Speaker Details
Prof. Andrei Irimia Keynote Speaker
Associate Professor of Gerontology and Neuroscience
USC Leonard Davis School of Gerontology, University of Southern California
Title: Deep learning of multimodal neuroimaging to study neurodegeneration and brain aging
Abstract
Understanding how the human brain ages—and how neurodegenerative diseases alter this trajectory—is fundamentally a computational imaging problem. Modern neuroimaging datasets are large, heterogeneous, multi-site, and multimodal, combining structural MRI, diffusion imaging, functional MRI, and clinical/biomarker data. This talk will present recent advances from our group on deep learning methods for modeling brain aging and neurodegeneration using such multimodal data. We develop 3D convolutional, graph, and transformer-based architectures that learn latent representations capturing both global and regional aging dynamics. These models enable us to estimate brain age at voxel-level resolution, quantify deviations from healthy aging, and predict future cognitive decline. A major emphasis is on interpretability and reliability: we introduce saliency benchmarking frameworks, counterfactual generative models, and diffusion-based perturbation methods that reveal the neuroanatomical features driving model predictions. I will also discuss how we integrate clinical covariates—such as vascular and metabolic risk factors, traumatic brain injury, and sex-specific variables—into multimodal learning pipelines to understand how they modulate brain aging trajectories. Together, these approaches illustrate how deep learning can bridge the gap between high-dimensional neuroimaging and precision neurology, enabling individualized markers of aging and early detection of neurodegenerative processes. The talk will highlight open challenges in multimodal fusion, harmonization across scanners and cohorts, and the design of robust, generalizable imaging models suitable for clinical translation.
Brief Biography
Andrei Irimia, PhD is an associate professor in the Leonard Davis School of Gerontology at the University of Southern California, with courtesy appointments in biomedical engineering and quantitative/computational biology. His research focuses on brain aging, traumatic brain injury, and Alzheimer’s disease, using advanced neuroimaging and quantitative methods to understand individual variability in aging trajectories and dementia risk. Dr. Irimia leads several NIH-funded studies examining how chronic disease variables and women's health factors influence brain aging and neurodegeneration. His work bridges population neuroscience and clinical neurology, with the goal of improving early detection and stratification of patients at risk for cognitive decline.
Prof. Ragini Verma
Professor of Radiology, University of Pennsylvania
Director, DiCIPHRLab
Title: AI-Enhanced dMRI: Reliable Data, Consistent Anatomy, and Translational Biomarkers
Abstract
Diffusion MRI (dMRI) has become an essential modality for probing tissue microstructure, mapping white-matter architecture, and quantifying network-level brain organization. As its use expands, from surgical planning to quantitative biomarker development, the field faces persistent challenges: ensuring high-quality data at scale, consistently extracting tracts across individuals and pathologies, interpreting subtle microstructural deviations through normative modeling, and integrating diffusion features and connectome structure into reliable, individualized markers.
AI is reshaping this landscape by making dMRI analysis more scalable, reproducible, and clinically meaningful. This tutorial will survey how a spectrum of AI architectures, including 3D CNNs, transformers, vision–language models, graph neural networks, and generative models, can be deployed across the entire diffusion MRI pipeline, from data reliability to tract-level anatomy to patient-specific biomarker derivation.
We begin with data quality, the fundamental prerequisite for any diffusion workflow. We show how CNN-based models can automatically detect artifacts, motion corruption, and implausible tensor patterns, replacing manual QC with fast, consistent assessment suitable for large, multi-site cohorts. We then examine AI-enabled tractography and anatomical segmentation, highlighting how LLMs and vision–language models can learn tract identities and anatomical descriptors to enable rapid, automated white-matter parcellation, even in the presence of disruptive pathologies such as tumors. These approaches support both surgical planning and population-scale studies requiring standardized tract definitions.
Next, we transition to biomarker discovery, illustrating how different classes of AI models address distinct biological and clinical questions. Examples include brain-age prediction from diffusion features; microstructural anomaly indices derived from normative modeling; and voxel-wise AI cancer maps that reveal microscopic tumor infiltration and tissue heterogeneity beyond conventional imaging. Finally, we demonstrate the use of graph-based CNNs for connectome-level normative modeling, leveraging the topology of diffusion-derived networks to detect network-level abnormalities, characterize developmental and aging trajectories, and predict clinical outcomes.
Together, these methods illustrate the diverse ways in which AI can enhance the information extracted from diffusion MRI, from QC to tract anatomy to network modeling and quantitative biomarkers, enabling a more robust and clinically relevant diffusion imaging ecosystem.
Biography
Ragini Verma, PhD, is Professor of Radiology at the University of Pennsylvania, a Distinguished Investigator of the Academy for Radiology & Biomedical Imaging Research, and a Fellow of the American Institute for Medical and Biological Engineering (AIMBE). She directs the DiCIPHRLab (Diffusion & Connectomics in Precision Healthcare Research), where her team develops and applies advanced neuroimaging methods to illuminate brain structure, connectivity, and pathology—paving the way for precision medicine and translational science.
Her research spans diffusion MRI, connectomics, and multimodal data integration, with a particular focus on brain tumors, traumatic brain injury, and developmental disorders. Dr. Verma has pioneered methods for modeling the tumor microenvironment using diffusion tensor imaging and free-water correction and has developed clinically translated tools for surgical planning. She is also advancing frameworks for distributed data analysis, normative brain health summaries, and AI-driven large-scale data integration to enable precision neuroimaging. Through her leadership of multi-site consortia, she remains deeply committed to bridging discovery with clinical impact.
Dr. Verma also serves as Associate Vice Chair for Translation and Commercialization in Penn Radiology. In this role, she is building a Translation Accelerator to connect academia, industry, and clinical practice, with the goal of speeding the adoption of imaging biomarkers into routine care, improving patient outcomes, and finding AI-based solutions for clinical needs.
Dr. Vaanathi Sundaresan
Assistant Professor, Department of Computational and Data Sciences (CDS), IISc Bangalore
Title: AI applications for structural neuroimaging data analysis in healthy and pathological brains
Abstract
Structural neuroimaging has become a cornerstone for understanding both healthy brain organization and the neurobiological signatures of disease. With rapid advances in artificial intelligence (AI), particularly machine learning and deep learning, the field is undergoing a major transformation toward scalable, automated, and more sensitive analysis of MRI-derived brain morphology. This tutorial provides a comprehensive overview of state-of-the-art AI methodologies for structural neuroimaging and their applications across the lifespan and in diverse neurological and psychiatric disorders. We first introduce foundational concepts in feature extraction, and quantitative morphometry, highlighting the shift from handcrafted features to end-to-end deep learning models. We then discuss supervised, weakly supervised, and self-supervised strategies for tasks such as tissue segmentation, cortical and subcortical parcellation, shape modeling, and anomaly detection. Special emphasis is placed on emerging approaches for harmonizing multi-site data, learning from limited or noisy labels, and integrating imaging with clinical or genetic information for precision phenotyping. Next, we explore clinical applications in neurodegeneration, cerebrovascular disease, developmental disorders, and brain aging. Finally, we address ongoing challenges—including bias mitigation, uncertainty estimation, explainability, and regulatory considerations for clinical deployment—and outline future research directions toward robust, interpretable, and equitable AI-driven neuroimaging analysis.
Biography
Dr. Vaanathi Sundaresan is an Assistant Professor at the Department of Computational and Data Sciences (CDS) , Indian Institute of Science (IISc), Bangalore. She is also the convenor of Biomedical image Analysis (BioMedIA) laboratory at CDS, IISc. Prior to this appointment, she was working as a postdoctoral research fellow at Athinoula A. Martinos Centre, Department of Radiology, Harvard Medical School and Massachusetts General Hospital. She received my doctorate degree at Oxford Centre for function MRI of Brain (FMRIB), Wellcome Centre for Integrative Neuroimaging (WIN), University of Oxford. Later, she continued her research at WIN as a Postdoctoral researcher, where she is currently affiliated as an Honorary Research Fellow. She has around 10 years of experience in open source tool development for medical imaging applications.
Dr. Chirag Ahuja
Additional Professor, Department of Radiodiagnosis and Imaging, PGIMER, Chandigarh
Title: Transformational role of AI in Neuroradiology: Current Perspectives and Future Scope
Biography
Dr. Chirag Ahuja is an Additional Professor in the Department of Radiodiagnosis and Imaging, Division of Neuroimaging and Interventional Neuroradiology, at the Post Graduate Institute of Medical Education and Research (PGIMER), Chandigarh. A clinician–scientist with expertise in stroke, aneurysm, and AVM interventions, he also serves as Treasurer of the Society for Emergency Radiology–India. His research spans cutting-edge applications of artificial intelligence in neuroradiology, including projects on traumatic brain injury analysis, intracranial aneurysm flow diversion, and AI-driven segmentation of the Circle of Willis.
Dr. Ahuja has received numerous national and international accolades, including the RSNA International Young Academic Scholarship (2014), the V. M. Qasim Best Paper Award in Neurointervention (2014), and the Society of Interventional Radiology International Scholar Award (2017). With over 250 indexed publications, 18 book chapters, and more than 150 invited lectures, he is widely recognized for his academic leadership and contributions to neurointerventional imaging.
Dr. Jitender Saini
Additional Professor of Radiology, NIMHANS, Bengaluru
Title: Role of AI in neuroimaging: Radiologists perspective
Biography
Dr. Jitender Saini, MD, DM, is an Additional Professor of Radiology at the National Institute of Mental Health and Neurosciences (NIMHANS), Bengaluru, one of India’s leading centers for neurological care, research, and education. A specialist in neuroradiology, he plays a key role in advanced imaging for neurological and psychiatric disorders, integrating clinical expertise with cutting-edge imaging technologies.
Dr. Saini’s work spans diagnostic neuroimaging, interventional radiology, and translational research aimed at improving the management of complex brain and spine conditions. He is actively involved in postgraduate training and academic development at NIMHANS, contributing to the mentorship of future radiologists and neuroscientists.