Ilenia Tinnirello
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Ilenia Tinnirello has a total of 172 co-authors. Here are the top co-authors by number of shared publications:
- Giuseppe Bianchi
University of Rome Tor Vergata, Rome, Italy
30 shared publications - Domenico Garlisi
CNIT, Parma, Italy
23 shared publications - Daniele Croce
University of Palermo
20 shared publications - Fabrizio Giuliano
CNIT, Parma, Italy
20 shared publications - Laura Giarre'
Department of Engineering “Enzo Ferrari”, University of Modena and Reggio Emilia, 41125 Modena, Italy
19 shared publications - Stefano Mangione
Department of Engineering, University of Palermo, Palermo, Italy
18 shared publications
Latest Publications
- Open Access
- 24 September 2026
- Published by MDPI AG in Diagnostics
- Vol. 16 (19) , 3095
Proactive Prediction of Alarms in Hemodialysis Machines: A Comparative Evaluation of AI Models for Clinical Prognostic Support
Background/Objectives: Hemodialysis is a treatment for patients with kidney disease, yet it remains risk-prone due to potential technical or physiological instabilities. Dialysis machines rely on reactive safety mechanisms that trigger alarms only when predefined thresholds are reached. This often alerts clinical staff only after a complication has already occurred, contributing to potential delays in interventions. This study aims to develop a proactive prognostic framework capable of anticipating system instabilities before they transform into clinical adverse events. Methods: This work proposes a predictive approach for the proactive detection of patient- or process-related alarm conditions using ten machine learning models and one deep learning architecture. A dataset consisting of over two million records sampled at 0.5 s intervals from 95 dialysis sessions was assembled by merging machine logs of sensor readings and the binary machine status (alarm/normal). Results: Among the models tested, Random Forest emerged as the most robust predictive tool, achieving an F1 score higher than 0.85 for predicting alarms up to one minute in advance, with precision (~0.83), recall (~0.87), specificity (~0.99), AUROC (~0.99), and AUPRC (~0.94) values indicating strong performance, and stable performance across validation and test sets. A progressive decrease in performance was observed for longer prediction time windows. Conclusions: These findings demonstrate the feasibility of predicting dialysis-machine-generated alarms in advance using machine learning and deep learning models. Providing advance warning could give clinicians the opportunity to anticipate impending alarm conditions and, where appropriate, implement timely corrective actions, potentially improving treatment management and the operational reliability of renal replacement therapies.
- 24 September 2026
- Published by MDPI AG in Diagnostics
- Vol. 16 (19) , 3095
Open AccessProactive Prediction of Alarms in Hemodialysis Machines: A Comparative Evaluation of AI Models for Clinical Prognostic Support
Background/Objectives: Hemodialysis is a treatment for patients with kidney disease, yet it remains risk-prone due to potential technical or physiological instabilities. Dialysis machines rely on reactive safety mechanisms that trigger alarms only when predefined thresholds are reached. This often alerts clinical staff only after a complication has already occurred, contributing to potential delays in interventions. This study aims to develop a proactive prognostic framework capable of anticipating system instabilities before they transform into clinical adverse events. Methods: This work proposes a predictive approach for the proactive detection of patient- or process-related alarm conditions using ten machine learning models and one deep learning architecture. A dataset consisting of over two million records sampled at 0.5 s intervals from 95 dialysis sessions was assembled by merging machine logs of sensor readings and the binary machine status (alarm/normal). Results: Among the models tested, Random Forest emerged as the most robust predictive tool, achieving an F1 score higher than 0.85 for predicting alarms up to one minute in advance, with precision (~0.83), recall (~0.87), specificity (~0.99), AUROC (~0.99), and AUPRC (~0.94) values indicating strong performance, and stable performance across validation and test sets. A progressive decrease in performance was observed for longer prediction time windows. Conclusions: These findings demonstrate the feasibility of predicting dialysis-machine-generated alarms in advance using machine learning and deep learning models. Providing advance warning could give clinicians the opportunity to anticipate impending alarm conditions and, where appropriate, implement timely corrective actions, potentially improving treatment management and the operational reliability of renal replacement therapies.
- 4 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. PP (18) , 1
From Digital Twins to Autonomous Local Agents: Rethinking AI for Smart Water Networks
Internet of Things (IoT) enabled Water Distribution Networks (WDNs) increasingly rely on Digital Twins (DTs) to support monitoring and decision-making, yet most DTs remain passive dashboards. Deploying Artificial Intelligence (AI) within these critical infrastructures requires efficient, autonomous solutions at the edge, where operational constraints and data sovereignty concerns limit the suitability of proprietary cloud-based services that expose sensitive telemetry beyond internal boundaries. At the same time, Large Language Models (LLMs) impose significant computational and energy demands, making lightweight and locally executable agentic pipelines essential for practical deployment under edge constraints. This work introduces a modular multi-agent framework that provides an on-demand DT service capable of executing complex “what-if” hydraulic analyses directly at the network edge. The system employs an autonomous agentic Retrieval-Augmented Generation (RAG) pipeline that enables locally deployed LLMs to interpret user intents, retrieve domain-specific context, generate simulation workflows, and self-correct execution errors without human intervention. This multi-agent design yields robust and trustworthy behavior even under strict resource constraints. Beyond introducing the framework, this work systematically assesses its trustworthiness and feasibility for edge deployment by benchmarking task reliability, hardware requirements, and energy efficiency. Evaluation across a representative benchmark of WDNs shows that the proposed agentic approach improves full-task success rates by up to 30-40% over zero-shot prompting and reduces execution failures by more than 50%. Moreover, 4-bit quantized local models deliver performance competitive with proprietary cloud solutions while lowering peak VRAM usage and energy consumption by over 60%. These results demonstrate the viability of effective, efficient, and trustworthy AI agents operating entirely on edge hardware to securely and autonomously enable advanced IoT infrastructure services.
- 4 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. PP (18) , 1
From Digital Twins to Autonomous Local Agents: Rethinking AI for Smart Water Networks
Internet of Things (IoT) enabled Water Distribution Networks (WDNs) increasingly rely on Digital Twins (DTs) to support monitoring and decision-making, yet most DTs remain passive dashboards. Deploying Artificial Intelligence (AI) within these critical infrastructures requires efficient, autonomous solutions at the edge, where operational constraints and data sovereignty concerns limit the suitability of proprietary cloud-based services that expose sensitive telemetry beyond internal boundaries. At the same time, Large Language Models (LLMs) impose significant computational and energy demands, making lightweight and locally executable agentic pipelines essential for practical deployment under edge constraints. This work introduces a modular multi-agent framework that provides an on-demand DT service capable of executing complex “what-if” hydraulic analyses directly at the network edge. The system employs an autonomous agentic Retrieval-Augmented Generation (RAG) pipeline that enables locally deployed LLMs to interpret user intents, retrieve domain-specific context, generate simulation workflows, and self-correct execution errors without human intervention. This multi-agent design yields robust and trustworthy behavior even under strict resource constraints. Beyond introducing the framework, this work systematically assesses its trustworthiness and feasibility for edge deployment by benchmarking task reliability, hardware requirements, and energy efficiency. Evaluation across a representative benchmark of WDNs shows that the proposed agentic approach improves full-task success rates by up to 30-40% over zero-shot prompting and reduces execution failures by more than 50%. Moreover, 4-bit quantized local models deliver performance competitive with proprietary cloud solutions while lowering peak VRAM usage and energy consumption by over 60%. These results demonstrate the viability of effective, efficient, and trustworthy AI agents operating entirely on edge hardware to securely and autonomously enable advanced IoT infrastructure services.
- Open Access
- 4 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Communications Letters
- Vol. 30 (10897798) , 1275-1279
Compact Attention-Augmented Neural Decoder for Chaos-Based Wireless Communications
Chaos-based modulation offers strong interference resilience but decoding remains challenging under low SNRs and fading channels. Existing deep learning receivers achieve promising results but are often too large for resource-constrained systems. We propose Ultra-CAN, an ultra-light convolutional–attention decoder for Chaos Shift Keying (CSK) that combines efficient convolutional feature extraction with a compact attention block to capture long-range dependencies. Our design achieves superior Bit Error Rate (BER) performance compared to state-of-the-art demodulators in AWGN and Rayleigh channels, with a footprint of 220.4 kB. Results demonstrate the viability of compact attention-based decoders for chaotic communications.
- 4 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Communications Letters
- Vol. 30 (10897798) , 1275-1279
Open AccessCompact Attention-Augmented Neural Decoder for Chaos-Based Wireless Communications
Chaos-based modulation offers strong interference resilience but decoding remains challenging under low SNRs and fading channels. Existing deep learning receivers achieve promising results but are often too large for resource-constrained systems. We propose Ultra-CAN, an ultra-light convolutional–attention decoder for Chaos Shift Keying (CSK) that combines efficient convolutional feature extraction with a compact attention block to capture long-range dependencies. Our design achieves superior Bit Error Rate (BER) performance compared to state-of-the-art demodulators in AWGN and Rayleigh channels, with a footprint of 220.4 kB. Results demonstrate the viability of compact attention-based decoders for chaotic communications.
- 2 January 2026
- Published by Springer Nature
- p. 1-15
A Federated Learning Approach for Predicting Marine Heat Waves
- 2 January 2026
- Published by Springer Nature
- p. 1-15
A Federated Learning Approach for Predicting Marine Heat Waves
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Communications Society
- Vol. 7, 10768-10790
Adversarial CSI Obfuscation Against Device-Free Activity Recognition in RIS-Assisted Systems
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Communications Society
- Vol. 7, 10768-10790
Open AccessAdversarial CSI Obfuscation Against Device-Free Activity Recognition in RIS-Assisted Systems
- Open Access
- 18 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 214223-214235
Investigating the Impact of Rational Dilated Wavelet Transform on Motor Imagery EEG Decoding With Deep Learning Models
Decoding neural signals from electroencephalogram (EEG) data is a challenging task due to the signals’ intrinsic non-stationarity, low signal-to-noise ratio, and complex spatio-temporal dynamic. The present study investigates the impact of the Rational DilatedWavelet Transform (RDWT) for implementing a de-noising operation before applying deep learning classifiers devised to recognize Motor Imagery (MI) EEG signals. A systematic paired evaluation (with/without RDWT) is conducted on four state-of-the-art deep learning architectures and carried out across three benchmark datasets. The performance of the models integrating RDWT is reported with subject-wise averages using accuracy and Cohen’s kappa, complemented by subject-level analyses to identify when RDWT is beneficial. Our results show improvements in classifier performance in almost all cases, and especially for MI tasks/trials that are difficult to classify. We then conclude thatRDWT-based signal preprocessing can mitigate localized noise and enhance the EEG classification performance without adding significant complexity to the classifier architecture.
- 18 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 214223-214235
Open AccessInvestigating the Impact of Rational Dilated Wavelet Transform on Motor Imagery EEG Decoding With Deep Learning Models
Decoding neural signals from electroencephalogram (EEG) data is a challenging task due to the signals’ intrinsic non-stationarity, low signal-to-noise ratio, and complex spatio-temporal dynamic. The present study investigates the impact of the Rational DilatedWavelet Transform (RDWT) for implementing a de-noising operation before applying deep learning classifiers devised to recognize Motor Imagery (MI) EEG signals. A systematic paired evaluation (with/without RDWT) is conducted on four state-of-the-art deep learning architectures and carried out across three benchmark datasets. The performance of the models integrating RDWT is reported with subject-wise averages using accuracy and Cohen’s kappa, complemented by subject-level analyses to identify when RDWT is beneficial. Our results show improvements in classifier performance in almost all cases, and especially for MI tasks/trials that are difficult to classify. We then conclude thatRDWT-based signal preprocessing can mitigate localized noise and enhance the EEG classification performance without adding significant complexity to the classifier architecture.
- 9 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Computer
- Vol. 58 (12) , 46-54
The Synergy of Digital Twins and Machine Learning in Sustainable Water Management: SWIM
Smart Water Interaction & Monitoring (SWIM) is a comprehensive tool that encapsulates digital twins, the Internet of Things, and machine learning to modernize smart water distribution networks. SWIM enables real-time monitoring and predictive analytics and lays the foundation for robust anomaly detection.
- 9 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Computer
- Vol. 58 (12) , 46-54
The Synergy of Digital Twins and Machine Learning in Sustainable Water Management: SWIM
Smart Water Interaction & Monitoring (SWIM) is a comprehensive tool that encapsulates digital twins, the Internet of Things, and machine learning to modernize smart water distribution networks. SWIM enables real-time monitoring and predictive analytics and lays the foundation for robust anomaly detection.
- Open Access
- 3 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 65-72
Towards Intelligent Mobility Management: Customized Handover in 5G O-RAN Networks
- 3 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 65-72
Open AccessTowards Intelligent Mobility Management: Customized Handover in 5G O-RAN Networks
- Open Access
- 3 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 105-112
An Energy-Harvesting Shield and Solar Testbed for IoT: LoRa Performance Insights and Power Measurements
- 3 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 105-112
Open AccessAn Energy-Harvesting Shield and Solar Testbed for IoT: LoRa Performance Insights and Power Measurements
- Open Access
- 3 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 113-120
Silent Drain: From Energy Profiling to Practical Denial-of-Energy Attacks in 5G
- 3 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 113-120
Open AccessSilent Drain: From Energy Profiling to Practical Denial-of-Energy Attacks in 5G