Software and hardware for real-time EEG-guided multi-locus TMS

In my thesis, I develop hardware and software that enable multi-locus transcranial magnetic stimulation (mTMS) to be guided in real time by feedback signals such as electroencephalography (EEG). I present a timing method that uses the internal clock of the TMS device, removing the need for precise coordination between the control computer and the stimulation device, thus avoiding communication-related timing imprecision. By synchronizing the internal clocks of the EEG and TMS devices, stimulation can be targeted to specific phases of brain rhythms, with timing errors below 0.3 ms. I also design a control interface that allows constructing complex stimulation pulse sequences from simple commands, such as individual pulses and capacitor charge/discharge actions. In addition, I contribute to the design, construction, and clinical deployment of the mTMS device.To enable EEG-guided stimulation, I develop open-source Linux software that performs real-time signal analysis and determines stimulation timing using Python. With a commercial TMS device, the software achieves timing errors below 1.4 ms, confirming its compatibility with conventional TMS systems. I evaluate several microservice frameworks for the control software and select ROS 2 for its realtime capabilities and flexible communication models, making it well-suited for EEG-guided TMS. The methods and tools introduced in this thesis increase the timing precision and simplify the control logic of EEG-guided transcranial magnetic stimulation, offering a foundation for the development of new stimulation protocols and, ultimately, more effective clinical treatments.