Synthesizing amyloid PET scans from the more widely available and accessible structural MRI modality offers a promising, cost-effective approach for large-scale Alzheimer's Disease (AD) screening. This is motivated by evidence that, while MRI does not directly detect amyloid pathology, it may nonetheless encode information correlated with amyloid deposition that can be uncovered through advanced modeling. However, the high dimensionality and structural complexity of 3D neuroimaging data pose significant challenges for existing MRI-to-PET translation methods. Modeling the cross-modality relationship in a lower-dimensional latent space can simplify the learning task and enable more effective translation. As such, we present CoCoLIT (ControlNet-Conditioned Latent Image Translation), a diffusion-based latent generative framework that incorporates three main innovations: (1) a novel Weighted Image Space Loss (WISL) that improves latent representation learning and synthesis quality; (2) a theoretical and empirical analysis of Latent Average Stabilization (LAS), an existing technique used in similar generative models to enhance inference consistency; and (3) the introduction of ControlNet-based conditioning for MRI-to-PET translation. We evaluate CoCoLIT's performance on publicly available datasets and find that our model significantly outperforms state-of-the-art methods on both image-based and amyloid-related metrics. Notably, in amyloid-positivity classification, CoCoLIT outperforms the second-best method with improvements of +10.5% on the internal dataset and +23.7% on the external dataset.

Deep generative models have emerged as a transformative tool in medical imaging, offering substantial potential for synthetic data generation. However, recent empirical studies highlight a critical vulnerability: these models can memorize sensitive training data, posing significant risks of unauthorized patient information disclosure. Detecting memorization in generative models remains particularly challenging, necessitating scalable methods capable of identifying training data leakage across large sets of generated samples. In this work, we propose DeepSSIM, a novel self-supervised metric for quantifying memorization in generative models. DeepSSIM is trained to: i) project images into a learned embedding space and ii) force the cosine similarity between embeddings to match the ground-truth Structural Similarity Index (SSIM) scores computed in the image space. To capture domain-specific anatomical features, training incorporates structure-preserving augmentations, allowing DeepSSIM to estimate similarity reliably without requiring precise spatial alignment. We evaluate DeepSSIM in two case studies using synthetic brain MRI and chest X-ray data generated by a Latent Diffusion Model (LDM) trained under memorization-prone conditions. Compared to state-of-the-art memorization metrics, DeepSSIM achieves superior performance, improving F1 scores by an average of +52.03% over the best existing method. Code and data are publicly available at https://github.com/brAIn-science/DeepSSIM.

The rapid advancements in Artificial Intelligence (AI) and edge devices have driven the proliferation of smart-world applications, many of which are deployed in distributed environments. In contexts where data privacy must be maintained, Federated Learning (FL) has emerged as a promising privacy-preserving collaborative learning paradigm, offering significant potential to harness the distributed data available at the network’s edge. However, the heterogeneity of devices and the non-Independent and Identically Distributed (IID) nature of data across them pose significant challenges, undermining the performance of FL systems. To address these issues, Knowledge Distillation (KD) has proven to be a valid solution. Initially developed as a teacher-student training paradigm, KD enhances the performance of smaller student networks by transferring knowledge from larger, more powerful teacher networks. When integrated with FL, KD offers additional benefits: it mitigates the effects of data and model heterogeneity and compensates for the absence of a centralized teacher model. This is achieved by creating a global knowledge representation derived from the aggregated knowledge of individual clients. Building on these principles, we propose Weighted-FD, a novel framework that introduces a quality-aware approach to global knowledge computation. Unlike conventional methods, Weighted-FD evaluates the quality of knowledge contributed by each client and dynamically adjusts their influence on the global knowledge representation. This ensures a more accurate and effective aggregation process. We detail the mathematical foundation of our framework and validate its efficacy through extensive experiments conducted on MNIST, FashionMNIST, and CIFAR-10 under different data heterogeneity settings (IID, weak non-IID, and strong non-IID). The results demonstrate that the proposed method consistently outperforms existing federated distillation approaches. In particular, Weighted-FD achieves substantial improvements under strong non-IID conditions, reaching accuracy gains of up to 57.12% over FedMD and 49.34% over Selective-FD on CIFAR-10, while maintaining low computational and memory requirements, making it well suited for deployment in resource-constrained edge environments. The source code for Weighted-FD is publicly available at the following link: https://anonymous.4open.science/r/weighted-fd-C6EC/ .

Accurate brain age prediction from MRI is a promising biomarker for brain health and neurodegenerative disease risk, but current deep learning models often lack anatomical specificity and clinical insight. We present a regional patch-based ensemble framework that uses 3D Convolutional Neural Networks (CNNs) trained on bilateral patches from ten subcortical structures, enhancing anatomical sensitivity. Ensemble predictions are combined with cognitive assessments to derive a cognitively informed proxy for cognitive reserve (CR-Proxy), quantifying resilience to age-related brain changes. We train our framework on a large, multi-cohort dataset of healthy controls and test it on independent samples that include individuals with Alzheimer’s disease and mild cognitive impairment. The results demonstrate that our method achieves robust brain age prediction and provides a practical, interpretable CR-Proxy capable of distinguishing diagnostic groups and identifying individuals with high or low cognitive reserve. This pipeline offers a scalable, clinically accessible tool for early risk assessment and personalized brain health monitoring. • Regional patch-based ensemble model enhances brain age prediction using 3D CNNs on 10 subcortical structures. • Cognitive Reserve Proxy (CR-Proxy) combines brain age estimates with MMSE scores for resilience assessment. • CR-Proxy distinguishes diagnostic groups: AD, MCI, and cognitively normal with high significance. • Ensemble model achieves MAE of 2.93 years, outperforming individual regional predictors. • Framework provides scalable biomarker for early neurodegenerative risk stratification.

This exploratory study analyses parameters influencing a tele-rehabilitation program using the Virtual Reality Rehabilitation System (VRSS) HomeKit for patients with Parkinson’s disease (PD). Data from 10 patients with idiopathic PD who completed 20 upper limb motor exercise sessions were analysed to assess correlations between system-generated metrics and clinical outcomes. Patients were clinically assessed with the Fugl-Meyer Assessment (FMA) and the Unified Parkinson’s Disease Rating Scale (UPDRS). In the experiments, we focused on two macro categories of exercises performed by patients, i.e., reaching and catching, while analysing how correlations among patients’ age, clinician-provided Hoehn-Yahr (HY) stage, and VRRS-provided metrics (such as repetitions, mean duration, correct responses, and omission errors) influenced exercise score assessments and other potential outcomes. This study is propaedeutic for advanced ML-based analyses to model patient progress and predict outcomes throughout treatment, laying the foundation for patient-centric precision medicine and the development of tailored remote rehabilitation strategies.

The growing availability of longitudinal Magnetic Resonance Imaging (MRI) datasets has facilitated Artificial Intelligence (AI)-driven modeling of disease progression, making it possible to predict future medical scans for individual patients. However, despite significant advancements in AI, current methods continue to face challenges including achieving patient-specific individualization, ensuring spatiotemporal consistency, efficiently utilizing longitudinal data, and managing the substantial memory demands of 3D scans. To address these challenges, we propose Brain Latent Progression (BrLP), a novel spatiotemporal model designed to predict individual-level disease progression in 3D brain MRIs. The key contributions in BrLP are fourfold: (i) it operates in a small latent space, mitigating the computational challenges posed by high-dimensional imaging data; (ii) it explicitly integrates subject metadata to enhance the individualization of predictions; (iii) it incorporates prior knowledge of disease dynamics through an auxiliary model, facilitating the integration of longitudinal data; and (iv) it introduces the Latent Average Stabilization (LAS) algorithm, which (a) enforces spatiotemporal consistency in the predicted progression at inference time and (b) allows us to derive a measure of the uncertainty for the prediction at the global and voxel level. We train and evaluate BrLP on 11,730 T1-weighted (T1w) brain MRIs from 2,805 subjects and validate its generalizability on an external test set comprising 2,257 MRIs from 962 subjects. Our experiments compare BrLP-generated MRI scans with real follow-up MRIs, demonstrating state-of-the-art accuracy compared to existing methods. The code is publicly available at: https://github.com/LemuelPuglisi/BrLP.

Large medical imaging data sets are becoming increasingly available. A common challenge in these data sets is to ensure that each sample meets minimum quality requirements devoid of significant artefacts. Despite a wide range of existing automatic methods having been developed to identify imperfections and artefacts in medical imaging, they mostly rely on data-hungry methods. In particular, the scarcity of artefact-containing scans available for training has been a major obstacle in the development and implementation of machine learning in clinical research. To tackle this problem, we propose a novel framework having four main components: (1) a set of artefact generators inspired by magnetic resonance physics to corrupt brain MRI scans and augment a training dataset, (2) a set of abstract and engineered features to represent images compactly, (3) a feature selection process that depends on the class of artefact to improve classification performance, and (4) a set of Support Vector Machine (SVM) classifiers trained to identify artefacts. Our novel contributions are threefold: first, we use the novel physics-based artefact generators to generate synthetic brain MRI scans with controlled artefacts as a data augmentation technique. This will avoid the labour-intensive collection and labelling process of scans with rare artefacts. Second, we propose a large pool of abstract and engineered image features developed to identify 9 different artefacts for structural MRI. Finally, we use an artefact-based feature selection block that, for each class of artefacts, finds the set of features that provide the best classification performance. We performed validation experiments on a large data set of scans with artificially-generated artefacts, and in a multiple sclerosis clinical trial where real artefacts were identified by experts, showing that the proposed pipeline outperforms traditional methods. In particular, our data augmentation increases performance by up to 12.5 percentage points on the accuracy, F1, F2, precision and recall. At the same time, the computation cost of our pipeline remains low - less than a second to process a single scan - with the potential for real-time deployment. Our artefact simulators obtained using adversarial learning enable the training of a quality control system for brain MRI that otherwise would have required a much larger number of scans in both supervised and unsupervised settings. We believe that systems for quality control will enable a wide range of high-throughput clinical applications based on the use of automatic image-processing pipelines.

Accurate and realistic simulation of high-dimensional medical images has become an important research area relevant to many AI-enabled healthcare applications. However, current state-of-the-art approaches lack the ability to produce satisfactory high-resolution and accurate subject-specific images. In this work, we present a deep learning framework, namely 4D-Degenerative Adversarial NeuroImage Net (4D-DANI-Net), to generate high-resolution, longitudinal MRI scans that mimic subject-specific neurodegeneration in ageing and dementia. 4D-DANI-Net is a modular framework based on adversarial training and a set of novel spatiotemporal, biologically-informed constraints. To ensure efficient training and overcome memory limitations affecting such high-dimensional problems, we rely on three key technological advances: i) a new 3D training consistency mechanism called Profile Weight Functions (PWFs), ii) a 3D super-resolution module and iii) a transfer learning strategy to fine-tune the system for a given individual. To evaluate our approach, we trained the framework on 9852 T1-weighted MRI scans from 876 participants in the Alzheimer's Disease Neuroimaging Initiative dataset and held out a separate test set of 1283 MRI scans from 170 participants for quantitative and qualitative assessment of the personalised time series of synthetic images. We performed three evaluations: i) image quality assessment; ii) quantifying the accuracy of regional brain volumes over and above benchmark models; and iii) quantifying visual perception of the synthetic images by medical experts. Overall, both quantitative and qualitative results show that 4D-DANI-Net produces realistic, low-artefact, personalised time series of synthetic T1 MRI that outperforms benchmark models.

Eye-tracking technology is an innovative tool that holds promise for enhancing dementia screening. In this work, we introduce a novel way of extracting salient features directly from the raw eye-tracking data of a mixed sample of dementia patients during a novel instruction-less cognitive test. Our approach is based on self-supervised representation learning where, by training initially a deep neural network to solve a pretext task using well-defined available labels (e.g. recognising distinct cognitive activities in healthy individuals), the network encodes high-level semantic information which is useful for solving other problems of interest (e.g. dementia classification). Inspired by previous work in explainable AI, we use the Layer-wise Relevance Propagation (LRP) technique to describe our network's decisions in differentiating between the distinct cognitive activities. The extent to which eye-tracking features of dementia patients deviate from healthy behaviour is then explored, followed by a comparison between self-supervised and handcrafted representations on discriminating between participants with and without dementia. Our findings not only reveal novel self-supervised learning features that are more sensitive than handcrafted features in detecting performance differences between participants with and without dementia across a variety of tasks, but also validate that instruction-less eye-tracking tests can detect oculomotor biomarkers of dementia-related cognitive dysfunction. This work highlights the contribution of self-supervised representation learning techniques in biomedical applications where the small number of patients, the non-homogenous presentations of the disease and the complexity of the setting can be a challenge using state-of-the-art feature extraction methods.

In recent years, endomicroscopy has become increasingly used for diagnostic purposes and interventional guidance. It can provide intraoperative aids for real-time tissue characterization and can help to perform visual investigations aimed for example to discover epithelial cancers. Due to physical constraints on the acquisition process, endomicroscopy images, still today have a low number of informative pixels which hampers their quality. Post-processing techniques, such as Super-Resolution (SR), are a potential solution to increase the quality of these images. SR techniques are often supervised, requiring aligned pairs of low-resolution (LR) and high-resolution (HR) images patches to train a model. However, in our domain, the lack of HR images hinders the collection of such pairs and makes supervised training unsuitable. For this reason, we propose an unsupervised SR framework based on an adversarial deep neural network with a physically-inspired cycle consistency, designed to impose some acquisition properties on the super-resolved images. Our framework can exploit HR images, regardless of the domain where they are coming from, to transfer the quality of the HR images to the initial LR images. This property can be particularly useful in all situations where pairs of LR/HR are not available during the training. Our quantitative analysis, validated using a database of 238 endomicroscopy video sequences from 143 patients, shows the ability of the pipeline to produce convincing super-resolved images. A Mean Opinion Score (MOS) study also confirms this quantitative image quality assessment.

Recent studies have shown that air pollution has a negative impact on people's health, especially for patients with respiratory and cardiac diseases (e.g. COPD, asthma, ischemic heart disease). Although there are already many air quality monitoring stations in major cities, such as London, these stations are sparsely located, and the periodic collection of information is insufficient to provide the granularity needed to assess the environmental risk for an individual (e.g. to avoid exacerbation). Wearable devices, on the other hand, are more suitable in this context, providing a better estimation of the air quality in the proximity of the person. Therefore, relevant warnings and information on health risks can be provided in real-time. As a proof of concept, we have developed a wearable sensor for continuous monitoring of air quality around the user, and a preliminary study was conducted to validate the sensor and assess the air quality in London underground stations. Based on the PM2.5 (particulate matter with a diameter of 2.5 μm), temperature and location information, a model is generated for predicting the air quality of each station at different times. Our preliminary results have shown that there are significant differences in air quality among stations and metro lines. It also demonstrates that wearable sensors can provide necessary information for users to make travel arrangements that minimize their exposure to polluted air.

With a massive influx of multimodality data, the role of data analytics in health informatics has grown rapidly in the last decade. This has also prompted increasing interests in the generation of analytical, data driven models based on machine learning in health informatics. Deep learning, a technique with its foundation in artificial neural networks, is emerging in recent years as a powerful tool for machine learning, promising to reshape the future of artificial intelligence. Rapid improvements in computational power, fast data storage, and parallelization have also contributed to the rapid uptake of the technology in addition to its predictive power and ability to generate automatically optimized high-level features and semantic interpretation from the input data. This article presents a comprehensive up-to-date review of research employing deep learning in health informatics, providing a critical analysis of the relative merit, and potential pitfalls of the technique as well as its future outlook. The paper mainly focuses on key applications of deep learning in the fields of translational bioinformatics, medical imaging, pervasive sensing, medical informatics, and public health.

The increasing popularity of wearable devices in recent years means that a diverse range of physiological and functional data can now be captured continuously for applications in sports, wellbeing, and healthcare. This wealth of information requires efficient methods of classification and analysis where deep learning is a promising technique for large-scale data analytics. While deep learning has been successful in implementations that utilize high-performance computing platforms, its use on low-power wearable devices is limited by resource constraints. In this paper, we propose a deep learning methodology, which combines features learned from inertial sensor data together with complementary information from a set of shallow features to enable accurate and real-time activity classification. The design of this combined method aims to overcome some of the limitations present in a typical deep learning framework where on-node computation is required. To optimize the proposed method for real-time on-node computation, spectral domain preprocessing is used before the data are passed onto the deep learning framework. The classification accuracy of our proposed deep learning approach is evaluated against state-of-the-art methods using both laboratory and real world activity datasets. Our results show the validity of the approach on different human activity datasets, outperforming other methods, including the two methods used within our combined pipeline. We also demonstrate that the computation times for the proposed method are consistent with the constraints of real-time on-node processing on smartphones and a wearable sensor platform.

Recent advances in hyperspectral imaging have made it a promising solution for intra-operative tissue characterization, with the advantages of being non-contact, non-ionizing, and non-invasive. Working with hyperspectral images in vivo, however, is not straightforward as the high dimensionality of the data makes real-time processing challenging. In this paper, a novel dimensionality reduction scheme and a new processing pipeline are introduced to obtain a detailed tumor classification map for intra-operative margin definition during brain surgery. However, existing approaches to dimensionality reduction based on manifold embedding can be time consuming and may not guarantee a consistent result, thus hindering final tissue classification. The proposed framework aims to overcome these problems through a process divided into two steps: dimensionality reduction based on an extension of the T-distributed stochastic neighbor approach is first performed and then a semantic segmentation technique is applied to the embedded results by using a Semantic Texton Forest for tissue classification. Detailed in vivo validation of the proposed method has been performed to demonstrate the potential clinical value of the system.

Human Activity Recognition provides valuable contextual information for wellbeing, healthcare, and sport applications. Over the past decades, many machine learning approaches have been proposed to identify activities from inertial sensor data for specific applications. Most methods, however, are designed for offline processing rather than processing on the sensor node. In this paper, a human activity recognition technique based on a deep learning methodology is designed to enable accurate and real-time classification for low-power wearable devices. To obtain invariance against changes in sensor orientation, sensor placement, and in sensor acquisition rates, we design a feature generation process that is applied to the spectral domain of the inertial data. Specifically, the proposed method uses sums of temporal convolutions of the transformed input. Accuracy of the proposed approach is evaluated against the current state-of-the-art methods using both laboratory and real world activity datasets. A systematic analysis of the feature generation parameters and a comparison of activity recognition computation times on mobile devices and sensor nodes are also presented.

Assessment of food intake has a wide range of applications in public health and life-style related chronic disease management. In this paper, we propose a real-time food recognition platform combined with daily activity and energy expenditure estimation. In the proposed method, food recognition is based on hierarchical classification using multiple visual cues, supported by efficient software implementation suitable for realtime mobile device execution. A Fischer Vector representation together with a set of linear classifiers are used to categorize food intake. Daily energy expenditure estimation is achieved by using the built-in inertial motion sensors of the mobile device. The performance of the vision-based food recognition algorithm is compared to the current state-of-the-art, showing improved accuracy and high computational efficiency suitable for realtime feedback. Detailed user studies have also been performed to demonstrate the practical value of the software environment.

Content-aware image resizing techniques allow to take into account the visual content of images during the resizing process. The basic idea beyond these algorithms is the removal of vertical and/or horizontal paths of pixels (i.e., seams) containing low salient information. In this paper, we present a method which exploits the gradient vector flow (GVF) of the image to establish the paths to be considered during the resizing. The relevance of each GVF path is straightforward derived from an energy map related to the magnitude of the GVF associated to the image to be resized. To make more relevant, the visual content of the images during the content-aware resizing, we also propose to select the generated GVF paths based on their visual saliency properties. In this way, visually important image regions are better preserved in the final resized image. The proposed technique has been tested, both qualitatively and quantitatively, by considering a representative data set of 1000 images labeled with corresponding salient objects (i.e., ground-truth maps). Experimental results demonstrate that our method preserves crucial salient regions better than other state-of-the-art algorithms.

Content-aware image resizing is an effective technique that allows to take into account the visual content of images during the resizing process. The basic idea beyond these algorithms is the resizing of an image by considering vertical and/or horizontal paths of pixels (i.e., seams) which contain low salient information. In this paper we exploit the Gradient Vector Flow (GVF) of the image to establish the paths to be considered during the resizing. The relevance of each path is derived from a saliency map obtained by considering the magnitude of the GVF associated to the image under consideration. The proposed technique has been tested, both qualitatively and quantitatively, by considering a representative set of images labeled with corresponding salient objects (i.e., ground-truth maps). Experimental results demonstrate that our method preserves crucial salient regions better than other state-of-the-art algorithms.

Red-eye artifact is a well-known problem in digital photography. Since the large diffusion of mobile devices with embedded camera and flashgun, automatic detection and correction of red-eyes have become an important task. In this paper we describe a technique that makes use of three steps to identify and correct red-eyes. First, red-eye candidates are extracted from the input image by using simple color segmentation coupled with geometrical constraints. A set of linear discriminant classifiers is then learned on the clustered patches space, and hence employed to distinguish between eyes and non-eyes patches. The proposed cluster-based Linear Discriminant Analysis is used to deal with the multi-modally nature of the input space. The third step of the pipeline is devoted to artifacts correction through de-saturation and brightness reduction. Experimental results on a large dataset of images demonstrate the effectiveness of the pro- posed pipeline that outperforms other existing solutions in terms of hit rates maximization, false positives reduction and ad-hoc quality measure.

Since the large diffusion of digital camera and mobile devices with embedded camera and flashgun, the red-eyes artifacts have de-facto become a critical problem. The technique herein described makes use of three main steps to identify and remove red-eyes. First, red eyes candidates are extracted from the input image by using an image filtering pipeline. A set of classifiers is then learned on gray code features extracted in the clustered patches space, and hence employed to distinguish between eyes and non-eyes patches. Once red-eyes are detected, artifacts are removed through desaturation and brightness reduction. The proposed method has been tested on large dataset of images achieving effective results in terms of hit rates maximization, false positives reduction and quality measure.

This paper proposes a method to recognize scene categories using bags of visual words obtained hierarchically partitioning into subregion the input images. Specifically, for each subregions the texton histogram and the extension of the sub-region is taken into account. The bags of visual words, obtained in this way, are weighted and used in a similarity measure during the categorization. Experimental tests using ten different scene categories show that the proposed approach achieves good performances with respect to the state of the art methods.