About
I am a Postdoctoral Research Fellow at the Department of Microbiology, Immunology and Cell Biology, WVU School of Medicine (PI: Dr. Gangqing Hu). My research centers on the measurement and evaluation of large language models and multimodal foundation models in clinical AI — identifying systematic failure modes that matter for patient safety.
Active research threads: sycophancy in chatbot clinical consultations, prompt framing bias in dermoscopy, text-dominant reasoning failures in multimodal models, and foundation model benchmarking for diabetic retinopathy grading. Clinical domains: dermatology, psychiatry, and ophthalmology.
News
May 2026
Paper published in
Diagnostics: "Comparative Analysis of General-Purpose vs. Domain-Specific Multimodal Models for Diabetic Retinopathy Classification."
doijournal
Apr 2026
Co-authored paper in
npj Digital Medicine (IF 15.1): AI-generated psychiatric vignette evaluation.
doijournal
Apr 2026
Comment accepted in
JAAD: evolving GPT performance in dermoscopic diagnosis.
doijournal
Nov 2025
Oral presentation at AAAI SECURE4H, Arlington VA: "Sensing Without Alerting."talk
Aug 2025
Paper at
IEEE IGARSS 2025: patch-based tiny object classification for aerial images.
doiconference
Selected Publications
The cost of AI sycophancy in dermoscopic diagnosis accepted
M. I. Nouyed, D. A. Adjeroh, D. Xu, M. S. Kolodney, G. Hu · JAAD
Sensing but Not Alerting: The Cost of Sycophancy in ChatGPT Psychodermatology Consultations under review
M. I. Nouyed et al. · JAAD International
Text-Dominant Decision-Making by Large Multimodal Models in Dermatology Clinical Challenges accepted
M. I. Nouyed, L. E. Keplinger, H. Akhter et al. · JAAD
M. I. Nouyed, M. Al-Mamun, D. A. Adjeroh, G. Hu · Diagnostics, 2026
W. Zheng, D. N. Chandran, D. E. Elswick, M. I. Nouyed, G. Hu · npj Digital Medicine, 2026
Full list → Google Scholar
Talks
Nov 2025
Sensing Without Alerting: The People-Pleaser Cost of General-Purpose Chatbots in Dermatology Patients with Psychiatric Comorbidity
AAAI SECURE4H Fall Symposium · Oral · Arlington, VA
Jul 2025
Using Multimodal LLMs for Studying Scientific Articles
WVU Summer Workshop on AI, Digital Health & BioML · Oral · Morgantown, WV
Aug 2025
Patch-Based Tiny Object Classification for High Resolution Aerial Images
IEEE IGARSS 2025 · Poster · Honolulu, HI
Jul 2024
Efficient Analysis of Gigapixel Histopathological Images
WVU Summer Workshop on AI & Digital Health · Oral · Morgantown, WV
Dec 2022
Efficient Classification of Very High Resolution Histopathological Images
IEEE BIBM 2022 · Oral · Las Vegas, NV
Projects
Multi-agent simulation framework evaluating sycophancy and psychiatric safety failures in ChatGPT psychodermatology consultations.
Evaluation of text-dominant reasoning failures in large multimodal models for clinical dermatology tasks.
Large-scale experiments on prompt framing effects on LLM melanoma classification across GPT-5, Gemini, and MedGemma.
Benchmarking general-purpose vs. domain-specific foundation models for diabetic retinopathy grading across zero-shot, few-shot, and fine-tuning paradigms.
Discriminative patch selection and ensemble decision fusion for gigapixel whole-slide image classification.
Teaching
Sp 2025
EE465: Digital Image Processing — sole instructor, undergraduate level
Lane Dept. of CSEE · West Virginia University
Fa 2024
Undergraduate research mentoring — cellular image preprocessing pipelines
Social Media and Data Analytics Lab · WVU
iqbalnaved@gmail.com ·
Dept. of Microbiology, Immunology and Cell Biology, WVU School of Medicine ·
Morgantown, WV
