Urban vulnerability assessment is crucial for understanding the spatial distribution of deprived areas and associated risks. Slum residents face significantly worse health outcomes than non-slum urban populations, with neighborhood effects being critical in social epidemiology. Identifying such areas is vital because they present public health challenges that climate change and increased air pollution can exacerbate. Accordingly, this study proposed an AI-driven methodology that integrates remote sensing data, socioeconomic indicators, and machine learning algorithms to identify and analyze vulnerable areas in Brazil. To create a vulnerability index, we incorporate multiple data sources, including Sentinel-2 and Sentinel-5P imagery, Brazilian socioeconomic indicators, and OpenStreetMap. Hence, we predicted pollution indicators using regression algorithms such as Random Forest, XGBoost, and Linear Regression. Our findings demonstrate that integrating multi-source data is a promising approach for better understanding deprived areas, indicating that slums (called “favelas” in Brazil) exhibit an intense concentration of the sociocconomic vulnerability index, a key determinant of deprivation. However, non-slum areas may present heterogeneous conditions, with some regions showing vulnerability levels comparable to those of slums while others show better conditions. Our results highlight the potential of AIdriven approaches for urban vulnerability assessment, offering insights for policymakers and researchers.

Traditional Deep Learning (DL) approaches for medical image classification rely on centralized, static datasets, which do not adequately reflect the dynamic, real-world medical practice where data is continually generated. In contrast, Federated Learning (FL) enables decentralized model training on localized data while preserving privacy. Yet, current FL methods struggle to handle Non-Identically Independently Distributed (Non-IID) data streams over time. This paper introduces Med-TimeSplit, a novel dataset partitioning strategy that integrates Online Continual Learning (OCL) with FL to simulate real-world medical data flows more realistically and effectively. MedTimeSplit partitions data into Non-IID, time-based increments, mimicking dynamic sourcing in medical environments. We evaluate its impact on FL model performance for medical image classification, focusing on skin lesions, and analyze the system’s resilience to backdoor attacks. Our experiments demonstrate that MedTimeSplit outperforms existing methods in both accuracy and robustness, offering a viable solution for real-world medical applications. Additionally, we propose new metrics to measure model behavior over time, including Average Bad Decisions (ABD) and Overall Changing Mistakes (OCM), which provide deeper insights into model performance specifically under OCL conditions. The results highlight the promise of combining OCL with FL in the medical domain, paving the way for more secure and adaptive healthcare solutions. The source code is available on GitHub1.

Medical image analysis is an important asset in the clinical process, providing resources to assist physicians in detecting diseases and making accurate diagnoses. Deep Learning (DL) models have been widely applied in these tasks, improving the ability to recognize patterns, including accurate and fast diagnosis. However, DL can present issues related to security violations that reduce the system’s confidence. Uncovering these attacks before they happen and visualizing their behavior is challenging. Current solutions are limited to binary analysis of the problem, only classifying the sample into attacked or not attacked. In this paper, we propose the RADAR-MIX framework for uncovering adversarial attacks using quantitative metrics and analysis of the attack’s behavior based on visual analysis. The RADAR-MIX provides a framework to assist practitioners in checking the possibility of adversarial examples in medical applications. Our experimental evaluation shows that the Deep-Fool and Carlini & Wagner (CW) attacks significantly evade the ResNet50V2 with a slight noise level of 0.001. Furthermore, our results revealed that the gradient-based methods, such as Gradient-weighted Class Activation Mapping (Grad-CAM) and SHapley Additive exPlanations (SHAP), achieved high attack detection effectiveness. While Local Interpretable Model-agnostic Explanations (LIME) presents low consistency, implying the most ability to uncover robust attacks supported by visual analysis.

Deep Learning (DL) is a valuable set of techniques that improve medical decision-making based on imaging exams, such as Chest X-rays (CXR), Computed Tomography (CT), and Optical Coherence Tomography (OCT). However, DL models may be susceptible to adversarial attacks when perturbed (tam-pered) examples sneak into the data, decreasing the model's confidence. In this paper, we evaluate the vulnerabilities of DL applied to medical images and analyze the effects of attacks on the Gradient-weighted Class Activation Mapping (GRAD-CAM). Our experiments were conducted on two scenarios: (i) CXR images with binary class; (ii) OCT images with multi-class. Vulnerabilities are described by Fooling Rate (FR) and visual analysis of Grad-CAM. We show that the PGD is the most malicious deed for multi-class, reaching an FR of up to 96%, whereas DeepFool is hurtful for binary classes, reaching an FR of up to 93%. Our analysis can be used to understand the adversarial attacks over medical images and their effects on explainability. The developed code is available at GitHub11https://github.com/eriksonJAguiar/Grad-Attacks-CBMS-2023.

Medical image analysis plays a major role in aiding physicians in decision-making. Specifically in detecting COVID-19, Deep Learning (DL) and radiomic approaches have achieved promising results separately. However, DL results are hard to interpret/visualize, and the radiomic approach encompasses successive steps, such as image acquisition, image processing, segmentation, feature extraction, and analysis. In this paper, we integrate DL with radiomic approaches, aiding in detecting COVID-19. We use DL models to extract 128 relevant deep radiomic features to assess COVID-19 from several image sources of 392 representative chest X-ray (CXR) exams. We avoid successive radiomic steps by employing DL (transfer learning) from Imagenet's VGG-16, ResNet50V2, and DenseNet201 networks. We considered a set of Machine Learning (ML) algorithms to further validate our results, providing an ensemble model to detect COVID-19. Our experimental results show that our approach achieved 95% AUC using 128 relevant features from DenseNet201. Conversely, our ensemble model presented 91% AUC, indicating that deep learning-based radiomics could increase binary classification performance in a real scenario. In addition, we highlight that our approach can be adapted to create other DL-based radiomics tools. For reproducibility, we made our code available at https://github.com/usmarcv/CBMS-DL-based-radiomics.

Bone mineral density (BMD) is the international standard for evaluating osteoporosis/osteopenia. The success rate of BMD alone in estimating the risk of vertebral fragility fracture (VFF) is approximately 50%, making BMD far from ideal in predicting VFF. In addition, whether or not a patient has been diagnosed with osteoporosis or osteopenia, he or she may suffer a VFF. For this reason, we conducted an extensive empirical study to assess VFFs in postmenopausal women. We considered a representative dataset of 94 T1- and T2-weighted routine spine MRI (with osteopenia or osteoporosis), split into 2,400 samples (slices). Comparing the classification results of machine learning and deep learning (DL) techniques showed that DL generally achieved better results at the cost of higher computational power and hard explainability. ResNet achieved the best results in discriminating patients from groups with and without VFFs with 83% accuracy and 90% AUC (with a confidence interval of 99%). Our results represent a significant step toward prospective and longitudinal studies investigating methods to achieve higher accuracy in predicting VFFs based on spine MRI features of vertebrae without fracture.