Esraa Sabeeh
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Esraa Sabeeh has a total of 6 co-authors. Here are the top co-authors by number of shared publications:
- Mohammed Zuhair Al-Taie
Faculty of Computing, Universiti Teknologi Malaysia, Johor Bahru, Malaysia
4 shared publications - Akram Qashou
Anglia Ruskin University, Cambridge, UK
1 shared publication Faten AbdulKareem Al-Azzaw
Qalat Saleh General Hospital, Laboratories Department, Bacterial Culture Unit, Misan, Iraq
1 shared publication- Firas Hazzaa
Anglia Ruskin University, Cambridge, UK
1 shared publication Heba Salah Khaled
Ministry of Education, Misan Education Directorate, Misan, Iraq
1 shared publicationKarrar Chasib Badr
Qalat Saleh General Hospital, Misan, Iraq
1 shared publication
Latest Publications
- Open Access
- 25 September 2026
- Published by Springer Nature in BMC Bioinformatics
RMGAT: a Relational Multi-Scale Graph Attention Network for spatial transcriptomics domain identification with systematic component ablation
Background: Mapping spatially coherent tissue domains from spatial transcriptomics data is a prerequisite for characterizing cell-type, developmental patterning, and disease-associated disruptions of tissue architecture. While graph neural network (GNN) methods have substantially improved domain identification over expression-only clustering, the effect of edge feature design (i.e. how spatial connections between tissue spots are represented) has never been systematically evaluated. All existing methods use a single scalar edge weight, leaving open the question of whether multi-dimensional edge features provide any benefit. Methods: We introduce RMGAT, a Relational Multi-Scale Graph Attention Network that applies multi-dimensional learned edge features to spatial transcriptomics, to the best of our knowledge for the first time. Each spatial edge is described by an 8-dimensional learned feature vector. We combine this with a three-component training objective (graph reconstruction, NT-Xent contrastive learning, and a self-expression decoder) and a V30 post-processing pipeline using consensus clustering, Hungarian label alignment, and spatial refinement. Results: Evaluated on the 12-section DLPFC (human dorsolateral prefrontal cortex) benchmark across 60 independent runs (5 seeds × 12 DLPFC sections), RMGAT achieves a mean adjusted Rand index (ARI) of 0.3422 ± 0.0402 (95% confidence interval [0.332, 0.353]), comparable to SpaGCN (~ 0.36). A systematic four-component ablation study with formal statistical testing identifies GAT attention as the only statistically confirmed essential contributor (ΔARI = − 0.028 when replaced with a graph convolutional network, GCN; p < 0.0001). Contrastive learning shows a directional benefit (ΔARI = − 0.017; p = 0.024 vs. scalar-weight baseline). The 8-dimensional EdgeMLP and self-expression decoder are both neutral (ΔARI ≈ 0.004; both p > 0.3), indicating neither contributes at PCA-50 input quality. Conclusions: The primary contribution of this work is a systematic ablation study providing evidence-based design guidance for spatial transcriptomics GNNs. At PCA-50 input quality, GAT attention is the statistically confirmed critical component; contrastive learning is beneficial ( p = 0.024); multi-dimensional edge features and additional decoder objectives do not improve performance under this input regime. For practitioners, these findings support using GAT attention ( p < 0.0001) and contrastive learning ( p = 0.024, non-significant after Bonferroni correction), and investing in node feature quality over edge feature complexity.
- 25 September 2026
- Published by Springer Nature in BMC Bioinformatics
Open AccessRMGAT: a Relational Multi-Scale Graph Attention Network for spatial transcriptomics domain identification with systematic component ablation
Background: Mapping spatially coherent tissue domains from spatial transcriptomics data is a prerequisite for characterizing cell-type, developmental patterning, and disease-associated disruptions of tissue architecture. While graph neural network (GNN) methods have substantially improved domain identification over expression-only clustering, the effect of edge feature design (i.e. how spatial connections between tissue spots are represented) has never been systematically evaluated. All existing methods use a single scalar edge weight, leaving open the question of whether multi-dimensional edge features provide any benefit. Methods: We introduce RMGAT, a Relational Multi-Scale Graph Attention Network that applies multi-dimensional learned edge features to spatial transcriptomics, to the best of our knowledge for the first time. Each spatial edge is described by an 8-dimensional learned feature vector. We combine this with a three-component training objective (graph reconstruction, NT-Xent contrastive learning, and a self-expression decoder) and a V30 post-processing pipeline using consensus clustering, Hungarian label alignment, and spatial refinement. Results: Evaluated on the 12-section DLPFC (human dorsolateral prefrontal cortex) benchmark across 60 independent runs (5 seeds × 12 DLPFC sections), RMGAT achieves a mean adjusted Rand index (ARI) of 0.3422 ± 0.0402 (95% confidence interval [0.332, 0.353]), comparable to SpaGCN (~ 0.36). A systematic four-component ablation study with formal statistical testing identifies GAT attention as the only statistically confirmed essential contributor (ΔARI = − 0.028 when replaced with a graph convolutional network, GCN; p < 0.0001). Contrastive learning shows a directional benefit (ΔARI = − 0.017; p = 0.024 vs. scalar-weight baseline). The 8-dimensional EdgeMLP and self-expression decoder are both neutral (ΔARI ≈ 0.004; both p > 0.3), indicating neither contributes at PCA-50 input quality. Conclusions: The primary contribution of this work is a systematic ablation study providing evidence-based design guidance for spatial transcriptomics GNNs. At PCA-50 input quality, GAT attention is the statistically confirmed critical component; contrastive learning is beneficial ( p = 0.024); multi-dimensional edge features and additional decoder objectives do not improve performance under this input regime. For practitioners, these findings support using GAT attention ( p < 0.0001) and contrastive learning ( p = 0.024, non-significant after Bonferroni correction), and investing in node feature quality over edge feature complexity.
- 30 July 2026
- Published by IGI Global Scientific Publishing
- p. 293-332
Automated Fecal Egg Counting for Cattle Parasite Surveillance
Gastrointestinal parasite surveillance in cattle relies on fecal egg counting, a task currently done by hand in most farm and laboratory settings. This chapter presents a proposed automated workflow for egg detection and counting using a Faster R-CNN model fine-tuned on the AI-KFM 2022 benchmark dataset. The workflow is built around four core principles: checking image quality before inference, applying separate thresholds for sensitive screening and accurate counting, referring borderline results to a human reviewer, and producing traceable offline reports. At the selected operating point, screening recall is 0.979 on the primary validation subset and 1.000 on a low-burden subset; counting-stage recall is 0.894 on the primary subset and 0.750 on the low-burden subset, with referral rules designed to escalate missed cases for human review. All results are based on cattle data only. The full workflow is a design pattern requiring field pilot validation before operational use.
- 30 July 2026
- Published by IGI Global Scientific Publishing
- p. 293-332
Automated Fecal Egg Counting for Cattle Parasite Surveillance
Gastrointestinal parasite surveillance in cattle relies on fecal egg counting, a task currently done by hand in most farm and laboratory settings. This chapter presents a proposed automated workflow for egg detection and counting using a Faster R-CNN model fine-tuned on the AI-KFM 2022 benchmark dataset. The workflow is built around four core principles: checking image quality before inference, applying separate thresholds for sensitive screening and accurate counting, referring borderline results to a human reviewer, and producing traceable offline reports. At the selected operating point, screening recall is 0.979 on the primary validation subset and 1.000 on a low-burden subset; counting-stage recall is 0.894 on the primary subset and 0.750 on the low-burden subset, with referral rules designed to escalate missed cases for human review. All results are based on cattle data only. The full workflow is a design pattern requiring field pilot validation before operational use.
- 2 May 2026
- Published by Springer Nature in Acta Parasitologica
- Vol. 71 (3) , 99
Leakage-Safe ITS1 Identification of Fasciola hepatica and Fasciola gigantica with Reverse-Complement-Invariant CNN Inference
- 2 May 2026
- Published by Springer Nature in Acta Parasitologica
- Vol. 71 (3) , 99
Leakage-Safe ITS1 Identification of Fasciola hepatica and Fasciola gigantica with Reverse-Complement-Invariant CNN Inference
- 1 July 2025
- Published by Egyptian Knowledge Bank in Microbes and Infectious Diseases
Study prevalence of E. Coli-associated urinary tract infections in Qalat Saleh Hospital / Maysan Province: A cross-sectional study
Background: Urinary tract infections (UTIs) represent one of the most common reasons for community-based antibiotic prescriptions globally. In low-income countries, they contribute substantially to morbidity and mortality. Escherichia coli (E. coli) is the predominant uropathogen causing these infections, and resistance to commonly used antibiotics continues to rise, especially in under-resourced settings. Objectives: This study aimed to evaluate the impact of sex, age, and residence (urban vs. rural) on the prevalence of E. coli-associated UTIs in the Qalat Saleh District, Maysan Province, Iraq. It also sought to assess antimicrobial sensitivity and resistance patterns among isolated E. coli strains. Methods: A cross-sectional study was conducted over a six-month period from January 2024 to June 2024. A total of 316 urine samples were collected from symptomatic patients aged 1 to 70 years who presented at healthcare facilities in the district. Bacterial cultures were performed, and E. coli was identified as the causative agent in confirmed UTI cases. Antibiotic susceptibility testing was carried out using standard disc diffusion methods. Results: Out of the 316 samples, 36 cases were confirmed to be caused by E. coli. The infection rate among females was 63.89% (23/36), compared to 36.11% (13/36) in males. The prevalence of E. coli infection was significantly higher in urban areas at 80.56% (29/36) versus 19.44% (7/36) in rural areas. In terms of antibiotic sensitivity, the highest susceptibility was noted for gentamicin (27%) and imipenem (16%), while doxycycline (1%), amikacin (2%), and ciprofloxacin (6%) showed the lowest sensitivity. The highest resistance rates were observed for cefotaxime and ceftriaxone (both 36%), followed by ampicillin and amoxicillin/clavulanic acid (both 29%). Conclusion: E. coli remains the predominant cause of UTIs in the Qalat Saleh District, with a higher burden among females and urban residents. Alarmingly low antibiotic sensitivity levels were observed, especially to commonly used antibiotics, indicating a pressing need for revised empirical treatment guidelines and antibiotic stewardship efforts in the region.
- 1 July 2025
- Published by Egyptian Knowledge Bank in Microbes and Infectious Diseases
Study prevalence of E. Coli-associated urinary tract infections in Qalat Saleh Hospital / Maysan Province: A cross-sectional study
Background: Urinary tract infections (UTIs) represent one of the most common reasons for community-based antibiotic prescriptions globally. In low-income countries, they contribute substantially to morbidity and mortality. Escherichia coli (E. coli) is the predominant uropathogen causing these infections, and resistance to commonly used antibiotics continues to rise, especially in under-resourced settings. Objectives: This study aimed to evaluate the impact of sex, age, and residence (urban vs. rural) on the prevalence of E. coli-associated UTIs in the Qalat Saleh District, Maysan Province, Iraq. It also sought to assess antimicrobial sensitivity and resistance patterns among isolated E. coli strains. Methods: A cross-sectional study was conducted over a six-month period from January 2024 to June 2024. A total of 316 urine samples were collected from symptomatic patients aged 1 to 70 years who presented at healthcare facilities in the district. Bacterial cultures were performed, and E. coli was identified as the causative agent in confirmed UTI cases. Antibiotic susceptibility testing was carried out using standard disc diffusion methods. Results: Out of the 316 samples, 36 cases were confirmed to be caused by E. coli. The infection rate among females was 63.89% (23/36), compared to 36.11% (13/36) in males. The prevalence of E. coli infection was significantly higher in urban areas at 80.56% (29/36) versus 19.44% (7/36) in rural areas. In terms of antibiotic sensitivity, the highest susceptibility was noted for gentamicin (27%) and imipenem (16%), while doxycycline (1%), amikacin (2%), and ciprofloxacin (6%) showed the lowest sensitivity. The highest resistance rates were observed for cefotaxime and ceftriaxone (both 36%), followed by ampicillin and amoxicillin/clavulanic acid (both 29%). Conclusion: E. coli remains the predominant cause of UTIs in the Qalat Saleh District, with a higher burden among females and urban residents. Alarmingly low antibiotic sensitivity levels were observed, especially to commonly used antibiotics, indicating a pressing need for revised empirical treatment guidelines and antibiotic stewardship efforts in the region.
- Open Access
- 5 April 2024
- Published by EDP Sciences in BIO Web of Conferences
- Vol. 97, 00013
Enhancing Agricultural Decision-Making through Data Analysis: Predicting Crop Health Outcomes
This research employs advanced data analysis techniques to predict crop health outcomes during harvest seasons, with a focus on insect count, pesticide use, and soil type. The study encompasses two main components: feature correlation and predictive modeling. Feature engineering techniques are applied to capture variations in pesticide use and insect infestation, enhancing predictive capabilities. Ensemble methods, including Random Forest, XGBoost, and Decision Trees, are employed to forecast patterns of crop damage based on identified trends. Decision Trees exhibit robust capabilities, achieving an impressive accuracy rate of 90.03%. Random Forest excels with a robust accuracy of 90.35%, highlighting its classification abilities. XGBoost stands out with an accuracy rate of 86.51%. In contrast, Logistic Regression, Naive Bayes, and Convolutional Neural Networks face challenges, displaying lower accuracy. The evaluation further emphasizes the strength of ensemble methods and Decision Trees through precision, recall, and F1-Score metrics, providing a comprehensive understanding of relationships within pesticide damage. The framework of the study introduced in this paper can be seen as a major step forward with regard to agricultural decision-making. We present actionable strategies to enhance crop health while reducing damage through the integration of feature correlation, predictive modeling and precise evaluation metrics. The innovativeness is in the use of ensemble methods and Decision Trees that are implemented to promote informed decision-making among stakeholders through a sustainable approach to agriculture.
- 5 April 2024
- Published by EDP Sciences in BIO Web of Conferences
- Vol. 97, 00013
Open AccessEnhancing Agricultural Decision-Making through Data Analysis: Predicting Crop Health Outcomes
This research employs advanced data analysis techniques to predict crop health outcomes during harvest seasons, with a focus on insect count, pesticide use, and soil type. The study encompasses two main components: feature correlation and predictive modeling. Feature engineering techniques are applied to capture variations in pesticide use and insect infestation, enhancing predictive capabilities. Ensemble methods, including Random Forest, XGBoost, and Decision Trees, are employed to forecast patterns of crop damage based on identified trends. Decision Trees exhibit robust capabilities, achieving an impressive accuracy rate of 90.03%. Random Forest excels with a robust accuracy of 90.35%, highlighting its classification abilities. XGBoost stands out with an accuracy rate of 86.51%. In contrast, Logistic Regression, Naive Bayes, and Convolutional Neural Networks face challenges, displaying lower accuracy. The evaluation further emphasizes the strength of ensemble methods and Decision Trees through precision, recall, and F1-Score metrics, providing a comprehensive understanding of relationships within pesticide damage. The framework of the study introduced in this paper can be seen as a major step forward with regard to agricultural decision-making. We present actionable strategies to enhance crop health while reducing damage through the integration of feature correlation, predictive modeling and precise evaluation metrics. The innovativeness is in the use of ensemble methods and Decision Trees that are implemented to promote informed decision-making among stakeholders through a sustainable approach to agriculture.