Research
Applied AI research focused on responsible, explainable, and impactful intelligent systems.
Research at AMIIE
The Applied Machine Intelligence Initiatives & Education (AMIIE) Lab conducts interdisciplinary research in artificial intelligence and machine learning with an emphasis on real-world impact. Our work spans healthcare and biomedical AI, responsible and explainable AI, computer vision and multimodal learning, graph-based machine learning, generative AI, and intelligent decision-support systems.
Across these areas, we are particularly interested in building AI systems that are accurate, interpretable, reliable, fair, and useful in practice. Our research combines methodological development with application-driven studies and often involves collaboration across computing, healthcare, and other scientific disciplines.
Core Research Programs
Our current research is organized around five complementary themes.
AI for Healthcare & Biomedical Research
Developing and evaluating AI methods for medical imaging, health data science, biomedical informatics, clinical prediction, patient-centered applications, and computational approaches that support biomedical discovery and healthcare decision-making.
Responsible, Explainable & Fair AI
Studying interpretability, explainability, reliability, fairness, and bias in AI systems, particularly in high-impact settings where model decisions must be transparent and trustworthy.
Computer Vision & Multimodal AI
Building learning systems for images and multimodal data, including medical image analysis, representation learning, image understanding, limited-data learning, and integration of visual and textual information.
Graph Machine Learning & Biomedical Networks
Exploring graph neural networks and network-based learning for complex biological and biomedical data, including the integration of molecular relationships, structured knowledge, and high-dimensional patient data for discovery and prediction.
Generative AI, LLMs & AI Agents
Investigating generative AI and agentic systems with a focus on reasoning, reliability, explainability, evaluation, and responsible use. Current interests include comparing conventional language models with agent-based AI workflows in complex decision-making tasks.
Current Research & Selected Publications
Current research directions at AMIIE, together with representative recent publications that illustrate each area.
AI for Healthcare & Biomedical Research
We develop AI and machine-learning methods for medical imaging, health data science, biomedical informatics, clinical prediction, and decision support. Our work emphasizes practical models that can contribute to biomedical discovery and healthcare applications.
Representative Recent Work
- KneeXNet-2.5D: a clinically-oriented and explainable deep learning framework for MRI-based knee cartilage and meniscus segmentation. npj Health Systems, 2026.
- Predicting cancer survival at different stages: Insights from fair and explainable machine-learning approaches. International Journal of Medical Informatics, 2025.
Responsible, Explainable & Fair AI
We study interpretability, explainability, fairness, reliability, and bias in AI systems, particularly in high-impact domains where predictions and model behavior must be transparent, trustworthy, and meaningfully evaluated.
Representative Recent Work
- State-of-the-art in responsible, explainable, and fair AI for medical image analysis. IEEE Access, 2025.
- Explainable Contrastive Learning for KL Grading Classification in Knee Osteoarthritis. IEEE EMBC, 2025.
Computer Vision & Multimodal AI
Our work in computer vision and multimodal AI includes medical image analysis, representation learning, limited-data learning, image understanding, and methods that integrate visual information with textual or structured data.
Research Emphasis
- Explainable and data-efficient medical image analysis.
- Image segmentation, classification, and representation learning.
- Multimodal learning that connects visual, textual, and structured information.
Graph Machine Learning & Biomedical Networks
We are exploring graph neural networks and network-based approaches for biological and biomedical data, where relationships among genes, proteins, clinical variables, patients, or other entities can be represented explicitly as structured networks.
Current Direction
- Graph neural networks for biomedical prediction and discovery.
- Biological interaction networks and patient-level data integration.
- Interpretable graph learning for biomarker and knowledge discovery.
Generative AI, LLMs & AI Agents
We investigate generative AI and agentic systems with a focus on reasoning, reliability, explainability, evaluation, and responsible use. An emerging research question is whether AI agents improve reasoning, reliability, and safety compared with conventional large language models.
Representative Recent Work
- Explainable Feature Engineering in Health Data Science: Empirical Comparison of ChatGPT-4o and Classical Machine Learning Methods. ACM/IEEE CHASE, 2025.
- Ongoing research on AI agents, agentic workflows, reasoning, reliability, and safety.
Computational Text Analytics & Scientific Knowledge
AMIIE also studies computational methods for extracting knowledge from scientific, biomedical, and health-related text. Earlier work in textual datasets, word embeddings, and scientific literature analysis now connects naturally with large language models, retrieval-based methods, and scientific knowledge discovery.
Research Emphasis
- Biomedical and scientific text mining.
- Open scientific datasets and terminology discovery.
- Connections between classical text analytics, LLMs, and knowledge discovery.
Research Philosophy
AMIIE emphasizes interdisciplinary, student-engaged research that connects methodological AI innovation with meaningful applications. We aim to develop research that is not only technically strong, but also interpretable, reproducible, responsible, and capable of contributing to real-world scientific and societal needs.