Patients’ engagement plays a crucial role in the effectiveness of robot-assisted gait training (RAGT), particularly in paediatric neurorehabilitation, where motivation and active participation strongly influence functional outcomes. Despite its clinical relevance, engagement assessment is still largely based on subjective observation, limiting objectivity and scalability. Objective modelling of engagement in children with cerebral palsy remains underexplored, especially when relying on physiological signals acquired during therapy, due to logistical, ethical, and technological challenges that constrain multimodal data collection and cohort sizes. This study proposes a novel multimodal framework to automatically estimate engagement in a paediatric cerebral palsy population undergoing RAGT, capturing clinically relevant variability across patients. A cohort of 20 paediatric patients with cerebral palsy undergoing RAGT with the Hocoma Lokomat® system was monitored across multiple therapy sessions. Heart rate variability (HRV), facial infrared thermography (IRT), and exoskeleton-derived biomechanical features were acquired to capture complementary physiological and behavioural responses related to engagement. Signals were segmented into time windows and processed to extract statistical, spectral, and complexity features. Engagement was annotated by an expert clinician using a three-class scale (not engaged, neutral, engaged). A supervised multiclass machine-learning framework was then developed to estimate engagement from multimodal data, evaluating several ensemble-based classifiers under strict training–testing separation with repeated stratified splits and class-weighted learning to tackle class imbalance. The combination of multimodal features with the Extra Trees classifier achieved the best performance, with a macro-F1 score of 0.658 ± 0.027. Class-wise analysis showed higher performance for extreme engagement levels (not engaged and engaged), with most misclassifications occurring between adjacent classes, particularly involving the neutral state. Feature selection and explainability analyses identified exoskeleton-derived torque features as the most influential predictors, with thermal and HRV features providing complementary autonomic and emotional information. This study presents a novel multimodal machine learning approach for automatic engagement estimation during RAGT in children with cerebral palsy. Beyond classification performance, it demonstrates the feasibility of objective engagement modelling in a complex clinical population by integrating physiological and robotic interaction data. The results highlight the potential of multimodal sensing to support adaptive rehabilitation systems for continuous monitoring and personalized human–robot interaction.

IntroductionAutism Spectrum Disorder (ASD) presents considerable diagnostic challenges due to its heterogeneous nature and early developmental onset. In recent years, the convergence of noninvasive neuroimaging modalities such as Electroencephalography (EEG) and Functional Near Infrared Spectroscopy (fNIRS) with machine learning (ML) and deep learning (DL) techniques has opened new avenues for uncovering objective biomarkers of ASD. EEG offers millisecond level resolution of brain electrical activity, while fNIRS tracks hemodynamic responses tied to neuronal function, making the two methods complementary. This review aims to investigate the state of the art of the applications of EEG and fNIRS to ASD patients combined with ML and DL approaches.MethodsTo this goal, Scopus and PubMed databases were searched, and following the PRISMA guidelines, 27 peer reviewed studies published between 2019 and 2024 were included in the survey.ResultsThe results showed consistent patterns across the studies, including alterations in neural oscillations and disruptions in connectivity within key brain regions related to social communication and cognition. However, a strong heterogeneity was assessed regarding probes montages, preprocessing workflows, and classification models employed, limiting the feasibility of a metanalysis.DiscussionThe results demonstrated the potential of DL and ML algorithms applied to EEG and fNIRS signals for early ASD assessment, supporting the development of personalized intervention strategies grounded in robust neurophysiological evidence.

Infrared thermography (IRT) is a non-invasive imaging able to detect irregularities in the temperature distribution of the skin. Particularly, IRT could be effective to detect anomalies related to the presence of breast cancer through the employment of artificial intelligence (AI) approaches. However, in order to effectively transfer this technological achievement in clinical practice, some explainable AI (XAI) approaches should be integrated to foster the trust of clinicians towards these approaches and to provide explanations regarding the identification process performed by the machinery. The objective of this research is to develop an automated explainable deep learning system for identifying IRT images of breast cancer (i.e., benign and malignant). Thermal images served as input for various convolutional neural networks (CNNs), including both a customized version and those based on transfer learning (i.e., VGG-16 and ResNet50 models). The customized CNN achieved around 85

Multiple sclerosis (MS) is a chronic disorder of the central nervous system characterized by progressive impairments in gait, balance, coordination, and fatigue. The Exopulse Mollii Suit (EMS) has been recently introduced as a non-invasive method for delivering peripheral surface electrical stimulation to alleviate motor dysfunction in individuals with neurological disorders. This case study evaluated the effects of a 1-Month home-based EMS intervention on functional and biomechanical outcomes in a woman with relapsing-remitting MS. Specifically, a 53-year-old woman, clinically stable and not receiving pharmacological treatment, used the EMS suit every other day for 60-minute sessions over one month. Assessments were conducted at Baseline, Post-Session (after one EMS use), after 1-Month of therapy, and at Follow-Up (one month after therapy cessation). Clinical evaluations included the 6 Min Walk Test (6MWT), 10-Meter Walk Test (10MWT), Timed Up and Go (TUG), and Modified Fatigue Impact Scale (MFIS). Additionally, instrumented balance and strength assessments were performed using the Hunova robotic platform. Progressive improvements were observed in all clinical outcomes: 6MWT distance increased by 28 meters, gait speed improved by 14%, TUG times decreased, and MFIS scores reflected a 30% reduction in fatigue. Balance robotic evaluations showed improvements in postural control, center of pressure metrics, and stabilization times. Notably, improvements were observed after a single session, became more consistent after one month, and were partially sustained at Follow-Up, but a regression toward baseline was observed for some outcomes. Hence, the EMS appears to be a promising home-based intervention for improving gait, balance, and fatigue in MS. However, further studies should be performed to validate these findings in larger cohorts.

This study introduces an integrative methodology combining satellite image segmentation and psychometric modeling to investigate how near-home environments influence psychological restoration and affective experiences. Using an ad-hoc clustering procedure on satellite imagery, we quantified environmental features (green spaces, gray areas, roofs, shadows) surrounding the home of 917 Italian university students. These objective features were then linked to self-reported perceptions of restorativeness and emotional states (pleasure, arousal, dominance). University and nature restorativeness were included as negative-control outcomes to test context-specificity; objective features were extracted only around home. Results revealed that gray spaces negatively predicted restorativeness, particularly diminishing psychological distancing ("being-away"), attentional engagement ("fascination"), and spatial openness ("scope"). Structural Equation Models (SEM) confirmed that these components significantly mediated the relationship between gray space and affective outcomes. Specifically, gray spaces indirectly reduced emotional states of pleasure and arousal through diminished restorativeness, while also exerting a positive association with emotions of dominance, possibly reflecting feelings of environmental control or adaptation in urban contexts. Our approach advances previous research by isolating the psychological pathways linking built environments to emotional well-being, and by demonstrating the value of combining environmental segmentation with latent variable modeling. The findings support the development of urban planning strategies aimed at reducing gray space exposure and enhancing restorative features in residential areas, thereby promoting emotional resilience and well-being.

Most people live in densely urbanized environments with limited access to nature. At the same time, digital technologies increasingly shape everyday life and cognition but may also induce technostress, particularly when users experience malfunction, intrusions, and reduced sense of agency. This study examined whether exposure to a natural green environment, compared with a built urban (gray) environment, modulates psychophysiological responses to technostress. Fifty-two healthy adults were assigned to either a green (park/garden-like) or gray (built urban) environment. Before and after two outdoor walking sessions in the assigned environment, participants completed a word-selection task manipulating technology-induced stress and sense of agency. High-resolution infrared thermal imaging was used as a contact-free method to monitor facial temperature dynamics, providing indirect indices of peripheral autonomic regulation during task performance. High technostress reduced task accuracy and increased perceived stress. Under the condition combining high technostress and reduced agency, perceived stress decreased following green, but not gray, exposure. Facial thermal responses, particularly in the nose-tip and perioral regions, also differed between the green and gray groups under stressful task conditions, indicating that environmental exposure modulated peripheral physiological responses associated with technostress. Heart rate variability showed no consistent effects. Our results suggest that brief exposure to natural green environments may reduce vulnerability to technostress by attenuating subjective stress and modulating peripheral physiological responses. These findings contribute to the growing literature on restorative environments, human–technology interaction, and psychophysiological resilience.

Parkinson’s disease (PD) is a progressive neurodegenerative disorder whose diagnosis still relies mainly on clinical observation and rating scales, which are limited by subjectivity and variability. Spiral drawing tests have recently gained attention as a simple yet informative task for detecting subtle motor impairments, although their diagnostic value depends critically on the analytical methods applied. This study investigates the potential of Hu and Legendre image moments as shape descriptors for classifying spiral drawings from PD patients and healthy controls using two publicly available datasets, one consisting of paper-based and the other of digital spirals. To ensure consistency, all images underwent cropping, resizing, and grayscale normalization before feature extraction. Eight Hu and eight Legendre moments were computed, and Recursive Feature Elimination combined with multiple classifiers under a Leave-One-Subject-Out protocol was used to identify the most informative features. Results showed that Legendre moments consistently outperformed Hu descriptors, achieving an accuracy of 78.8

Cerebral palsy (CP) is a permanent neurological disorder that frequently causes motor impairments, balance dysfunction, and reduced mobility in children. Robotic-assisted gait training (RAGT) has shown promise for improving gait and motor function, but little is known about how neuroplasticity adaptation influence outcomes. This study evaluated the individualized effects of a four-week RAGT program in two pediatric CP patients (GMFCS level 2 and level 5) using multimodal monitoring. Functional near-infrared spectroscopy (fNIRS) signal was recorded at the first RAGT session (T0), sixth session (T1), and twelfth session (T2), to assess cerebral plasticity as well as robotic torque outputs to assess leg (hip and knee) muscle force improvements. Clinical assessments (Modified Ashworth Scale, GMFM-88, WeeFIM®, PedsQL™ CP Module) complemented these measures. The GMFCS level 5 patient, despite severe initial limitations, demonstrated greater improvements, with enhanced prefrontal cortical activation, reduced robotic assistance torque, and decreased spasticity, alongside modest functional gains. In contrast, the GMFCS level 2 patient showed subtler fNIRS variation and torque reduction, reflecting a more stable neuromotor profile. These preliminary findings underscore the value of integrating neurovascular metrics to personalize pediatric neurorehabilitation. Real-time monitoring of brain responses, beyond motor performance alone, may help optimizing RAGT by ensuring interventions are functionally effective.

Osteopathic manipulative treatment (OMT) has shown efficacy in various clinical conditions and age groups. Understanding its neurobiological, particularly autonomic, mechanisms of action remain limited. Preliminary studies suggested a parasympathetic effect of OMT, evidenced by heart-rate-variability analysis. A cross-over RCT on healthy adults was conducted to compare OMT with sham therapy. Thirty-seven participants underwent two sessions (OMT and sham), comprising baseline, tactile treatment, and post-touch. Novel thermal imaging data analyses in combination with seed correlation analyses (SCA) were employed to explore the OMT effects on autonomic parameters. Particularly, the sham group exhibited an elevated warming effect on the cheeks, nose, and chin. Inversely, for the OMT group a conspicuous cooling trend in the nose, but not in the cheeks and chin was observed. Considering SCA maps, the intensity of the correlation for nose tip, glabella and GSR seeds showed higher values in the OMT compared to the sham group. The comparative analysis of thermal maps and SCA results represents a significant advancement in our understanding of the physiological mechanisms underlying OMT’s effects on autonomic functions. By elucidating specific patterns of temperature change, correlation intensity and specific clusters, this research provides valuable insights for optimizing clinical practice and refining theoretical models of manual therapy.

This study focuses on predicting air pollutants on construction sites, which is an essential aspect for preserving the health of workers and people who live nearby. We carried out through data pre-processing, handling missing values and transforming categorical variables. The focus is forecasting different key air pollutants like PM2.5, PM10, $\text{SO}_{2}$, $\text{NO}, \text{CO}_{2}$, and $\mathrm{O}_{3}$. Hence, to achieve this, we compare different statistical, machine learning and deep learning approaches. The novelties of the proposed work from the previous works are: (i) the prediction of multiple pollutants, (ii) the use of multiple predictive models, and (iii) a larger prediction window of 12 hours. We compare the models by computing the Root Mean Squared Error and the $\mathrm{R}^{2}$ to assess their performances. This study provides a comparative analysis of well-known models in the literature for predicting air quality in construction sites. In conclusion, the findings show that training LSTM models can significantly enhance air pollution predictions, providing valuable insights for improving environmental monitoring and forecasting accuracy.

Multiple sclerosis (MS) is a chronic neurological disease often resulting in motor and autonomic dysfunction. This case report investigates the acute and subacute effects of the EXOPULSE Mollii Suit (EMS), a wearable device capable of delivering transcutaneous electrical nerve stimulation to multiple anatomical regions, in a 43-year-old woman with MS. The patient underwent a clinical evaluation before the EMS treatment, during which central nervous system (CNS) and autonomic nervous system (ANS) responses were monitored using electroencephalography (EEG), heart rate variability (HRV), and infrared thermography (IRT). Immediately after the first EMS application, the clinical evaluation was repeated. The intervention continued at home for one month, followed by a post-treatment evaluation similar to the pre-intervention assessment. Functional evaluations showed improvements in sit-to-stand performance (from 8 s to 6 s), muscle tone (MAS scale for the right side from 3 to 2 and for the left side from 2 to 1), clonus, and spasticity (from 3 to 2). EEG results revealed decreased θ-band power (on average, from 0.394 to 0.253) and microstates’ reorganization. ANS activity modifications were highlighted by both HRV (e.g., RMSSD from 0.118 to 0.0837) and IRT metrics (e.g., nose tip temperature sample entropy from 0.090 to 0.239). This study provides the first integrated analysis of CNS and ANS responses to EMS in an MS patient, combining functional scales with multimodal instrumental measurements, emphasizing the possible advantages EMS for MS treatment. Although preliminary, these results demonstrated the potentiality of the EMS to deliver effective and personalized rehabilitative interventions for MS patients.

Air pollution, largely caused by activities in the construction sites, poses serious health and environmental risks to workers and people living nearby. This study focuses on predicting the concentrations of six major pollutants, i.e. PM2.5, PM10, NO2, CO, SO2, and O3. We train a Long Short-Term Memory network (LSTM) on each pollutant to forecast its levels twelve hours in advance. A window generator is used to map data into sequences, enabling the model to capture temporal patterns effectively. Extensive data pre-processing ensures accuracy, including handling missing values and transforming categorical variables. Specifically, the analysis of the pollutants is composed by the following steps: i) preparing the data, ii) building and training the model, iii) evaluating the model performance in terms of Root Mean Square Error (RMSE). We prove that LSTM performs outstandingly over other models, i.e. Random Forest and Artificial Neural Network. The obtained RMSE values ensure credibility and reliability of LSTM in air quality predictions. This predictive framework offers a practical approach for construction sites to manage air pollution and mitigate health and environmental impacts proactively.

IntroductionCerebral palsy (CP) is a group of permanent disorders of movement development that may cause activity limitations. In this context, robot-assisted therapy might play a key role in clinical management. This comprehensive systematic review aimed to investigate the efficacy of robotic systems in improving upper limb (UL) functions in children with CP.MethodsPubMed, EMBASE, Scopus, and PEDro were searched from inception to February 2024. The risk of bias was assessed with the Joanna Briggs Institute critical appraisal tools battery.ResultsOf 756 articles identified, 14 studies involving 193 children with CP with a judged to be of good methodological quality, but with a lack in the study design, were included in the final synthesis. In the included studies a wide range of devices was used, both exoskeletons and end-effectors, both wearable and non-wearable. The CP children who underwent robot-assisted therapy reported a significant overall increase in clinical assessment, specifically in UL movements and manual dexterity. The clinical improvement was often accompanied by a gain also in instrumental assessments (i.e., kinematic analysis, EMG).DiscussionThe present review suggested that robot-assisted therapy can improve UL motor functions in children with CP. Moreover, the availability of different devices with adjustable parameters can represent an important resource in proposing patient-centered-personalized rehabilitation protocols to enhance the efficacy of rehabilitation and integration into daily life. However, the limited sample size and lack of standardized and clearly reproducible protocols impose to recommend the use of robot-assisted therapy as an integration to usual rehabilitation and not as a replacement.Systematic review registrationhttps://osf.io/a78zb/.

BackgroundChildren with cerebral palsy (CP) may present motor and gait impairment.ObjectiveThis systematic review aims to assess the potential of robot-assisted gait training (RAGT) with Lokomat® exoskeleton to improve gait in children with CP.MethodsThe search was conducted and repoted according to PRISMA guidelines on PubMed, Scopus, Cochrane Library and PEDro databases. All randomised controlled trials (RCT) including children with CP who underwent RAGT with Lokomat® were considered eligible. Risk of bias was assessed with the Rob2 tool by two blinded reviewers. The review was previously registered on the PROSPERO database (CRD42023488699).Results948 articles were found, and 9 studies involving 403 children with CP met the inclusion criteria. We found a heterogeneity in the RAGT protocol and a higher risk of bias for two included studies. Seven out nine studies reported a statistically significant improvement (p < 0.05) on gait, balance, or global functions with respect to control groups. Specifically, walking speed and stride length were improved after RAGT.ConclusionsChildren with CP can benefit from the add-on therapy with RAGT through Lokomat® to improve walking and balance function. There is a need for RCTs with better patient stratification and with less heterogeneity in outcomes to improve the quality of the pooled evidence.

This paper presents a novel interface enabling seamless human-robot interaction by integrating the NAO social robot with OpenAI’s generative language model, ChatGPT. The interface, developed in Python, allows users to interact verbally with the NAO robot, which responds with context-aware and natural language outputs generated by ChatGPT. The system architecture includes a graphical user interface and dual-script integration to overcome compatibility issues between Python 2.7 (for NAO) and Python 3.x (for OpenAI API). Experimental evaluations demonstrated an average response time of 1.37 seconds per 10 tokens and confirmed stable performance under moderate background noise conditions. These results highlight the potential of combining generative AI with social robots to enhance engagement and communication in real-world applications, including education, healthcare, and assistive robotics.

Cerebral palsy (CP) is a permanent, non-degenerative neurological condition that often leads to motor impairments, balance disorders, and reduced functional mobility in children. Robotic-assisted gait training (RAGT) has shown promising results in enhancing motor function and gait performance in individuals with neurological conditions, including CP. Beyond motor improvements, therapy effectiveness is increasingly linked to patient engagement. In this case study, we investigated the impact of a four-week RAGT program on two children with CP classified as level 2 and level 5 on the Gross Motor Function Classification System (GMFCS) by analyzing both physiological and neurophysiological signals. The main aim of the study is to investigate the effects of RAGT on autonomic activity, brain plasticity, and therapy responsiveness in relation to individual functional levels. Heart rate variability (HRV), infrared thermography (IRT), and functional near-infrared spectroscopy (fNIRS) were used to monitor autonomic nervous system activity, emotional engagement, and cortical hemodynamics at three time points: first session (T0), sixth session (T1), and twelfth session (T2). The results demonstrated a higher level of efficacy in the subject with GMFCS 5 particularly in terms of cortical activation and physiological engagement. Moreover, the integration of multimodal monitoring techniques offers a comprehensive perspective on therapy outcomes, supporting the use of RAGT in personalized rehabilitation strategies.

Dementia affects millions of people worldwide, placing a significant burden on healthcare systems. Early and accurate diagnosis is crucial for managing the disease, with clinical tests like the Mini-Mental State Examination (MMSE) playing a key role. To alleviate pressure on healthcare resources and support telemedicine, remote administration of such tests through artificial agents must be explored. However, different administration modalities may evoke varying psychophysiological responses in patients. This study aims to evaluate the feasibility of administering the MMSE via a tablet and compare the psychophysiological responses elicited by tablet-based and psychologist-led administration. Contactless technologies, such as thermography, are used to measure these responses, providing valuable insight into stress and engagement. The Wilcoxon signed rank test delivered a lower engagement state (zval=-3.895, p=9.84.10-5) during the psychologist-administered MMSE when compare with the tablet-based administration. The study explores the potential for tablet-based MMSE administration, demonstrating its reliability and fostering the shifting to digital methods.

A stroke is a critical medical disease characterized by the abrupt cessation of blood flow to the brain, resulting in cellular injury or necrosis. The effects of stroke on people might range from minor impairments to significant disabilities. Stroke treatment often requires gait rehabilitation. The efficacy of rehabilitative treatment is often assessed using clinical scales. Among them, the Performance-Oriented Mobility Assessment (POMA) is commonly used for assessing balance and gait in stroke patients. Importantly, evaluating muscle activation and kinematic patterns by electromyography (EMG) and stereophotogrammetry during ambulation might provide insights into gait impairments. This study intends to use a machine learning-based regression to predict the POMA total score using EMG and kinematic data in stroke patients. The model achieved correlations of 0.70 and 0.67 throughout the validation and testing stages, respectively. The t-test indicated no bias between the estimated and measured POMA values, but the Bland-Altmann plot revealed a systematic error in the model. Although preliminary, these results indicate the potential to construct models that might provide significant assistance to clinicians by offering accurate evaluations of motor deficits reported post-stroke.