Saeed S. Alahmari

The use of deep learning for food recognition has attracted significant attention due to its applications in dietary monitoring, automated food logging, and nutrition analysis. While deep learning has demonstrated impressive results across various cuisines, some, including Saudi Arabian cuisine, have not been thoroughly explored. This paper examines the effectiveness of deep learning models in accurately classifying and recognizing thirteen different Saudi Arabian traditional food dishes. We applied convolutional neural networks (CNNs) and transformer-based architectures, taking advantage of large pre-trained models for fine-tuning on our dataset. Additionally, we have curated the first dataset of traditional Saudi Arabian food, consisting of 13 categories and 3,000 images. Experimental results show that deep learning models achieve high accuracy in distinguishing among diverse food categories, even under challenging conditions such as occlusions and varying lighting. Our curated dataset is available publicly at https://huggingface.co/datasets/ssalahmari/Saudi_Traditional_Food.

Background: This paper investigates the use of the Segment Anything Model (SAM) for chest X-ray (CXR) image segmentation, with a focus on improving its performance using low-rank adaptation (LoRA). Methods: We evaluate three versions of SAM: two zero-shot methods (using coordinate and bounding box prompts) and a fine-tuned SAM using LoRA. To support these approaches, we also trained two standard convolutional neural networks (CNNs), U-Net and DeepLabv3+, to generate draft lung segmentations that serve as input prompts for the SAM methods. Our fine-tuning approach uses LoRA to add lightweight trainable adapters within the Transformer blocks of the SAM, allowing only a small subset of parameters to be updated. The rest of the SAM remains frozen, helping preserve its pre-trained knowledge while reducing memory and computational needs. We tested all models on a dataset of CXR images labeled for COVID-19, viral pneumonia, and normal cases. Results: Results show that fine-tuned SAM with LoRA outperforms zero-shot SAM methods and CNN baselines in terms of segmentation accuracy and efficiency. Conclusions: This demonstrates the potential of combining LoRA with SAM for practical and effective medical image segmentation.

Large Language Models (LLMs) have demonstrated impressive performance across various natural language processing (NLP) tasks, including text summarization, classification, and generation. Despite their success, LLMs are primarily trained on curated datasets that lack human-induced errors, such as typos or variations in word choice. As a result, LLMs may produce unexpected outputs when processing text containing such perturbations. In this paper, we investigate the resilience of LLMs to two types of text perturbations: typos and word substitutions. Using two public datasets, we evaluate the impact of these perturbations on text generation using six state-of-the-art models, including GPT-4o and LLaMA3.3-70B. Although previous studies have primarily examined the effects of perturbations in classification tasks, our research focuses on their impact on text generation. The results indicate that LLMs are sensitive to text perturbations, leading to variations in generated outputs, which have implications for their robustness and reliability in real-world applications.

Recent advances in foundation models have enabled strong zero-shot performance across vision tasks, yet their effectiveness for fine-grained domains such as food image segmentation remains underexplored. This study investigates the zero-shot capabilities of the Segment Anything Model (SAM) for segmenting food images. We evaluate two prompting strategies—bounding boxes and coordinate points—and compare SAM’s results against established deep learning baselines, U-Net and DeepLabv3+, both trained on the FoodBin-17k dataset with pixel-wise binary annotations. Using bounding box prompts, zero-shot SAM achieved a Dice coefficient of 0.790, closely matching U-Net (0.759), while DeepLabv3+ attained 0.918. Coordinate-based prompting yielded comparable results, with the optimal configuration (75 coordinate points) also reaching a Dice coefficient of 0.790. These results highlight SAM’s strong potential for generalizing to novel visual categories without task-specific retraining. This study further provides insights into how prompt type influences segmentation quality, offering practical guidance for applying SAM in food-related and other fine-grained segmentation applications.

Large Language Models (LLMs) have demonstrated impressive performance across various natural language processing (NLP) tasks, including text summarization, classification, and generation. Despite their success, LLMs are primarily trained on curated datasets that lack human-induced errors, such as typos or variations in word choice. As a result, LLMs may produce unexpected outputs when processing text containing such perturbations. In this paper, we investigate the resilience of LLMs to two types of text perturbations: typos and word substitutions. Using two public datasets, we evaluate the impact of these perturbations on text generation using six state-of-the-art models, including GPT-4o and LLaMA3.3-70B. Although previous studies have primarily examined the effects of perturbations in classification tasks, our research focuses on their impact on text generation. The results indicate that LLMs are sensitive to text perturbations, leading to variations in generated outputs, which have implications for their robustness and reliability in real-world applications.

Molecular assays can resolve clonal structure, but they are expensive and typically sparse in time, whereas phenotypic observations such as imaging can be collected frequently but often are not preserved in the context needed for later interpretation. We present CLONEID , an event-based framework for organizing clone-resolved phenotypic, molecular, and specimen-context records so that genotype-to-phenotype interpretation can be maintained across time. CLONEID links time-stamped Events, assay-specific Perspectives, and reconciled Identities through structured ingestion, provenance-aware retrieval, and reproducible export, complementing upstream clone-calling methods. In a long-term gastric cancer density-selection experiment, CLONEID linked repeated culture events, growth measurements, and late karyotypic profiling within a shared record, supporting longitudinal interpretation of phenotypic adaptation together with underlying chromosomal state.

Several methods for cell cycle inference from sequencing data exist and are widely adopted. In contrast, methods for classification of cell cycle state from imaging data are scarce. We have for the first time integrated sequencing and imaging derived cell cycle pseudo-times for assigning 449 imaged cells to 693 sequenced cells at an average resolution of 3.4 and 2.4 cells for sequencing and imaging data respectively. Data integration revealed thousands of pathways and organelle features that are correlated with each other, including several previously known interactions and novel associations. The ability to assign the transcriptome state of a profiled cell to its closest living relative, which is still actively growing and expanding opens the door for genotype-phenotype mapping at single cell resolution forward in time.