Ebook: Envisioning the Future of Health Informatics and Digital Health

This volume contains the accepted papers of the ICIMTH (International Conference on Informatics, Management, and Technology in Healthcare) for the year 2024, and presents to the community of Biomedical and Health Informatics the scientific outcomes of the ICIMTH 2024 Conference, which was held on 13-15 December 2024 as a virtual event.

The event was characterized as a Anniversary Event to commemorate the meeting of the conference founders, the late Professor Kyriakos Kioulafas and Professor John Mantas, to agree to initiate a conference which was based on knowledge and experience gained from the joint Inter-University Master’s programme on Healthcare Management and Health Informatics organised by the National and Kapodistrian University of Athens. The meeting was held in December of 2001, and after a trial period, the first organised conference was held as an international event wishing to exchange ideas and experiences with international colleagues in early July of 2003 on the island of Samos in Greece. This series of conferences after the passing away of Professor Kioulafas in 2011 continued to be held in Athens. So, the 22nd conference commemorates this founding event. The organisers are thankful and grateful to the participants of this virtual event. Of the 160 submissions, 99 full papers and 11 short communications papers were accepted for publication. An invited panel was held to review the events of the last 22 years of the ICIMTH conferences. The panelists were Professor Reinold Haux, former president of IAHSI and IMIA; Professor Kaija Saranto, former president of IAHSI; Professor George Mihalas, former president of EFMI; Professor John Mantas, former president of EFMI; Professor Lacramioara Stoicu-Tivadar, former EFMI president; and Professor Panos Bamidis, HL7 Hellas chair.

It should be noted that the Proceedings titled “Envisioning the Future of Health Informatics and Digital Health” will be published with Open Access and submitted (for evaluation) for indexing by MEDLINE/Pubmed and Scopus. The expected date of publication is early April 2025.

The fields of Biomedical and Health Informatics and Digital Health are studied in this conference at a very broad perspective, from cells to populations, including several technologies such as Imaging, Sensors, and Biomedical Equipment and Management and Organisational and ethical aspects. Essentially, Data Science, Informatics, and Technology and mainly Artificial Intelligence inspire health professionals and informaticians to improve healthcare for the benefit of patients.

The Editors would like to thank the Members of the Scientific Programme Committee, the Organising Committee, and all Reviewers, who have performed a very professional, thorough, and objective refereeing of the scientific work to achieve a high- quality publishing achievement for a successful scientific event.

Athens, 08.02.2025

The Editors,

John Mantas, Arie Hasman, Emmanouil Zoulias, Konstantinos Karitis, Parisis Gallos, Marianna Diomidous, Spyridon Zogas, Martha Charalampidou.

We compared emotional valence scores as determined via machine vs human ratings from a survey conducted from April to May 2024 on perceived attitudes on the use of artificial intelligence (AI) for African American family caregivers of persons with Alzheimer’s disease and related dementias (ADRD) (N=627). The participants answered risks, benefits and possible solutions qualitatively on the open-ended questions on ten AI use cases, followed by a rating of each. Then, we applied three machine learning algorithms to detect emotional valence scores from the text data and compared their mean to the human ratings. The mean emotional valence scores from text data via natural language processing (NLP) were negative regardless of algorithms (AFINN: -1.61 ± 2.76, Bing: -1.40 ± 1.52, and Syuzhet: -0.67 ± 1.14), while the mean score of human ratings was positive (2.30 ± 1.48, p=0.0001). Our findings have implications for the practice of survey design using self-rated instruments and open-ended questions in an NLP era.

This study investigated healthcare utilization patterns prior to prostate cancer diagnoses, aiming to develop machine learning models for early prediction of cancer diagnosis. Data from the All of Us Research Program was used, focusing on adult patients diagnosed with prostate cancer between 2010 and 2019. Key variables were derived from procedure, measurements, and condition records, including PSA values, comorbidity index, and symptoms. Multiple machine learning models were tested to predict prostate cancer 3, 6, 9, and 12 months ahead of time. The dataset included 1,276 cancer patients and 1,232 non-cancer patients. The XGBoost model performed best at 3 months, achieving an accuracy and F1 score of 0.73 and an AUC of 0.82. At 6 months, the model had an accuracy and F1 score of 0.71 and an AUC of 0.78. Performance declined with longer prediction windows. PSA values were consistently the most important predictor across all timeframes, along with other factors like triglyceride and creatinine levels.

Introduction:

This study developed and validated a Bayesian method which detects pairs of clinical events likely to be temporally ordered.

Methods:

Association mining rules between medical procedures were extracted into a research-generated database and, for each pair of procedures A,B, the conditional probability P(A|B), its inverse, and the difference from its inverse (ConfDiff) were calculated. The study hypothesized that the higher the ConfDiff is, the more likely it is for A and B to be temporally ordered. The actual calendar date of each medical procedure served as ground truth.

Results:

ConfDiff is the strongest predictor of %Tseq (r=0.278), followed by P(B|A) (r=0.129). This association continued to be present after controlling for the confidence, leverage and conviction metrics.

Conclusion:

Findings substantiate the assumption that, in a structured process-based domain (e.g., clinical care) if an attribute is strongly associated with another one, but not the other way around, this could imply temporality.

Artificial Intelligence (AI) holds great promise for healthcare, promising improved patient outcomes and streamlining processes. Nevertheless, this transformational journey comes with numerous potential pitfalls that warrant attention. This comprehensive review explores some key challenges involved with integrating AI into medicine. First and foremost is the risk of over-reliance on AI systems. Users often rely on recommendations provided by AI to follow without question, potentially causing automation bias. Human oversight is essential to avoid mistakes and patient harm; failure to provide such oversight could have serious repercussions that necessitate having someone in control at all times - emphasizing the necessity for having a human-in-the-loop approach. Ethical considerations must always come first when developing AI systems, with privacy, informed consent, and data protection as non-negotiable obligations for patients and organizations. Transparency and accountability within AI systems are necessary to quickly identify biases or errors to enable AI development with integrity that mitigates bias, ensures fairness, and maintains transparency. Ethical AI development involves ongoing efforts made with great diligence by developers to mitigate any bias, ensure fairness, and maintain transparency. These principles form the bedrock upon which ethical development depends. Collaboration between healthcare providers and AI developers is of utmost importance for patient safety and well-being; healthcare providers must protect patient data while developers must ensure AI systems adhere to legal and ethical requirements. AI and healthcare present significant challenges. Ethical frameworks, bias mitigation techniques, and transparency measures must all be pursued to advance AI’s role within healthcare delivery systems. We can unleash AI’s full potential by overcoming such hurdles while upholding patient safety, ethics, and quality care as the cornerstones of healthcare innovation.

Medical and paramedical training is still a major issue in Africa, given the growing needs that contrast with the shortage of teachers, hospitals and educational facilities. The aim of the present work is to design an artificial intelligence that enables a virtual patient to hold a conversation with a doctor or a student. To achieve this, we first carried out a literature review on the various approaches to designing artificial intelligence of this type. Next, we selected the tools needed to develop the intelligent model. Finally, we developed and trained our intelligent model to hold a conversation with a healthcare professional during a medical consultation.

The adaptation of a breast cancer detection platform based on artificial intelligence, designed for use on Android devices, is an initiative driven by the particular challenges faced in Africa, where access to computers is often limited due to their high cost and limited availability, a significant issue in Burkina Faso. It is especially crucial to find tailored and more efficient solutions for healthcare professionals in such environments. This mobile adaptation aims to make this advanced technology more accessible to healthcare professionals across the country, with mobile devices being far more common and accessible, with around 86% coverage in Burkina Faso. Our goal is to simplify the work of pathologists by enabling them to benefit from the advantages of AI for early and accurate breast cancer detection, directly from mobile devices, without requiring expensive infrastructure.

This study explores the readiness of Dutch general practitioners (GPs) to adopt artificial intelligence (AI) technologies in primary healthcare (PHC) using an expanded Technology Acceptance Model (TAM). The model assesses perceived usefulness (PU), perceived ease of use (PEU), and perceived societal impact (PSI) of AI. A self-administered online questionnaire gathered responses from 224 GPs regarding their demographics, AI knowledge, and perceptions of AI in PHC. Significant relationships were found between PU and “intention to use AI”, as well as between “intention to use” and “actual use”. Knowledge about AI was significantly correlated with PU. However, no significant relationships were found between PEU and “intention to use” and between PSI and “intention to use”. Age, gender, and years of experience had significant relation with PU, with younger age, male gender reporting higher PU. Overall, the study highlights that Dutch GPs, particularly younger GPs are ready to adopt AI-based technologies.

This short communication presents preliminary findings on the application of Large Language Models (LLMs) for sentiment analysis in educational settings. By analyzing qualitative descriptions derived from student reports, we aimed to assess students’ emotional states and attitudes towards their academic performance. The sentiment analysis provided valuable insights into student engagement and areas requiring attention. Our results indicate that LLMs can effectively process and analyze textual data, offering a more nuanced understanding of student sentiments compared to traditional coding methods. This approach highlights the potential of LLMs in enhancing educational assessments and interventions.

This short communication presents preliminary findings on the integration of Large Language Models (LLMs) and wearable technology to generate personalized recommendations aimed at enhancing student well-being and academic performance. By analyzing diverse student data profiles, including metrics from wearable devices and qualitative feedback from academic reports, we conducted sentiment analysis to assess students’ emotional states. The results indicate that LLMs can effectively process and analyze textual data, providing actionable insights into student engagement and areas needing improvement. This approach demonstrates the potential of LLMs in educational settings, offering a more nuanced understanding of student needs compared to traditional methods.

Artificial Intelligence (AI) is increasingly incorporated into medical devices, revolutionizing diagnostics, treatment planning, and patient monitoring. To ensure AI’s safe and ethical use, the European Commission published the AI Act in 2024, which places stringent obligations on AI systems, especially those classified as high-risk, such as medical devices. This paper evaluates the impact of the EU AI Act on existing regulations such as the Medical Device Regulation (MDR) and the In Vitro Diagnostic Regulation (IVDR). It explores challenges related to compliance, certification processes, and potential conflicts between the AI Act and existing medical device frameworks while providing recommendations for harmonization.

This paper introduces a novel approach for predicting symptom escalation in chemotherapy patients by leveraging Convolutional Neural Networks (CNNs). Accurate forecasting of symptom escalation is crucial in cancer care, as it enables timely interventions and enhances symptom management, ultimately improving patients’ quality of life during treatment. The analytical dataset consists of daily self-reported symptom logs from chemotherapy patients, capturing a variety of symptoms such as nausea, fatigue, and pain. However, the data was significantly imbalanced, with approximately 84% of entries showing no symptom escalation. To address this issue and enhance the model’s ability to identify symptom escalation, the data was resampled into varying interval lengths, ranging from 3 to 7 days. This resampling allows the model to detect notable changes in symptom severity over different time frames. The study’s results show that shorter intervals (3 days) delivered the best performance, achieving an accuracy of 79%, a precision of 85%, a recall of 79%, and an F1 score of 82%. As the interval length increased, both accuracy and recall declined, though precision remained relatively consistent. These findings illustrate the capability of CNN-based models to capture temporal patterns in symptom progression effectively. Incorporating such predictive models into digital health platforms could empower healthcare providers to offer more personalized, proactive care, allowing for earlier interventions that may reduce symptom severity and improve adherence to treatment

The European Union’s Artificial Intelligence Act (EU AI Act) proposes a comprehensive regulatory framework for rtificial ntelligence (AI), emphasizing transparency, accountability, and ethical considerations, particularly in high-risk sectors like healthcare. This study assesses the readiness of European healthcare institutions for compliance with the EU AI Act. Using a systematic analysis of news articles and publications, we examine the current discourse on AI governance, compliance challenges, and implementation strategies within the European healthcare context. Our study findings indicate a big gap between AI adoption and compliance preparedness; hence, healthcare institutions should have increased awareness and strategic planning.

Introduction:

Gut microbiota (GM) is implicated in the remnant liver regeneration (LR) after partial hepatectomy (PH) and affects outcomes. Our study shifts the algorithmic computational modeling from the classical knowledge of (LR) to that of (GM) implication, integrating Artificial Intelligence/Machine Learning (AI/ML) for risk/benefit analysis to optimize outcomes.

Methods:

The best model predicting postoperative liver volume (LR) has been developed upon the classic biological knowledge. This phenomenological model predicts, whether liver size would recover or remain irreversibly reduced and it is not perfect.

Results:

Focusing on the impact of (GM) on (LR) after PH and the current articles upon (GM) and its impact on the change of the medical dogma integrated (GM), (AI/ML) to provide new predictive and therapeutics capabilities after PH.

Conclusion/Discussion:

Personalized and precise preoperative preparation for PH can optimize anatomic PH, pre-operative planning and outcomes upon AI/ML risk/benefit analysis integrating the impact and measurements of (GM).

We used data mining to predict the attitudes of 527 caregivers towards the pros and cons of using robotics and artificial intelligence (AI) for dementia care in African American families, with a focus on family-level factors. African American family caregivers would prefer using AI home attendant for caregiving, even though there are associated costs, and see the benefits of using AI robots to improve family dynamics, despite the need for the AI to collect sensitive data. In contrast, white family caregivers aged 25-34 are more likely to perceive the risks of using AI robots for this purpose. The proposed AI smart home system evaluates care quality and assists families in nursing home decisions. However, specific groups are hesitant to embrace its benefits. This highlights the need for in-depth research to address concerns and communicate potential advantages effectively.

The integration of Artificial Intelligence (AI) into healthcare has the potential to revolutionize patient care, diagnostics, and treatment planning. However, this integration also introduces significant challenges related to data governance, privacy, and compliance with emerging regulations. The European Union’s (EU) AI Act proposes a comprehensive regulatory framework aimed at ensuring that AI systems are trustworthy and respect fundamental rights. This paper provides an in-depth analysis of the data governance requirements stipulated by the EU AI Act specifically within the context of healthcare AI. Furthermore, it explores strategies for compliance, examines the interplay with existing regulations such as the General Data Protection Regulation (GDPR), and addresses the ethical considerations inherent in deploying AI in healthcare settings.

Large language models (LLMs) have increasingly been used to extract critical information from unstructured clinical notes, which often include important details not captured in the structured sections of electronic health records (EHRs). This study assesses the performance of the GPT-4o LLM in extracting signs and symptoms (S&S) from clinical notes, focusing on both general and organ-specific (urological and cardiorespiratory) contexts. Clinical notes from the MTSamples corpora were manually annotated for comparison with the S&S extraction results using LLM. GPT-4o was applied to extract S&S using named entity recognition techniques. Key performance metrics—precision, recall, and F1-score—were used to evaluate and compare general and organ-specific results. The model showed high precision in general S&S extraction (78%) and achieved the highest precision for organ-specific tasks in the cardiorespiratory dataset (87%). For the urinary dataset, precision was also strong (81%), with balanced recall and F1-scores across analyses. These findings underscore GPT-4o’s effectiveness in both general and domain-specific S&S extraction but highlight the need for domain-specific tuning and optimization to further improve recall and generalizability in specialized medical contexts.

The integration of artificial intelligence (AI) into medical informatics presents significant opportunities to enhance healthcare through data-driven diagnostics, predictive analytics, and personalized therapeutic recommendations. This paper examines the role of general intelligence in improving the effectiveness and adaptability of AI systems in complex clinical environments. We explore various levels of generalization – local, broad, and extreme – highlighting their respective contributions and limitations in healthcare. Local generalization provides robust assessments based on well-defined risk factors, while broad generalization allows for nuanced patient stratification across diverse populations. Extreme generalization, however, presents the greatest challenge, requiring AI systems to adapt to entirely new contexts without prior exposure. Despite advancements, existing metrics for assessing generalization difficulty remain inadequate, necessitating the development of new evaluation methodologies.

Introduction:

Dementia is a major cause of disability among the elderly, imposing significant financial burdens on healthcare systems. Traditional care approaches contribute to rising costs, especially in high-income countries. Artificial intelligence (AI) offers potential solutions by enhancing various areas of dementia care.

Methods:

This scoping review follows the Arksey and O’Malley framework, identifying studies from PubMed, Scopus, and Web of Science that examine AI applications in dementia care with economic impacts. Eight studies met criteria, focusing on cost reduction in diagnosis, monitoring, personalized care, and resource management.

Results:

AI reduces healthcare costs by enabling timely interventions, optimizing resources, and tailoring care. Technologies, including machine learning for diagnosis and wearable devices for monitoring, showed significant cost-saving potential.

Discussion:

AI holds promise for reducing dementia care costs, though challenges like data privacy, bias, and system integration remain. Addressing these and further research is essential to maximize AI’s impact on dementia care.

Accurate extraction of patient symptoms and signs from clinical notes is essential for effective diagnosis, treatment planning, and research. In this study, we evaluate the capability of GPT-4, specifically GPT-4o, in extracting symptoms and signs from nursing notes within the MIMIC-III dataset. We experimented with two temperature settings (1 and 0.3) to explore the impact of model diversity and consistency on extraction accuracy. Performance metrics include precision, specificity, recall, and F1-score. The results show that a higher temperature (1) led to more creative and varied outputs, with a mean precision of 79% and specificity of 96%, but also exhibited variability, with a minimum precision of 24%. Conversely, at a lower temperature (0.3), precision was more conservative but dropped significantly, with a mean precision of 45% and minimum of 0%. High recall and specificity at optimal temperature setting indicates that GPT-4 holds promise as an assistive tool in clinical practice for symptom and sign extraction tasks.

We applied machine learning techniques to build models that predict perceived risks and benefits of using artificial intelligence (AI) algorithms to recruit African American informal caregivers for clinical trials and general health disparity research via social media platforms. In a U.S. sample of 572 family caregivers of a person with Alzheimer’s disease and related dementias (ADRD), our application of the J48 algorithm (C4.5) revealed an interesting trend. African American family members of a person with ADRD were more likely to see the benefits of using AI on social media to ease the burden of recruitment, regardless of age, ethnicity, gender, and level of education. However, white family caregivers, particularly those aged 25-34 with graduate degrees, were more cautious and prone to perceive risks of using AI on social media for recruitment in research. This caution underscores the need for further research and understanding in this area.

Chronic lymphocytic leukemia (CLL) exhibits a heterogeneous clinical course. Prognostic markers that impact patient outcomes have been identified, including MYC gene abnormalities. This study investigates machine learning (ML) models for predicting survival in CLL, comparing the performance of Random Survival Forest (RSF), Decision Tree (DT), and Cox proportional hazards models across two cohorts: MYC-positive patients and a general CLL population. Three time-to-event outcomes were assessed: 10-year from diagnosis, 10-year from cytogenetic assessment, and time to first treatment. Model performance was evaluated using the C-index and AUC, revealing that RSF and DT models outperformed Cox models in predictive accuracy. Permutation importance highlighted key predictive variables; however, RSF and DT models pose interpretability challenges compared to Cox models, which offer clear hazard ratios. Additionally, an interactive application is available via Streamlit, and the source code is open access on GitHub. Despite limitations in dataset size and external validity, ML models show promise for personalized survival predictions in CLL, especially for MYC-positive cases, underscoring the potential for further model refinement to enhance clinical usability.

Humanoid robots are increasingly being used in a number of domains. This paper focuses on reviewing the use of the Pepper humanoid robot in healthcare. This robot has begun to be used in a range of settings and combines speech recognition with artificial intelligence to create meaningful interactions with users. In this paper we describe a scoping review conducted to assess the type and range of applications Pepper has been used for. The focus of the review was to determine what the uses of Pepper have been, how it has impacted healthcare and what the challenges and limitations are of using Pepper in healthcare. The results of the review indicate that Pepper has successfully been applied to an increasing range of areas which include its use in dementia care, neurodevelopmental disorders, chronic illness education, caregiver shortages as well as for cognitive stimulation therapy.

Evaluating the blood smear test images remains the main route of detecting the type of leukaemia, accurate diagnosis is fundamental in providing effective treatment. The changes in the structure of the white blood cells present different morphological characteristics translated into extractable features. This paper explores techniques for manipulating a reduced dataset to increase the classification with CNN (Convolutional neural Network) and feature extraction. Extracting ROI (Regions of Interest) divides the leukaemia images into points of interest respective white blood cells, expanding the dataset an important factor for CNN’s performance. Segmenting the initial dataset into ROI through computation after applying Otsu thresholding results in a new dataset of images. The two datasets are analysed, feature extraction performs better on the initial dataset while CNN’s accuracy is higher for ROI images. Further steps will divide the images into filtered regions of interest where more specific characteristics are extracted to increase the accuracy.

Objectifying the quality of microsurgical technique is both crucial and challenging. The aim of this study was to evaluate whether microinstrument motion metricscan reflect the complexity of microsurgical tasks. The laboratory experiment involved 13 right-handed neurosurgeons tasked with using microsurgical scissors to cut a white thread at a spot marked by a purple dot under the microscope. Each participant completed the task under four consecutive conditions: with or without wrist stabilization on a support, both before and after muscle load. Using the promptable transformer model, we segmented microsurgical instruments from video recordings and extracted their skeletons and centers of mass. From the time series of the center of mass X and Y coordinates, we derived seven additional time series for velocity, acceleration, and the jerk along the X and Y axes, as well as the smoothness metric. We generated thirty-three statistical features for each time series using the feasts R package. These motion features were then compared pairwise across various tasks. Of the 1782 tests conducted, 164 (or 9.2%) revealed statistically significant differences in 66 motion features. Our results provide a proof-of-concept, showing that AI-derived microsurgical motion features can reflect the complexity of conditions encountered by the microsurgeon during surgery.