With the increasing adoption of predictive healthcare systems, Federated Learning (FL) has emerged as a privacy-preserving paradigm for collaborative model training, which is crucial for sensitive health data. This study investigates the integration of Kolmogorov-Arnold Network (KAN) with FL to enhance predictive healthcare applications. KAN’s innovative implementation of spline-based activation functions offers enhanced flexibility, interoperability, and efficiency in capturing complex nonlinear data patterns with fewer parameters than traditional neural networks. When benchmarked against standard Multilayer Perceptrons (MLPs), KANs show significant performance improvements in both real-world and synthetic electrocardiogram (ECG) datasets. Notably, KAN’s adaptive activation functions, particularly quadratic splines with a grid size of 30, effectively capture the complexities of time-series data, achieving a test accuracy of 93.73% and an F1-score of 92.90% on the MIT-BIH arrhythmia dataset. Furthermore, KAN outperformed MLP on synthetic ECG data, highlighting its potential for broader medical applications. Despite its higher computational overhead, federated KAN shows considerable promise in healthcare settings that require both privacy and high accuracy in modeling complex data patterns. The findings underscore the need for optimization to address computational inefficiencies in KAN-based FL environments while establishing a foundation for future applications of KAN in time-series-based medical diagnostics.

To mitigate the impact of COVID-19, the government has adopted various NPIs (Non-Pharmaceutical Interventions) ranging from wearing masks to social distancing. Over the due course of time, it has been proven that these NPIs are effective but assessing the effectiveness of the NPIs on the COVID-19 spread is still discussed. Till now, case confirmation and hospitalization have been incorporated as indicators to determine the success of NPIs. Here, we compare the effectiveness of two indicators such as the death rate and the number of cases, and their variation with human mobility, in assessing the effectiveness of NPIs to control the impact of COVID-19. The study includes the daily number of COVID-19 cases and deaths, Google Mobility Reports, and information on NPIs in 9 Italian regions for over 2 years from 2020. Similar considerations can be applied to other countries. The intent is to improve the method proposed by Wang et al. in 2020. Our findings suggest that in combination with human mobility, the death rate works better than the number of cases in assessing the effectiveness of NPIs. These findings can help policymakers formulate the best data-driven approaches for tackling confinement issues and structuring future scenarios in case of new outbreaks.

Artificial intelligence is transforming healthcare with a profound paradigm shift impacting diagnostic techniques, drug discovery, health analytics, interventions and much more. In this paper we focus on exploiting AI-based chatbot systems, mainly based on machine learning algorithms and Natural Language Processing, to understand and respond to needs of patients and their families. In particular, we describe an application scenario for an AI-chatbot delivering support to pregnant women, mothers, and families with young children, by giving them help and instructions in relevant situations.

Fetal biometry is considered the keystone in fetal well-being assessment. In particular, fetal growth curves built by means of ultrasound images and reference charts (defining the normal and pathological sizes for each biometric parameter and for each gestational age) are extensively adopted to track fetal sizes from the early phases of pregnancy up to delivery. In literature a large variety of reference charts are reported to consider the differences among different ethnic groups, but they are up to five decades old and they do not consider environmental factors such as foods, lifestyle, smoke, familial aspects, physiological and pathological variables, temporal parameters etc., which cannot be disregarded in a correct diagnosis. Therefore, current reference charts are rapidly becoming inadequate to support the melting pot of ethnic groups and lifestyles of our society, while customized reference charts can provide an accurate fetal assessment for the different fetal anthropometrical variables. Starting from a detailed analysis of the limits of classical reference charts, the paper presents a new method, based on multidimensional analysis for creating personalized fetal growth curves. A simple implementation, based on Open Source software and simulated data, shows the need of Big Data techniques in order to scale up the problem.