Nicholas E. Protonotarios

Medical image reconstruction began with an exact mathematical result. In 1917, Radon showed how to recover a function from its line integrals, providing the mathematical basis for computed tomography, positron emission tomography, and single-photon emission computed tomography. This review follows the development of reconstruction from analytical inversion to modern deep learning. We first present the Radon transform, its attenuated and three-dimensional generalizations, and the corresponding inversion formulae, including filtered backprojection. We then review iterative and optimization-based methods, including maximum-likelihood expectation-maximization, variational regularization, and compressed sensing. The discussion proceeds to deep learning, covering convolutional and transformer architectures, algorithm unrolling, physics-informed networks, neural operators, and generative diffusion priors. Our central argument is that analytical, iterative, and learned reconstruction are different formulations of the same inverse problem rather than competing paradigms. Many learned components derive from classical formulations, and many classical operators admit learnable counterparts. We conclude by discussing open challenges related to theoretical guarantees, trustworthiness, data efficiency, federated learning, and regulatory validation. The most reliable learned systems are likely to retain the mathematical and physical structure of established reconstruction methods.

The year 1958 marked the beginning of a new era, as Rosenblatt introduced the Perceptron, a computational model for supervised learning through error-driven weight adaptation inspired by biological neurons. The backpropagation algorithm, introduced by Rumelhart et al. in 1986, transformed multilayer networks by allowing for weight updates. In 1997, Maass proposed spiking neural networks to emulate the temporal and sequential learning of biological neurons, while Schultz et al. proposed reinforcement learning by integrating brain-like reward mechanisms for decision making. The 2000s witnessed significant advancements, with the development of convolutional neural networks and recurrent neural networks inspired by brain’s hierarchical and recurrent mechanisms, respectively. More recently, in 2020, Feinman and Lake presented hybrid neurosymbolic models to capture the brain’s ability to learn from sensory data and perform structured reasoning. This work explores the remarkable progression of breakthroughs toward brain-inspired learning, aiming to offer a new perspective in the dynamic interplay between neuroscience and computer science.

We introduce the augmented total variation () regularization method for Radon‐type inverse problems. Our novel approach incorporates a dual variable into the regularization process, thereby extending and essentially augmenting traditional regularization techniques. The proposed method is robust, requiring only one algorithmic iteration to achieve accurate reconstructions. Numerical experiments on a modified Shepp–Logan phantom demonstrate that the augmented regularization consistently yields higher structural similarity index metric (SSIM) values and lower mean absolute difference (MAD) values compared to filtered backprojection and standard regularization. These findings indicate that our method not only reduces reconstruction errors but also preserves structural details.

As vehicle technologies rapidly advance, video streaming capabilities emerge as a significant feature of modern on-board vehicular systems. The integration of 360 $^{\circ }$ video streaming in vehicles is expected to enhance road situational awareness, by providing panoramic live streaming and recording capabilities. This paper introduces the concept of roads digital twin , combining, for the first time, 360 $^{\circ }$ video streaming with graph neural networks, in order to enhance predictive situational awareness in road environments. To this end, we have developed eGAT, a selective edge-enhanced graph attention network architecture, that uses graph neural networks with attention mechanisms. eGAT is capable of effectively predicting road coverage, considering uplink bandwidth limitations that may affect video streaming quality and user quality of experience (QoE). For the evaluation of our novel method, we utilized four different datasets, considering several vehicular scenarios. For performance comparison purposes, we employed three metrics, namely the overall percentage of the covered region, the normalized mutual information (NMI), and the precision-recall scores. In terms of overall coverage percentage, eGAT provided superior coverage performance against similar studies in all nine scenarios and for all numbers of streaming vehicles investigated, reaching an increase of up to 32.3%. In terms of NMI score, for low values of prediction horizon, eGAT outperformed similar attention-based algorithms, with an increase of up to 29.6%, whereas for larger prediction horizons, eGAT presented lower, however comparable, coverage performance. Our results suggest that eGAT is a promising solution for scenarios involving streaming vehicles, showcasing its potential for applications such as autonomous vehicles and traffic management systems.

Kinase inhibitors are potent therapeutic agents in cancer treatment, but their effectiveness is frequently restricted by the inability to image the tumor microenvironment. To address this constraint, kinase inhibitor–fluorophore conjugates have emerged as promising theranostic agents, allowing for simultaneous cancer diagnosis and treatment. These conjugates are gaining attention for their ability to visualize malignant tissues and concurrently enhance therapeutic interventions. This review explores the design principles governing the development of multimodal inhibitors, highlighting their potential as platforms for kinase tracking and inhibition via bioimaging. The structural aspects of constructing such theranostic agents are critically analyzed. This work could shed light on this intriguing field and provide adequate impetus for developing novel theranostic compounds based on small molecule inhibitors and fluorophores.

Federated learning (FL) has emerged as a pivotal paradigm for training machine learning models across decentralized devices while maintaining data privacy. In the healthcare domain, FL enables collaborative training among diverse medical devices and institutions, enhancing model robustness and generalizability without compromising patient privacy. In this paper, we propose DPS-GAT, a novel approach integrating graph attention networks (GATs) with differentially private client selection and resource allocation strategies in FL. Our methodology addresses the challenges of data heterogeneity and limited communication resources inherent in medical applications. By employing graph neural networks (GNNs), we effectively capture the relational structures among clients, optimizing the selection process and ensuring efficient resource distribution. Differential privacy mechanisms are incorporated, to safeguard sensitive information throughout the training process. Our extensive experiments, based on the Regensburg pediatric appendicitis open dataset, demonstrated the superiority of our approach, in terms of model accuracy, privacy preservation, and resource efficiency, compared to traditional FL methods. The ability of DPS-GAT to maintain a high and stable number of client selections across various rounds and differential privacy budgets has significant practical implications, indicating that FL systems can achieve strong privacy guarantees without compromising client engagement and model performance. This balance is essential for real-world applications where both privacy and performance are paramount. This study suggests a promising direction for more secure and efficient FL medical applications, which could improve patient care through enhanced predictive models and collaborative data utilization.

MDCG 2019-16 is intended to assist practitioners in compliance with the Medical Device Regulation and the In-Vitro Device Regulation. This paper presents a gap analysis of MDCG 2019-16, identifying key gaps and proposing a robust set of recommendations to enhance the IoMT regulatory framework. This work has been undertaken by a selection of current (2023-2025) projects, all funded under the Horizon Europe call “Enhancing cybersecurity of connected medical devices”: HORIZON-HLTH-2022-IND-13-01, and this paper summarises observations and recommendations from across these projects. There is considerable consensus across the projects in many recommendation themes, notably; linking cybersecurity with patient safety and privacy; keeping the guidelines current; and usage recipes for the guidelines. The paper also suggests toolkit solutions to address some of the recommendations.