IEEE Transactions on Biomedical Engineering

Authors: Sungjin Oh, Jose Roberto Lopez Ruiz, Kanghwan Kim, Nathan Slager, Eunah Ko, Mihály Vöröslakos, Hyunsoo Song, Wangbo Chen, Sung-Yun Park, Euisik Yoon Recent neuroscientific research craves for understanding sophisticated brain networks formed by neuron ensembles across multiple regions. An ideal way to unveil the complex connectome is bidirectionally interacting (simultaneous recording and stimulation) with neurons at high spatiotemporal resolutions.

Authors: Hao Guan, David Bates, Li Zhou Artificial intelligence (AI) is increasingly integrated into healthcare, supporting diagnosis, patient monitoring, outcome prediction, and treatment planning. However, after deployment, medical AI systems may experience performance degradation as data distributions shift, patient populations evolve, clinical practices change, and data quality varies over time.

Authors: Anna Tegon, Nicholas Lehmann, Yawei Li, Andrea Cossettini, Luca Benini, and Thorir Mar Ingolfsson FEMBA addresses the gap between large EEG foundation models and the severe memory, latency, and power limits of wearable neuro-monitoring hardware. The proposed FEMBA-Tiny encoder uses two bidirectional Mamba blocks to model long EEG sequences with linear complexity, while a lightweight Transformer decoder is used only during pre-training.

Authors: Weijie Qi, Matthew Hammink, Andrew Ooi, David B. Grayden, Lindsea C. Booth, Brooke L. Farrugia, and Sam E. John Endovascular neural interfaces place tiny electrodes on a stent inside a brain vein, enabling neural recording or stimulation without open-brain surgery. Their long-term success, however, depends on how the vein responds to the implant. Excess tissue can grow around a stent, narrowing the vessel and potentially affecting device performance.

Authors: Anna Tegon, Nicholas Lehmann, Yawei Li, Andrea Cossettini, Luca Benini, and Thorir Mar Ingolfsson FEMBA addresses the gap between large EEG foundation models and the severe memory, latency, and power limits of wearable neuro-monitoring hardware. The proposed FEMBA-Tiny encoder uses two bidirectional Mamba blocks to model long EEG sequences with linear complexity, while a lightweight Transformer decoder is used only during pre-training.

Authors: Andrew S. Perley and Todd P. Coleman Phase-amplitude coupling (PAC) describes how the phase of a slow rhythm modulates faster activity. Because PAC may reflect neural excitability and communication, resolving when it changes is important for studying transient and event-related physiology. We introduce dynamic gamma PAC (dgPAC), a probabilistic state-space method for point-in-time PAC estimation.

Authors: Ioana-Georgiana Vasilaș, Anagha Navale, Paul Čvančara, Ali Usama, Bastian Rapp, Thomas Stieglitz Can we trust what we observe after a neural implant is explanted? Delamination and structural alterations are frequently reported, yet distinguishing changes associated with implantation from those introduced during routine post-mortem processing remains challenging.

Authors:  Stephan Strassle Rojas, Travis C. Singh, Jimena Martín Tempestti, Alessandro Veneziani, Brooks D. Lindsey Percutaneous coronary intervention (PCI) is a minimally invasive procedure for alleviating symptoms of myocardial ischemia by stenting stenotic coronary arteries. Wall shear stress (WSS) is an established indicator of plaque rupture risk and likelihood of restenosis, however, it is not currently used to guide stenting.

Authors: Viktor van der Valk, Douwe Atsma, Roderick Scherptong and Marius Staring Accurate interpretation of ECG signals is essential for diagnosing cardiac conditions, but traditional expert-driven analysis is time-consuming, costly, and prone to overlooking subtle diagnostic features. While AI  has demonstrated strong potential for automating ECG interpretation, most existing models lack the explainability that clinicians require to trust and act on their predictions.

Authors: Kaleab A. Kinfu, Carolina Pacheco, Alice D. Sperry, Deana Crocetti, Bahar Tunçgenç, Stewart H. Mostofsky, René Vidal Autism is a highly complex condition affecting approximately 1 in 100 people, yet demand for timely diagnosis and support outpaces available services. This gap calls for scalable, objective behavioral measures that capture individual differences and complement clinical assessments.