The application of Artificial Intelligence (AI) in healthcare is a burgeoning field with the potential to revolutionise clinical practices and patient outcomes [1,2]. Our laboratory, the Augmented Reality for Health Monitoring Laboratory (ARHeMLab) at the Università di Napoli Federico II, sits at the forefront of this exciting research landscape.
ArnhemLab explores the potential applications of artificial intelligence and augmented reality within a scholarly setting, focusing on advancements in healthcare knowledge and development of novel tools. Our research is guided by a deep awareness of the challenges and op-portunities AI presents within the healthcare domain. We recognise its complexity and are committed to conducting thorough, ethical research to uncover real-world solutions.
This research group has been involved in developing AI-powered systems for non-invasive cardiovascular risk assessment with wearable technology [3] and automated fracture detection in maxillofacial trauma patients [4]. These projects contribute to the ongoing exploration of various artificial intelligence applications in improving patient care and diagnostic procedures.
As the use of artificial intelligence in healthcare is still developing, ArnhemLab operates in an environment characterised by both unknowns and potential applications. Our projects, encompassing cognitive monitoring with EEG-based systems and AI deployment for complex disease diagnosis, represent steps towards understanding how technology can be effectively integrated into healthcare.
The following sections detail our current research activities. We focus on harnessing AI's power to push boundaries in healthcare, particularly through innovative monitoring and diagnostic technologies. This paper is structured to first introduce ARHeMLab's core research areas, highlighting our significant advancements in applying AI and machine learning within healthcare. We will then explore specific examples of how we integrate these technologies into practical healthcare solutions. Through this exploration, we aim to provide a clear overview of ARHeMLab's contribution to AIdriven healthcare advancement, offering insights into our methodologies, achievements, and future research directions.
This section outlines our core research initiatives, each addressing a distinct topic critical to the broader field of AI in healthcare. These topics encompass the development of EEG-based systems for cognitive function monitoring and the application of AI to improve gait analysis and rehabilitation, among others. Each subsection provides a brief introduction to our contributions in these areas, laying the groundwork for a more in-depth exploration of their significance, methodologies, and potential impact on transforming healthcare practices and patient outcomes. These projects represent our efforts to harness the power of machine learning for advancing medical interventions. By integrating sophisticated analytical techniques, we aim, for instance, to achieve higher precision in surgeries and customise non-invasive treatments for more personalised and effective patient care.
The research within this topic explores the application of artificial intelligence, machine learning, and electroencephalography (EEG) in healthcare. The goal is to improve diagnostic accuracy and enable real-time patient monitoring.
We focus on analysing complex health data, such as the link between nutrition and chronic conditions, and monitoring cognitive states in high-pressure medical professions. The potential benefits include personalised medicine approaches and improved performance and safety for healthcare professionals.
We are developing systems for predicting health states and assessing cognitive load. By identifying specific biomarkers and cognitive indicators, this research aims to contribute to more data-driven and responsive healthcare practices.
Here the focus is on applying machine learning techniques to refine and improve the effectiveness of medical treatments and procedures. Our current research projects explore innovative ways to leverage data-driven insights in both surgical and non-invasive therapeutic contexts. One area of focus is optimising blood perfusion quality during laparoscopic colorectal surgeries. Machine learning algorithms analyse intraoperative data to predict tissue blood flow adequacy, assisting surgeons in making real-time decisions that can directly impact surgical outcomes and patient recovery.
We are also conducting research to understand the effects of non-invasive treatments like transcranial Electrical Stimulation (tES) on brain activity. Machine learning helps identify patterns and correlations between treatment parameters and neurophysiological responses. This research aims to tailor treatments to better suit individual patient profiles and enhance therapeutic efficacy.
The increasing application of Artificial Intelligence (AI) and Machine Learning (ML) in healthcare is driving significant advancements in diagnostic accuracy and patient monitoring. This research area focuses on utilising these computational technologies to gain a deeper understanding of complex health conditions and optimise healthcare interventions.
The research initiatives under this theme exemplify a broader shift in healthcare towards data-driven, personalised medicine. These projects, by connecting computational science with clinical practice, are not simply theoretical exercises; they are laying the foundation for potential transformations in healthcare delivery. As these technologies develop and integrate more seamlessly into healthcare systems, they could usher in a new era of diagnostics and patient monitoring. This era might be characterised by increased precision and a potential focus on tailoring care to individual patient needs and contexts. Looking ahead, there is a possibility that AI and machine learning technologies could become central to improvements in healthcare delivery and patient outcomes.
One area of research within this field focuses on the relationship between dietary intake and blood glucose levels in individuals with type 1 diabetes [5,6]. Traditional models for predicting glycemic response have limitations, often using linear approaches that don't account for the complex interplay between various dietary components and individual metabolic variations. Current research explores the use of advanced machine learning algorithms, such as Random Forest and Support Vector Machines. These algorithms are trained on large datasets encompassing diverse nutritional profiles, glycemic indices, and patient-specific metabolic responses. The algorithms identify subtle correlations between these factors that may not be captured by conventional analysis (see Figure 1.
Furthermore, the research incorporates explainable Artificial Intelligence (XAI) principles to ensure the model's outputs are interpretable. This provides both patients and healthcare professionals with actionable insights into how different foods and meal timing affect glycemic control. This personalised dietary planning tool could represent a significant advancement in managing T1DM by offering a tailored approach that potentially mitigates the risk of glycemic spikes and improves long-term health outcomes.
Alongside research on nutrition, another area of focus explores the use of wearable EEG-based systems to monitor cognitive load and fatigue in neurosurgeons [7]. This research utilises high-resolution EEG caps designed to capture real-time brain activity across various frequency bands (see Figure 2).
The high-dimensional nature of the data necessitates sophisticated signal processing techniques. Algorithms perform spectral analysis, transforming EEG signals into power densities across theta, alpha, and beta frequency bands. These bands are known to be associated with different cognitive states, ranging from deep concentration to approaching fatigue.
A key aspect of this approach involves applying machine learning classifiers. These classifiers are trained on labelled datasets to distinguish between these cognitive states with high accuracy. The resulting dynamic monitoring tool aims to alert surgeons to the onset of cognitive fatigue, potentially improving surgical precision and reducing the risk of errors associated with diminished cognitive capacity.
To further refine the detail of cognitive load assessment, some projects are exploring the use of Sequential Feature Selector (SFS) algorithms. These algorithms identify the most informative EEG features that reflect the cognitive demands specific to neurosurgical tasks [8]. Unlike simpler methods, SFS takes a more nuanced approach. It iteratively evaluates the predictive power of each feature and its interaction with others, ultimately constructing a subset of features that maximises the model's performance. This meticulous selection process, combined with machine learning classifiers such as Deep Neural Networks and Gradient Boosting Machines, facilitates the development of robust models for real-time cognitive load assessment. These targeted monitoring systems offer potential benefits not only in improving surgical out- comes but also for applications in other high-pressure professions where cognitive performance is crucial [9].
Research using Explainable Artificial Intelligence (XAI) to analyse EEG features associated with critical cognitive functions, such as inhibition and working memory activation, represents a new area of investigation in cognitive neuroscience. By employing models that provide insights into how algorithms make decisions, researchers can link specific EEG patterns to cognitive processes, leading to a deeper understanding of brain function. This convergence of AI and neuroscience not only advances our understanding of cognitive health but also opens doors for developing interventions to promote cognitive resilience, potentially improving professional performance across various fields.
The research initiatives under the theme "AI and Machine Learning for Enhanced Diagnostics and Monitoring" exemplify a broader shift in healthcare towards datadriven, personalised medicine. These projects, by connecting computational science with clinical practice, are not simply theoretical exercises; they are laying the foundation for transformative healthcare solutions.
As these technologies develop and integrate more seamlessly into healthcare systems, they hold promise for ushering in a new era of diagnostics and patient monitoring. This era would be characterised by increased precision and a greater focus on tailoring care to individual patient needs and contexts. The future suggests a possibility where AI and machine learning technologies become central to improving healthcare delivery and patient outcomes.
This research area explores the application of artificial intelligence and machine learning techniques in medical interventions. This approach aims to improve the precision, efficiency, and personalisation of both surgical and non-invasive treatments. Current research projects within this domain utilise ML algorithms to investigate new possibilities in medical treatments while also setting a focus on improving patient care and safety.
A significant portion of this research is dedicated to improving outcomes in laparoscopic colorectal surgery through the machine learning-assisted assessment of blood perfusion quality [10]. Perfusion, the process of blood delivery to tissue, is a critical determinant of tissue health and recovery post-surgery. Traditional methods for assessing perfusion rely on visual inspection, which can be subjective and variable. The integration of ML offers a paradigm shift towards a more objective, datadriven approach. Technical methodologies involve the utilisation of intraoperative imaging technologies, such as fluorescence angiography, combined with advanced image processing algorithms. Machine learning models, particularly convolutional neural networks (CNNs), are trained on vast datasets comprising images labelled with perfusion outcomes. These models learn to identify features and patterns correlated with optimal and sub-optimal perfusion, such as tissue colour, brightness, and contrast changes indicative of blood flow.
We are currently engaged in a comprehensive research effort aimed at exploring and identifying potential methods to accurately predict and estimate the risk factors associated with Anastomotic Leakage following colorectal surgery. Anastomotic Leakage is a significant and serious postoperative complication, where the connection between two sections of the intestines (anastomosis) fails to heal properly, leading to the leakage of intestinal contents into the abdominal cavity. This can result in severe infection, sepsis, and in some cases, can be life-threatening.
Parallel to surgical innovations, research efforts are also concentrated on enhancing the effectiveness of non-invasive treatments like transcranial Electrical Stimulation (tES). tES has shown promise for various neurological conditions by modulating brain activity. However, the variability in patient response poses a challenge to its widespread adoption. This challenge is met with the development of ML models capable of analysing electroencephalography data to identify biomarkers predictive of treatment success. By employing supervised learning techniques, models are trained on pre-and post-treatment EEG recordings, alongside clinical outcome measures. Feature selection algorithms, such as principal component analysis (PCA) and mutual information, reduce dimensionality and isolate the most predictive features of treatment response, such as specific frequency bands or connectivity patterns between brain regions.
Advanced classification algorithms, including support vector machines and gradient boosting machines, are then utilised to classify patients based on their likelihood of benefiting from tES. This personalised approach not only enhances patient outcomes but also contributes to the understanding of the underlying mechanisms of action of tES, paving the way for optimised protocols and broader applicability [11].
ARHeMLab's research applies machine learning to improve clinical practices, focusing on EEG-based systems and ML algorithms for detecting cognitive decline. This enhances diagnostic accuracy and monitoring, especially for medical professionals in high-stress environments. Initial studies with healthy subjects performing cognitive tasks show promise for real-time cognitive state assessment during complex activities like surgery.
Our research identifies specific EEG features linked to cognitive activation levels, paving the way for preventive measures and targeted cognitive rehabilitation programs for at-risk populations. Additionally, ARHeMLab explores ML to assess blood perfusion quality during laparoscopic surgeries, leading to a novel decision-support system to increase surgical safety and efficiency.
Looking ahead, ARHeMLab's research will involve recruiting a diverse participant pool and utilizing a broader spectrum of EEG features to refine detection capabilities and broaden system applicability. We will also investigate wearable EEG systems to assess cognitive load during motor tasks, aiming for a comprehensive understanding of cognitive states in dynamic environments.
Future research will explore neural correlates of treatments like transcranial Electrical Stimulation (tES) for conditions such as Multiple Sclerosis, aiming to correlate EEG measurements with treatment outcomes and develop adaptive, personalized tES protocols. Additionally, we aim to improve the ML-based decision-support system for blood perfusion assessment in surgery by increasing resolution and automating ROI selection.
Finally, we plan to further investigate the application of machine learning (ML) in complex medical assessments. This expanded research will focus on more intricate and multifaceted evaluations, including analysing how underlying medical conditions might influence the results of standard procedures and assessments.
This work was financially supported by the Italian Ministry of Health, through the project HubLife Science -Digital Health (LSH-DH) PNC-E3-2022-23683267 -DHEAL-COM -CUP E63C22003790001, within the "National Plan for Complementary Investments -Innovative Health Ecosystem" -Unique Investment Code: PNC-E.3.