Rahimeh Rouhi

The infant hippocampus plays a pivotal role in early brain development and is linked to cognitive and memory functions. Accurate delineation of the hippocampus is essential for studying normal brain development and detecting early abnormalities associated with various neurodevelopmental disorders. In this paper, different deep neural network models were trained for 3D-automatic segmentation of the hippocampus based on cross-validation on a cohort of T1-Weighted (T1W) images acquired from 100 subjects with ground truth. The models were tested on another image cohort of 86 subjects without ground truth. Ensembling the single-trained models during cross-validation resulted in the final segmentation. Among all the trained networks, nnUNet and SegResNet achieved the best average Dice Similarity Coefficient (DSC)=0.82±0.01 and 95th Hausdorff Distance (95HD)=2.45±1.10 mm, respectively, in 5-fold cross-validation. We presented a comprehensive comparison between different architectures in terms of their generalizability and effectiveness, suggesting the potential for developing on-the-fly automated segmentation of the hippocampus in pediatric MRI.

Given that a significant fraction of buildings and architectural heritage in Europe’s historical centers are masonry structures, the selection of proper diagnosis, technological surveys, non-destructive testing, and interpretations of crack and decay patterns is paramount for a risk assessment of possible damage. Identifying the possible crack patterns, discontinuities, and associated brittle failure mechanisms within unreinforced masonry under seismic and gravity actions allows for reliable retrofitting interventions. Traditional and modern materials and strengthening techniques create a wide range of compatible, removable, and sustainable conservation strategies. Steel/timber tie-rods are mainly used to support the horizontal thrust of arches, vaults, and roofs and are particularly suitable for better connecting structural elements, e.g., masonry walls and floors. Composite reinforcing systems using carbon, glass fibers, and thin mortar layers can improve tensile resistance, ultimate strength, and displacement capacity to avoid brittle shear failures. This study overviews masonry structural diagnostics and compares traditional and advanced strengthening techniques of masonry walls, arches, vaults, and columns. Several research results in automatic surface crack detection for unreinforced masonry (URM) walls are presented considering crack detection based on machine learning and deep learning algorithms. In addition, the kinematic and static principles of Limit Analysis within the rigid no-tension model framework are presented. The manuscript sets a practical perspective, providing an inclusive list of papers describing the essential latest research in this field; thus, this paper is useful for researchers and practitioners in masonry structures.

Atrial Fibrillation (AF) is the most common type of cardiac arrhythmia. Early diagnosis of AF helps to improve therapy and prognosis. Machine Learning (ML) has been successfully applied to improve the effectiveness of Computer-Aided Diagnosis (CADx) systems for AF detection. Presenting an explanation for the decision made by an ML model is considerable from the cardiologists' point of view, which decreases the complexity of the ML model and can provide tangible information in their diagnosis. In this paper, a range of explanation techniques is applied to hand-crafted features based ML models for heart rhythm classification. We validate the impact of the techniques by applying feature selection and classification to the 2017 CinC/PhysioNet challenge dataset. The results show the effectiveness and efficiency of SHapley Additive exPlanations (SHAP) technique along with Random Forest (RF) for the classification of the Electrocardiogram (ECG) signals for AF detection with a mean F-score of 0.746 compared to 0.706 for a technique based on the same features based on a cascaded SVM approach. The study also highlights how this interpretable hand-crafted feature-based model can provide cardiologists with a more compact set of features and tangible information in their diagnosis.

The popularity of social networks (SNs), amplified by the ever-increasing use of smartphones, has intensified online cybercrimes. This trend has accelerated digital forensics through SNs. One of the areas that has received lots of attention is camera fingerprinting, through which each smartphone is uniquely characterized. Hence, in this paper, we compare classification-based methods to achieve smartphone identification (SI) and user profile linking (UPL) within the same or across different SNs, which can provide investigators with significant clues. We validate the proposed methods by two datasets, our dataset and the VISION dataset, both including original and shared images on the SN platforms such as Google Currents, Facebook, WhatsApp, and Telegram. The obtained results show that k-medoids achieves the best results compared with k-means, hierarchical approaches, and different models of convolutional neural network (CNN) in the classification of the images. The results show that k-medoids provides the values of F1-measure up to 0.91% for SI and UPL tasks. Moreover, the results prove the effectiveness of the methods which tackle the loss of image details through the compression process on the SNs, even for the images from the same model of smartphones. An important outcome of our work is presenting the inter-layer UPL task, which is more desirable in digital investigations as it can link user profiles on different SNs.

Clustering the images shared through Social Network (SN) platforms according to the acquisition cameras embedded in smartphones is regarded as a significant task in forensic investigations of cybercrimes. The Sensor Pattern Noise (SPN) caused by camera sensor imperfections during the manufacturing process can be extracted from the images and used to fingerprint the smartphones. The process of content compression performed by SNs causes loss of image details and weakens the SPN, making the clustering task even more challenging. In this paper, we present a hybrid algorithm capable of clustering the images captured and shared through SNs without prior knowledge about the types and number of the acquisition smartphones. The hybrid method exploits batch partitioning, image resizing, hierarchical and graph-based clustering approaches to cluster the images. Using Markov clustering, the hierarchical clustering is conducted in such a way that the representative clusters with a higher probability of belonging to the same camera are selected for merging, which accelerates the clustering. For merging the clusters, the adaptive threshold updated iteratively through the hybrid clustering is used, which results in more precise clusters even for images from the same model of smartphones. The results on the VISION dataset, including both native and shared images, prove the effectiveness and efficiency of the hybrid method in comparison with the state-of-the-art SPN-based image clustering algorithms.