Mulham Fawakherji

Early and accurate identification of malignant pulmonary nodules is essential to reduce lung cancer mortality, yet segmentation of these nodules in CT scans remains challenging due to their small size, variable morphology, low contrast, and extreme class imbalance. To address these challenges, we propose a privacy-preserving federated learning framework that integrates a hybrid U-Net–Transformer architecture with a customized Dice-Focal loss function. The hybrid model combines residual convolutional blocks for local texture feature extraction with Transformer-based self-attention for capturing global contextual dependencies, while skip connections, Layer Normalization, and Dropout enhance gradient flow and generalization across heterogeneous client datasets. The Dice-Focal loss mitigates class imbalance and improves boundary sensitivity, yielding stable convergence and precise segmentation of solid nodules. Using the LUNA16 dataset, our federated setup demonstrated high segmentation fidelity for nodules between 15mm and 25mm, achieving Dice coefficients up to 0.93 and robust precision-recall trade-offs across clients. These results highlight the framework’s potential for scalable, privacy-preserving lung cancer screening, while also identifying challenges in segmenting part-solid or low-contrast nodules for future methodological improvements.

Autism Spectrum Disorder (ASD) is a neurodevelopmental condition whose diagnosis relies heavily on subjective behavioral assessments. Electroencephalography (EEG) provides an inexpensive and non-invasive method of objective biomarker detection but is plagued by high dimensionality and noise that complicates automated analysis. This article introduces a hybrid Transformer–Quantum Neural Network (QNN) model that combines deep learning and quantum computing to enhance P300 targeted ASD responses from non-targeted responses. The model uses transformer capabilities captures cross-domain correlations and feature-level dependencies across heterogeneous feature sources and the quantum heads introduced via parameterized quantum circuits detects high-order, nonlinear relations in a Hilbert space via entanglement and superposition. Experimental results on the BCIAUT-P300 dataset better classical alternatives with higher diagnostic accuracy at higher accuracy (0.921) and F1-score (0.915). Statistical comparisons confirm that the hybrid approach yields significant improvements in discriminative reliability and diagnostic odds ratio, justifying quantum feature embedding’s expressivity and efficiency. The results bear witness to the feasibility and potential of hybrid quantum–classical systems as potential future intelligent diagnostic devices for ASD responses from EEG signals.

Floods are among the most destructive and costly natural disasters worldwide, necessitating accurate and timely mapping for emergency response and impact assessment. Remote sensing offers a valuable means to monitor inundation, yet trade-offs in spatial, spectral, and temporal resolution limit single-sensor approaches. This study presents a comparative analysis of flood mapping using multimodal sensor fusion and semantic segmentation. We integrate high-resolution UAV RGB imagery with Synthetic Aperture Radar (SAR) data from Sentinel-1 and UAVSAR, focusing on the classification of four land cover classes relevant to flooding: inundated vegetation, dry vegetation, open water, and others. A modified U-Net architecture was developed to handle multimodal stacked inputs, allowing systematic evaluation of different sensor combinations. The results show that the combination of RGB UAV with UAVSAR VV polarization at 5 m resolution achieves the highest performance (mean IoU: 84. 9%), outperforming RGB UAV alone (77. 8%) and significantly surpassing the combinations of higher resolution UAVSAR (1 m, 61. 2%) and Sentinel-1. These findings highlight the complementary strengths of optical and SAR features, the importance of contextual information, and the counterintuitive benefit of medium-resolution UAVSAR for scene-level understanding. Incorporation of SAR indices such as VH/VV ratios provided modest gains, particularly in vegetation-rich areas. Cross-domain testing with Hurricane Matthew imagery revealed a notable drop in generalization (best mean IoU: 46.2%), underscoring challenges of domain shift and the need for domain-adaptive modeling. Overall, the results emphasize the advantages of multimodal fusion, especially UAV RGB with UAVSAR VV (5 m), for fine-grained flood mapping and suggest future directions in multi-temporal and adaptive frameworks for robust flood monitoring.

Automating agricultural processes holds significant promise for enhancing efficiency and sustainability in various farming practices. This paper contributes to the automation of agricultural processes by providing a dedicated mushroom detection dataset related to automated harvesting, 3D pose estimation, and growth monitoring of the button mushroom produced using Agaricus Bisporus fungi. With a total of 2000 images for object detection, instance segmentation, and 3D pose estimation—containing over 100,000 mushroom instances—and an additional 3838 images for yield estimation featuring eight mushroom scenes covering the complete growth period, it fills the gap in mushroom-specific datasets and serves as a benchmark for detection and instance segmentation as well as 3D pose estimation algorithms in smart mushroom agriculture. The dataset, featuring realistic growth environment scenarios with comprehensive 2D and 3D annotations, is assessed using advanced detection and instance segmentation algorithms. This paper details the dataset’s characteristics, presents detailed statistics on mushroom growth and yield, evaluates algorithmic performance, and, for broader applicability, makes all resources publicly available, including images, code, and trained models, via our GitHub repository. (accessed on 22 March 2025).

Automating agricultural processes holds significant promise for enhancing efficiency and sustainability in various farming practices. This paper contributes to the automation of agricultural processes by providing a dedicated mushroom detection dataset related to automated harvesting, 3D pose estimation, and growth monitoring of the button mushroom produced using Agaricus Bisporus fungus. With a total of 2,000 images for object detection, instance segmentation, and 3D pose estimation containing over 100,000 mushroom instances and an additional 3,838 images for yield estimation containing 8 mushroom scenes covering the complete growth period, it fills the gap in mushroom-specific datasets and serves as a benchmark for detection and instance segmentation as well as 3D pose estimation algorithms in smart mushroom agriculture. The dataset, featuring realistic growth environment scenarios with comprehensive 2D and 3D annotations, is assessed using advanced detection and instance segmentation algorithms. The paper details the dataset’s characteristics, presents the detailed statistics of mushroom growth and yield, evaluates algorithmic performance, and for broader applicability, we have made all resources publicly available including images, codes, and trained models via our GitHub repository: https://github.com/abdollahzakeri/m18k

The use of deep learning methods for precision farming is gaining increasing interest. However, collecting training data in this application field is particularly challenging and costly due to the need of acquiring information during the different growing stages of the cultivation of interest. In this paper, we present a method for data augmentation that uses two GANs to create artificial images to augment the training data. To obtain a higher image quality, instead of re-creating the entire scene, we take original images and replace only the patches containing objects of interest with artificial ones containing new objects with different shapes and styles. In doing this, we take into account both the foreground (i.e., crop samples) and the background (i.e., the soil) of the patches. Quantitative experiments, conducted on publicly available datasets, demonstrate the effectiveness of the proposed approach. The source code and data discussed in this work are available as open source.

Oral squamous cell carcinoma is the most common oral cancer. In this paper, we present a performance analysis of four different deep learning-based pixel-wise methods for lesion segmentation on oral carcinoma images. Two diverse image datasets, one for training and another one for testing, are used to generate and evaluate the models used for segmenting the images, thus allowing to assess the generalization capability of the considered deep network architectures. An important contribution of this work is the creation of the Oral Cancer Annotated (ORCA) dataset, containing ground-truth data derived from the well-known Cancer Genome Atlas (TCGA) dataset.