Computer Vision News Computer Vision News 40 Congrats, Doctor Skylar At the core of Skylar’s work is a challenge faced by many researchers in neurotechnology: How can we model individual human heads with sufficient accuracy and speed to guide interventions like transcranial direct current stimulation (tDCS)? This matters because traditional “one-sizefits-all” protocols often fail to deliver consistent results. tDCS trials are usually judged based on group averages—but the brain is anything but average. To deliver clinically meaningful tDCS, we need to know if the right amount of current is reaching the Skylar E. Stolte (center) has recently earned her PhD from the SMILE Lab at the University of Florida under the supervision of Ruogu Fang (right). Her dissertation bridges artificial intelligence, cognitive aging, and neuromodulation, focusing on the development of trustworthy digital twins for precision modeling in non-invasive brain stimulation. She is now preparing to continue her training as a postdoctoral researcher, with the ultimate goal of leading her own research group at a top-tier R1 university. Congratulations, Doctor Skylar! reaching the right brain region in the right person. Finite Element Method (FEM) models are a step forward, but they require specialized expertise and can break down on real-world, diverse data—especially in older populations. That’s where Skylar’s research comes in. Her thesis tackled three fundamental questions: Can deep learning be used to rapidly and accurately segment the full head into tissue types essential for tDCS modeling? How can we improve model generalization, especially for underrepresented populations and unseen anatomical variations? Is it feasible to create “digital twins” of individual brains that generate electric field maps—without massive computational cost or lengthy modeling process? To answer these, Skylar built a suite of AI-powered tools—available opensource—with memorable names like GRACE, DOMINO, DOMINO++, and tDCS-DT. These tools integrate calibrated deep learning, domain
RkJQdWJsaXNoZXIy NTc3NzU=