict the patient\u0026rsquo;s behavior and future states. \u003cb\u003eExample-7\u003c/b\u003e: for the 40-years-old female melanoma patient example presented in Section 3, it may be desirable to compare the performance of the treatment after three months of receiving Pembrolizumab 200 and contrast with her historical information.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eAnswering specific disease questions\u003c/em\u003e: For example, PMDT instances can be used to answer questions such as \u0026lsquo;What is the expected QoL of a particular melanoma patient after one year of receiving immunotherapy treatment?\u0026rsquo;. Also, to predict the IR-AEs of a specific cancer patient with specific characteristics, like the analytical example presented in Section 3.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cem\u003eCreating disease assessment reports\u003c/em\u003e: \u003cb\u003eExample-8\u003c/b\u003e: For instance, the creation of a melanoma assessment report that summarizes the impact of drinking and smoking habits on the diagnosis, clinical management, and the outcomes of combining multiple cancer treatment therapies.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e \u003cp\u003eUsing the AI tools in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, the PMDT offers a risk-free playground to explore innovations and test limitless \u0026ldquo;what-if\u0026rdquo; scenarios. Predictive AI techniques can be used to analyze each instantiation of the PMDT model to understand how ongoing lifestyle changes are impacting a specific patient\u0026rsquo;s health and suggest new tweaks that could speed up those improvements and reverse the effects of long-term conditions.\u003c/p\u003e \u003c/div\u003e"},{"header":"6 Implementation and validation","content":"\u003cp\u003eThe implementation\u003csup\u003e5\u003c/sup\u003e and validation of the PMDT digital health ecosystem presented in this article (cf. Figure \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e) is a challenging yet requisite effort to ascertain the validity and soundness of the proposed approach. The following two subsections present the implementation and evaluation \u0026amp; validation efforts, respectively.\u003c/p\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003ch2\u003e6.1 Implementation\u003c/h2\u003e\n \u003cp\u003eFigure \u003cspan class=\"InternalRef\"\u003e9\u003c/span\u003e presents a standard UML component diagram showing the overall system architecture and its logical components. The identified and implemented components interact with each other via Application Programming Interfaces (APIs) using the Python programming language. As shown in Fig. \u003cspan class=\"InternalRef\"\u003e9\u003c/span\u003e, the implementation comprises of four main components: (i) Portal component, (ii) Data Homogenization component, (iii) Federated Query Processing component, and (iv) Distributed Hospitals Database System components. In the following discussion, each component is discussed by showing its interactions with other components, which is followed by highlighting the iterative validation efforts conducted in the context of the QUALITOP project.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eData Homogenization Component\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eThis component represents the backbone of the whole system. It consists of\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cem\u003ePatientMedicalDigitalTwinKB\u003c/em\u003e, which implements the detailed conceptual model of the PMDT (the detailed conceptual model of the PMDT is outside the scope of this article) in the Ontology Web Language\u003csup\u003e6\u003c/sup\u003e (OWL) standard using prot\u0026eacute;g\u0026eacute;\u003csup\u003e7\u003c/sup\u003e framework. The OWL implementation provides semantic support to the PMDT models, which is a vital requirement to surmount not only the syntactic and schematic data heterogeneity problem, but also the semantic heterogeneity of diverse medical data sources [24]. This brings intelligence to healthcare as discussed in Section 4.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cem\u003eMetaDataManagement\u003c/em\u003e, which requests and imports meta-data from different involved HospitalDataBaseSystem(s). Such technical meta-data includes information about tables, columns, indexes, constraints, and any other relevant details in the disparate databases and is associated with the relevant blueprint entities in the unifying knowledge model so that they can be discovered in their respective databases. It also requests the \u0026ldquo;\u003cem\u003eMetaDataMapping\u003c/em\u003e\u0026rdquo; component to carry out the meta-data mapping of the local databases against the PMDT schema, which acts as a global view over the local database systems (cf. Figure \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e and Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e). This leads to simple query processing strategies [24]. The cross-schema meta-data mapping is then stored and maintained by the \u0026ldquo;\u003cem\u003eMetaDataManagement\u003c/em\u003e\u0026rdquo; component.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003ePortal Component\u003c/strong\u003e: The PMDT digital health ecosystem is designed and implemented as a cross, private platform using Humhub\u003csup\u003e8\u003c/sup\u003e and by using the Python programming language. Thus, the platform is built to give intended users the tools to make communication and collaboration easy and successful. The portal component incorporates and implements a wide range of analytical (AI-powered) queries as described in Section 5. This component comprises of three sub-components: (i) \u003cem\u003ePortalInterface\u003c/em\u003e, which includes all the GUIs between the user and the system to submit descriptive/analytical queries and display and manipulate the results. For example, the predication of QoL analytical query example presented in Section 3, named as query\u003csub\u003e1\u003c/sub\u003e thereafter, (ii) \u003cem\u003eQueryTranslator\u003c/em\u003e: once query\u003csub\u003e1\u003c/sub\u003e is submitted via the GUI as shown in Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e, this component will then translate it into a SPARQL\u003csup\u003e9\u003c/sup\u003e query, which is the standard query language and protocol of OWL (the PMDT implementation), and (iii) \u003cem\u003eVisualization component\u003c/em\u003e: once the \u0026ldquo;\u003cem\u003eFederatedQueryProcessing\u003c/em\u003e\u0026rdquo; component processes the query as demonstrated below, the result will be visualized and presented on the portal\u0026rsquo;s integrated dashboards using appropriate graphs, charts, tables and advanced visualization capabilities (cf. Figure \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eHybrid Federated Data Mesh Query Processing Component\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003ethe \u0026ldquo;\u003cem\u003eQueryDecomposition\u003c/em\u003e\u0026rdquo; sub-component receives the SPQRQL representation of query\u003csub\u003e1\u003c/sub\u003e from the \u0026ldquo;\u003cem\u003eQueryTranslator\u003c/em\u003e\u0026rdquo; of the \u0026ldquo;\u003cem\u003ePortal\u003c/em\u003e\u0026rdquo;. the \u0026ldquo;\u003cem\u003eQueryDecomposition\u003c/em\u003e\u0026rdquo; sub-component then decomposes or replicates the query into a number of sub-queries, corresponding to the physical location of the queried data. That\u0026rsquo;s, for example Hospital-1 (e.g., in France), Hospital-2 (e.g., in Spain),\u0026hellip;, Hospital-n. This is done by consulting the \u0026ldquo;\u003cem\u003eDataHomogenization\u0026rdquo;\u003c/em\u003e component for metadata mapping. For example, \u003cem\u003equery\u003c/em\u003e\u003csub\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sub\u003e will be replicated by the the \u0026ldquo;\u003cem\u003eQueryDecomposition\u003c/em\u003e\u0026rdquo; sub-component into \u003cem\u003esub-query\u003c/em\u003e\u003csub\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003esub-query\u003c/em\u003e\u003csub\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sub\u003e, assuming that Melanoma data exists in Hospotal-1 and Hospital-2.\u003c/p\u003e\n \u003cp\u003eThe \u0026ldquo;\u003cem\u003eQueryTranslator\u003c/em\u003e\u0026rdquo; sub-component then translates each sub-query into the query languages accepted by the local database system, e.g., Structured Query Language (SQL). The \u0026ldquo;\u003cem\u003eRoutingComponent\u003c/em\u003e\u0026rdquo; is then responsible for requesting local database systems of hospitals to process each respective SQL sub-query locally, and receives the sub-queries results. This is followed by the \u0026ldquo;\u003cem\u003eResultAggregation\u003c/em\u003e\u0026rdquo; sub-component aggregating and performing any further processing on the sub-queries results, and then forwarding the aggregated results to the \u0026ldquo;\u003cem\u003eVisualization\u003c/em\u003e\u0026rdquo; sub-component of the \u0026ldquo;\u003cem\u003ePortal\u003c/em\u003e\u0026rdquo; component. The \u0026ldquo;\u003cem\u003eResultAggregation\u003c/em\u003e\u0026rdquo; also sends the aggregated query result to the \u0026ldquo;\u003cem\u003eDataHomogenization\u003c/em\u003e\u0026rdquo; component to be saved and maintained.\u003c/p\u003e\n \u003cp\u003eThe Federated Query Processing Component is implemented in Python using the Flask framework\u003csup\u003e10\u003c/sup\u003e for web integration. Its primary function is to decompose the incoming SPARQL query into multiple sub-queries that correspond to the physical locations of the queried data.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eDistributed Hospitals Database System Components\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eThe implementation architecture in Fig. \u003cspan class=\"InternalRef\"\u003e9\u003c/span\u003e is designed as a federation of domain-oriented data products. Each data product is owned and managed by a specific medical domain. These data products expose standardized interfaces for other domain-specific data sources to access and use their data. This ensures the autonomy of the participating hospitals and preserves the privacy of medical data. Standardized APIs are implemented in Python based on JSON\u003csup\u003e11\u003c/sup\u003e. The use of APIs ensures effective data retrieval and enhances the interoperability of the system across different hospital databases.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\n \u003ch2\u003e6.2 Validation and Evaluation\u003c/h2\u003e\n \u003cp\u003ePMDTs represent the nerve center and the enabler of the digital health ecosystem presented in the article. PMDTs are based on a rigour formal foundation of Description Logic [25] through its design and implementation in OWL. Furthermore, the building blocks of the proposed PMDT-based health ecosystem, including the PMDT models, the platform, its collaboration features and analytical capabilities, were continuously evaluated, refined and improved by using it in its applications (see Section 5) and by discussing it with medical partners and patients\u0026rsquo; representatives in the QUALITOP project. These efforts embody an agile/iterative knowledge engineering and system development approach for the development of the PMDT models and the collaborative analytical platform, respectively.\u003c/p\u003e\n \u003cp\u003eBased on the analysis of the pilot study presented in Section 3, and the heavy involvement of QUALITOP medical partners in the agile/iterative analysis, design and development of the PMDT digital health ecosystem, through regular online and physical meetings spanning the lifetime of the QUALITOP project, medical partners and patients\u0026rsquo; representatives:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003evalidated the applicability and efficacy of the PMDT knowledge models conceptually and technically, which went through continuous cycles of refinement and improvement.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eiteratively identified and validated the descriptive/analytical query capabilities presented in this article.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eexperimented and validated the digital health platform as a user-friendly collaborative healthcare ecosystem. Usability is a major issue in today\u0026rsquo;s systems development, which measures how easily a user can accomplish their goals when using a service. This is achieved through research methodologies of customer satisfaction, which is attested by the medical partners.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cp\u003eMedical partners and patients\u0026rsquo; representatives appreciated the collaborative and advanced analytical capabilities of the smart digital platform, which are enabled by the formal PMDT as the nerve center of the ecosystem presented in Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e. Specifically, a team member said that \u0026ldquo;[\u0026hellip;] AI-generated solutions have immense potential when developed with humility and a clear focus on the end-users\u0026mdash;be it patients or healthcare professionals. Ensuring that these tools address well-defined healthcare challenges is critical to their success. [\u0026hellip;]. I am truly impressed by the outstanding work your team and consortium have accomplished. It\u0026rsquo;s essential that you leverage this success to make the tool accessible to both patients and doctors. [\u0026hellip;]\u0026rdquo;.\u003c/p\u003e\n \u003cp\u003eHowever, the following points were raised to further improve the developed solutions. Some are ongoing work:\u003c/p\u003e\n \u003cp\u003e\u003cem\u003ePiloting the platform with real patients in a controlled environment\u003c/em\u003e to gather feedback and make necessary adjustments: discussions started with the participating hospitals in Europe to initiate this experiment.\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eEnsuring that Data Protection Officer (DPO) and regulatory experts at participating medical institutions/hospitals assess the legal framework\u003c/em\u003e applicable to the project before its launch. Compliance with privacy laws, such as GDPR, is essential for building patient trust. Cited from one of the patients\u0026rsquo; representatives: \u0026ldquo;[\u0026hellip;] Patients care deeply about whether an ethics committee has reviewed the project or if a privacy expert has confirmed its GDPR compliance [\u0026hellip;]\u0026rdquo;: communications have started with participating hospitals in this regard.\u003c/p\u003e\n \u003cp\u003eMoreover, as discussed in this article, the proposed PMDT ecosystem is privacy-preserving by nature, as the incorporated analytical capabilities illustrated in Section \u0026lrm;5.2 are implemented following a federated data analytics approach [26]. Furthermore, the platform incorporates security and privacy measures at different levels including: (i) a multi-factor authentication method using Time-based one-time passwords or email verification, as well as manually verifying the identity and the role of the person registering to the platform, which is necessary in such a highly regulated domain to ensure the identity and role of the person signing up to the platform (ii) role-based access Control, and (iii) and input validation.\u003c/p\u003e\n \u003cp\u003eAdditionally, a GDPR notice is incorporated in all the GUIs of the platform. Figure \u003cspan class=\"InternalRef\"\u003e11\u003c/span\u003e shows a screenshot of the GDPR privacy notice in a \u0026ldquo;Melanoma Patients\u0026rdquo; support group page, a collaborative feature supported by the platform. Patient-to-patient (P2P) engagement shown in Fig. \u003cspan class=\"InternalRef\"\u003e11\u003c/span\u003e encourages controlled interaction among patients, fostering a supportive community that promotes sharing of experience and emotional support. To control P2P interactions, \u0026ldquo;Private chats\u0026rdquo; is disabled in this collaboration mode, furthermore, Natural Language Processing (NLP) techniques will be employed to ensure that the interactions between patients is only for sharing experiences and for emotional support, and no medical advice are given in such communications (details of the collaborative capabilities of the platform is outside the scope of this article)\u003c/p\u003e\n \u003cp\u003e[5] \u0026nbsp;Video demonstrations of the platform and its analytical module are available at: https://drive.google.com/drive/folders/1BmFfh2C1yEB5PJeCX0sdgxDdpydsEgKg?usp=sharing\u003c/p\u003e\n \u003cp\u003e[6] OWL: www.w3.org/OWL/\u003c/p\u003e\n \u003cp\u003e[7] https://protege.stanford.edu/\u003c/p\u003e\n \u003cp\u003e[8] HumHub: https://www.humhub.com/en\u003c/p\u003e\n \u003cp\u003e[9] SPARQL 1.1 Query Language: https://www.w3.org/TR/sparql11-query/\u003c/p\u003e\n \u003cp\u003e[10] Flask framework: https://flask.palletsprojects.com/\u003c/p\u003e\n \u003cp\u003e[11] JavaScript Object Notation: https://www.json.org/json-en.html\u003c/p\u003e\n\u003c/div\u003e"},{"header":"7 Conclusions and Future Work","content":"\u003cp\u003eThis paper proposed a DT as a full personalized in silico replica of a patient. This is a virtual model (data plus algorithms) with special features not found in traditional models and simulations, one that dynamically pairs the physical (patient and treatment) and digital worlds, and AI in order to support improved collaboration between patients and healthcare professionals to achieve disease prevention goals and understanding of when one is in danger of developing a disease or their condition deteriorates. By making highly resolved digital representations of individuals available, the DT concept enables precise personalized health protection and medical care. The PMDT is fully implemented in OWL as part of a collaborative platform and was continuously validated by medical partners and patients\u0026rsquo; representatives to ensure it applicability, efficacy and usability.\u003c/p\u003e \u003cp\u003e \u003cem\u003eIt is crucial to work against cancer inequalities\u003c/em\u003e. Tools have to be produced for the benefit of the patients, equally accessible for all, no matter the country of residence, the economical and social status of the patient, the beliefs, the health care unit he/she received treatment: theoretically the developed technological solutions presented in this article satisfy this requirement, as the PMDT, the platform, its collaborative features, and analytical capabilities are extensible in nature, because they are built on rigor formal models of the PMDTs and by following a structured agile system development approach, which enforces the loose coupling design principle as well as the open-closed-principle (software systems must be open for extensions, closed for modifications).\u003c/p\u003e \u003cp\u003eFuture work directions include: (i) the utilization of Blockchain technology with PMDT containing information to help health information exchanges (HIEs) by relieving security concerns, (ii) the incorporation of virtual and augmented reality technologies, and (iii) experimenting the proposed approach with colorectal cancer screening as part of the ONCOSCREEN\u003csup\u003e12\u003c/sup\u003e project.\u003c/p\u003e\n\u003cp\u003e[12] ONCOSCREEN project: https://cordis.europa.eu/project/id/101097036\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding\u0026nbsp;\u003c/strong\u003eThis research is partially funded by the EC Horizon 2020 project QUALITOP, under contract number H2020 - SC1-DTH-01-2019 \u0026ndash; 875171.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of Data and Materials\u0026nbsp;\u003c/strong\u003eQUALITOP cohort study has been registered as an observational study at www.clinicaltrials.gov. Trial registration number NCT05626764.\u003c/p\u003e\n\u003cp\u003eStatements and Declarations\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical Approval\u003c/strong\u003e Not applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u0026nbsp;\u003c/strong\u003eThe authors declare no financial or non-financial competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe wish to thank Hospital Clinic Barcelona (IDIBAPS), Hospices Civils de Lyon (HCL), University Medical Center Groningen \u0026nbsp;(UMCG), and Instituto Portugu\u0026ecirc;s de Oncologia, Lisboa (IPOL) for providing the essentials of the pilot study in this paper, and for the continuous evaluation and validation of the reported results. Special thanks goes to Dr Roxana Albu, Chief Scientific Officer, Association of European Cancer Leagues (ECL), and Dr Menia Koukougianni, Fellow of the European Patients\u0026apos; Academy, for their insightful feedback in evaluating and validating the work presented in this paper.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eZahid A, Poulsen JK, Sharma R, Wingreen SC (2021) A systematic review of emerging information technologies for sustainable data-centric health-care. Int J Med Informatics 149:104420. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ijmedinf.2021.104420\u003c/span\u003e\u003cspan address=\"10.1016/j.ijmedinf.2021.104420\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Health Organization. Noncommunicable diseases (2020) April \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases\u003c/span\u003e\u003cspan address=\"https://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJain P, Agarwal A, Behara RS, Baechle C (2019) HPCC based framework for COPD readmission risk analysis. J Big Data 6:1\u0026ndash;13\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLopez-Perez L, Hern\u0026aacute;ndez L, Ottaviano M, Martinelli E, Poli T, Licitra L et al BD2Decide: Big Data and Models for Personalized Head and Neck Cancer Decision Support. 2019 IEEE 32nd International Symposium on Computer-Based Medical Systems (CBMS)2019. pp. 67\u0026thinsp;\u0026ndash;\u0026thinsp;8\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAkbar W, Wu WP, Faheem M, Saleem S, Javed A, Saleem MA Predictive Analytics Model Based on Multiclass Classification for Asthma Severity by Using Random Forest Algorithm. 2020 International Conference on Electrical, Communication, and Computer Engineering (ICECCE)2020. pp. 1\u0026ndash;4\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJing S, Hou K, Yan J, Ho Z-P (2020) An application of big data mining and fuzzy prediction of total knee replacement surgery patients attributes --Fuzzy TKR patients attributes prediction. J NonLinear Convex Anal 21(8):1689\u0026ndash;1702\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSatti FA, Ali T, Hussain J, Khan WA, Khattak AM, Lee S (2020) Ubiquitous Health Profile (UHPr): a big data curation platform for supporting health data interoperability. Computing 102(11):2409\u0026ndash;2444. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00607-020-00837-2\u003c/span\u003e\u003cspan address=\"10.1007/s00607-020-00837-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSivaparthipan CB, Muthu BA, Manogaran G, Maram B, Sundarasekar R, Krishnamoorthy S et al (2019) Innovative and efficient method of robotics for helping the Parkinson's disease patient using IoT in big data analytics. Trans Emerg Telecommunications Technol. ;31(12)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAli F, El-Sappagh S, Islam SMR, Ali A, Attique M, Imran M et al (2021) An intelligent healthcare monitoring framework using wearable sensors and social networking data. Future Gener Comput Syst 114:23\u0026ndash;43\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSun K, Yang D (2021) Microprocess Microsyst 81:103793. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.micpro.2020.103793\u003c/span\u003e\u003cspan address=\"10.1016/j.micpro.2020.103793\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Human health big data evaluation based on FPGA processor and big data decision algorithm\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChakshu NK, Carson J, Sazonov I, Nithiarasu P (2019) A semi-active human digital twin model for detecting severity of carotid stenoses from head vibration-A coupled computational mechanics and computer vision method. Int J Numer Methods Biomed Eng 35(5):e3180. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/cnm.3180\u003c/span\u003e\u003cspan address=\"10.1002/cnm.3180\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKovatchev B (2019) A Century of Diabetes Technology: Signals, Models, and Artificial Pancreas Control. Trends Endocrinol Metabolism 30(7):432\u0026ndash;444. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.tem.2019.04.008\u003c/span\u003e\u003cspan address=\"10.1016/j.tem.2019.04.008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrande Gutierrez N, Mathew M, McCrindle BW, Tran JS, Kahn AM, Burns JC et al (2019) Hemodynamic variables in aneurysms are associated with thrombotic risk in children with Kawasaki disease. Int J Cardiol 281:15\u0026ndash;21. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ijcard.2019.01.092\u003c/span\u003e\u003cspan address=\"10.1016/j.ijcard.2019.01.092\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang Y, Fu T, Xu Y, Ma Z, Xu H, Du B et al (2024) TWIN-GPT: Digital Twins for Clinical Trials via Large Language Model. ACM Trans Multimedia Comput Commun Appl. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1145/3674838\u003c/span\u003e\u003cspan address=\"10.1145/3674838\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSmarter devices (2020) better patient care. MIT Technology Review InsightsJuly\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMazumder O, Roy D, Bhattacharya S, Sinha A, Pal A Synthetic PPG generation from haemodynamic model with baroreflex autoregulation: a Digital twin of cardiovascular system. 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC)2019. pp. 5024-9\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAllen A, Siefkas A, Pellegrini E, Burdick H, Barnes G, Calvert J et al (2021) A Digital Twins Machine Learning Model for Forecasting Disease Progression in Stroke Patients. Appl Sci 11(12):5576\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCorral-Acero J, Margara F, Marciniak M, Rodero C, Loncaric F, Feng Y et al (2020) The 'Digital Twin' to enable the vision of precision cardiology. Eur Heart J 41(48):4556\u0026ndash;4564. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/eurheartj/ehaa159\u003c/span\u003e\u003cspan address=\"10.1093/eurheartj/ehaa159\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eŁukaniszyn M, Majka Ł, Grochowicz B, Mikołajewski D, Kawala-Sterniuk A (2024) Digital Twins Generated by Artificial Intelligence in Personalized Healthcare. Appl Sci 14(20):9404\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVinke PC, Combalia M, de Bock GH, Leyrat C, Spanjaart AM, Dalle S et al (2023) Monitoring multidimensional aspects of quality of life after cancer immunotherapy: protocol for the international multicentre, observational QUALITOP cohort study. BMJ open 13(4):e069090. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bmjopen-2022-069090\u003c/span\u003e\u003cspan address=\"10.1136/bmjopen-2022-069090\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRahman A, Hossain MS, Muhammad G, Kundu D, Debnath T, Rahman M et al (2022) Federated learning-based AI approaches in smart healthcare: concepts, taxonomies, challenges and open issues. Cluster Comput 26(4):2271\u0026ndash;2311. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s10586-022-03658-4\u003c/span\u003e\u003cspan address=\"10.1007/s10586-022-03658-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHayden PJ, Roddie C, Bader P, Basak GW, Bonig H, Bonini C et al (2022) Management of adults and children receiving CAR T-cell therapy: 2021 best practice recommendations of the European Society for Blood and Marrow Transplantation (EBMT) and the Joint Accreditation Committee of ISCT and EBMT (JACIE) and the European Haematology Association (EHA). Annals oncology: official J Eur Soc Med Oncol 33(3):259\u0026ndash;275. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.annonc.2021.12.003\u003c/span\u003e\u003cspan address=\"10.1016/j.annonc.2021.12.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGordon R, Perlman M (2018) Shukla. The hospital of the future: How digital technologies can change hospitals globally. Deloitte Center for Health Solutions\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCruz IF, Xiao H (2005) The role of ontologies in data integration. Int J Eng Intell Syst Electr Eng Commun 13:245\u0026ndash;252\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarkus K, Frantisek S, Ian HA (2012) Description Logic Primer. eprint arXiv:1201.4089. Cornell University\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAledhari M, Razzak R, Parizi RM, Saeed F (2020) Federated Learning: A Survey on Enabling Technologies, Protocols, and Applications. IEEE Access 8:140699\u0026ndash;140725. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1109/ACCESS.2020.3013541\u003c/span\u003e\u003cspan address=\"10.1109/ACCESS.2020.3013541\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Healthcare Data Integration, Interoperability, Medical Digital Twin, Knowledge Representation and Processing, Federated Analytics, Medical Intelligence","lastPublishedDoi":"10.21203/rs.3.rs-5826330/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5826330/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis work introduces the concept of Patient Medical Digital Twins (PMDTs) to simulate treatment outcomes, optimize drug dosages, and deliver personalized chronic care. The PMDT model, supported by an interconnected ecosystem, is validated iteratively by medical institutions to ensure its efficacy and applicability.\u003c/p\u003e \u003cp\u003eAt its core, the PMDT leverages expressive knowledge structures to capture a patient\u0026rsquo;s psychosomatic, cognitive, biometric, and genetic data, creating a comprehensive personal digital footprint. This enables medical professionals to run simulations predicting health issues over time and to proactively implement personalized preventive interventions.\u003c/p\u003e \u003cp\u003eThe PMDT ecosystem integrates big data analytics, continuous monitoring, cognitive simulation, and AI technologies. By connecting stakeholders across the care continuum, it provides deeper insights into a patient\u0026rsquo;s medical history and supports informed, shared decision-making. Validated in a pilot study through an EU-funded healthcare initiative, the PMDT demonstrates its transformative potential at the intersection of Big Data and AI, positioning itself as a critical tool for advancing personalized preventive care.\u003c/p\u003e","manuscriptTitle":"Medical Digital Twins for Personalized Chronic Care","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-01-16 08:13:23","doi":"10.21203/rs.3.rs-5826330/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"06247ddb-0649-42ef-91ea-39d894524d3c","owner":[],"postedDate":"January 16th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-01-20T23:53:10+00:00","versionOfRecord":[],"versionCreatedAt":"2025-01-16 08:13:23","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5826330","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5826330","identity":"rs-5826330","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}