Field-Based Localization: Applications in Cardiac and Wireless Sensing

Many localization problems reduce, in some form, to finding a point source from measurements taken elsewhere: a pacing electrode exciting the heart, or a transmitter moving through a room. This talk examines two such problems — activation-site localization in electrocardiographic imaging (ECGI) and indoor positioning from wireless channel state information (CSI). In both problems, we use learned mappings to transform indirect measurements into a spatial representation from which the underlying source location can be inferred.

In ECGI, our work addresses the inverse problem of recovering cardiac electrical activity and activation sites from body-surface potentials. We explore both physics-based and learning-based approaches, including a framework that integrates a physics-based inverse formulation with a learned proximal operator, and a FastKAN-based model that reconstructs epicardial potentials directly from body-surface measurements without requiring a patient-specific forward model at inference. In CSI-based localization, we similarly address the challenge of inferring location from indirect measurements, where multipath and environmental effects can make the mapping between CSI and location inherently ambiguous. Rather than forcing this ambiguity into a single coordinate, we represent it explicitly as a continuous spatial probability field and refine this field over consecutive observations using a FastKAN-based architecture.

————————

Gözde B. Akar is a Professor in the Department of Electrical and Electronics Engineering at Middle East Technical University (METU), where she leads research in artificial intelligence, signal processing, medical imaging, and intelligent sensing. She is a member of the METU Digital Transformation Center and Director of the Smart-Flexible Manufacturing Systems Laboratory.  In 2023-2024 he was a visiting researcher at Weill Cornell Medicine. She is  the co-founder of technology companies including ParanaVision and Vestia and advising several medical and defense-industry companies on AI and intelligent sensing technologies.

She  has also been active in the IEEE community through a range of leadership and professional service roles. She has served on the IEEE ICIP 2020/24 SAC and MMSP 2026 TPC, had various other technical and organizational roles in international conferences. She is also a member of the Editorial Board of Digital Signal Processing.Professor Akar's research is interdisciplinary, spanning AI-based medical imaging, inverse problems, computer vision, and industrial AI. Through her academic, professional, and entrepreneurial activities, Professor Akar has worked to bridge fundamental research with real-world applications, particularly in AI, intelligent sensing, and digital transformation.