Short Communications

AI-Based Liver Iron Assessment in Sickle Cell Disease

Presenter: Khaled Baroudi (10 minutes including Q&A)
Co-authors: Tala El-Alami, Miled Yaacoub, Rita Zoghbi, Antoine Tarabay, Hilda Ghadieh, Leila Abs, Adlette Inati
Affiliations: Nini Hospital, Tripoli; University of Balamand, Koura; Lebanese American University, Byblos
Category: Admissions and Assessment Strategies

Abstract
Background: Sickle cell disease patients risk hepatic iron overload. AI-powered MRI (FerriSmart) provides precise LIC quantification.
Objectives: To investigate associations between LIC and transfusion history, labs, and complications.
Methods: Prospective study with 72 patients. LIC measured via FerriSmart MRI, correlated with clinical parameters.
Results: LIC correlated with cumulative transfusions, inversely with transfusion-free interval, and associated with vaso-occlusive crises, osteonecrosis, and cholecystectomy.
Conclusions: Cumulative transfusions drive hepatic iron. AI MRI is crucial for monitoring iron burden in SCD.
Keywords: Sickle Cell Disease, Liver Iron Concentration, AI MRI, Transfusion Patterns

Standardized Exam of the Abdomen Protocol (SEAP) Through Telemedicine in the Emergency Department

Presenter: Souraya Arabi (10 minutes including Q&A)
Co-authors: Sarah Abdul-Nabi, Hind Anan, Rasha Sawaya, Ahmad Zaghal, Hani Tamim, Jean-Marie El Semaani, Zahi Hamdan, Maha Makki, Moustafa Al Hariri, Afif-Jean Mufarrij
Affiliations: American University of Beirut, University of Balamand, Alfaisal University, Tamayuz Simulation Center
Category: Technological Advances: AI, Virtual Reality, Digital Health, and Simulation

Abstract
Background: Abdominal pain is a frequent ED complaint. Remote exams are difficult. SEAP was developed to guide structured patient self-exam.
Objectives: To validate SEAP by comparing self-exam with physician exam and assess feasibility, accuracy, and satisfaction.
Methods: Prospective study at AUBMC with 103 patients. Agreement measured with Cohen’s kappa. Patient feedback collected.
Results: High patient satisfaction with video guidance. Best agreement in RLQ (κ = 0.36–0.47). Moderate agreement in suprapubic and left flank.
Conclusions: SEAP is feasible, accurate, and patient-friendly, enhancing telemedicine triage.
Keywords: abdominal pain, telemedicine, emergency medicine

Assessing Lebanese Nurses’ Knowledge and Perceptions Toward AI Adoption in Healthcare Delivery

Presenter: Abir Abi Ghannam (10 minutes including Q&A)
Co-authors: Abbas Mourad, Saada Alame, Wafaa Takash Chamoun
Affiliations: Lebanese University – Faculty of Medical Sciences, Beirut, Lebanon
Category: Technological Advances: AI, Virtual Reality, Digital Health, and Simulation

Abstract
Background: Healthcare systems require cost-effective, efficient models. AI integration offers solutions. In Lebanon, little is known about nurse readiness for AI.
Objectives: To assess Lebanese nurses’ knowledge and perceptions toward AI in healthcare.
Methods: 300 nurse practitioners completed a validated questionnaire. Data were analyzed with SPSS using descriptive and inferential statistics.
Results: Knowledge average = 53.3%, perception = 58.1%. Higher education linked to higher scores. Female nurses had more positive perceptions. Work unit influenced results. Marital status impacted knowledge but not perception.
Conclusions: Lebanese nurses show moderate knowledge and positive attitudes toward AI. Education, gender, and work unit predict receptivity.
Keywords: AI, Lebanese Healthcare, Nurses, Knowledge, Perception

Locally Deployed Language Models for Pathology Education

Presenter: Mahmoud Atwi (10 minutes including Q&A)
Co-authors: Rachad Atat, Fouad Honeine
Affiliations: Lebanese American University, Beirut, Lebanon; Hammoud Hospital University Medical Center, Sidon, Lebanon
Category: Research in Medical Education

Abstract
Background: Large Language Models (LLMs) show promise for creating educational tools like flashcards for pathology. Few studies assessed their direct application on local hardware.
Objectives: To compare Baseline, LoRA fine-tuned, and LoRA-RAG hybrid pipelines on consumer hardware for pathology education.
Methods: Using a 7B model, pathology flashcards were generated and evaluated by metrics, AI judge, and pathologist blind review.
Results: LoRA-only model was most reliable. Hybrid had depth but 20% flagged errors. Baseline unreliable with 70% flagged.
Conclusions: Specialized fine-tuning yields high reliability; RAG adds variability. Future work will explore hybrid consistency.
Keywords: Medical Education, LLM, Fine-Tuning, Retrieval-Augmented Generation, Local AI

WhatsApp Use to Monitor Diabetic Foot/Leg Wounds Treated with Negative Pressure Wound Therapy

Presenter: Kaissar Yammine (10 minutes including Q&A)
Co-authors: Mariana Helou
Affiliation: Beirut, Lebanon – Lebanese American University
Category: Technological Advances: AI, Virtual Reality, Digital Health, and Simulation

Abstract
Background: Monitoring progress is the mainstay of wound treatment, yet it is known to be costly and time consuming. Tele-medicine has been proposed as an alternative to face-to-face consultations in diabetic ulcer monitoring. However, very few papers explored its utility and effectiveness through instant messaging application such as WhatsApp.
Objectives: To evaluate the validity of WhatsApp in monitoring diabetic wounds treated with negative pressure wound therapy (NPWT).
Methods: Twenty-two patients were prospectively recruited. All patients had an initial face-to-face consultation and debridement with or without surgery. Dressings were conducted twice per week remotely and media files were sent via SMS. A subsequent face-to-face consultation was scheduled whenever a complication was suspected. The primary outcomes were the percentages of accurate cases when a new or recurrent infection was suspected, and when a debridement was believed needed.
Results: Complete healing in 10 patients, skin graft procedure in 4, surgical debridement in 5, and 3 cases of failure to heal. WhatsApp monitoring showed 71.5% accuracy in infection detection and 73% accuracy in debridement need detection.
Conclusions: WhatsApp is a reliable method to document and track wound process and complications when using NPWT. It is valuable for patients in rural areas and in case of pandemics.
Keywords: diabetic foot ulcer; negative wound pressure therapy; WhatsApp; wound healing

AI-Aided Evaluation of Patient Satisfaction with Telemedicine for Type 2 Diabetes in PHCCs

Presenter: Zeinab Hazime (10 minutes including Q&A)
Co-authors: Hussein Hazimeh, Mohamad Chahrour, Rana El Haidari
Affiliations: Lebanese University, Ektidar Program, INSPECT-LB
Category: Technological Advances: AI, Virtual Reality, Digital Health, and Simulation

Abstract
Background: Telemedicine supports T2D management in primary healthcare centers.
Objectives: To assess satisfaction with PHCC telemedicine services for T2D and explore barriers.
Methods: Mixed-methods design. Phase I: 300 patients surveyed. Phase II: 30–40 interviews. AI applied for predictive analytics and sentiment analysis.
Results: Ongoing. Anticipated outcomes include robust satisfaction insights and predictors.
Conclusions: Study will provide evidence on patient satisfaction with PHCC telemedicine for T2D.
Keywords: Telemedicine, Patient Satisfaction, AI