Abstract:

Three magnetometers based on dc superconducting quantum interference devices (SQUIDs) fabricated from YBa2Cu3O7 x have been operated in a magnetically shielded room using a flux-locked loop involving additional positive feedback with bias current reversal. Two of these devices, integrated multiloop dc SQUIDs with outer diameters of 7 mm, achieved white noise levels of 10 fT/√Hz for bicrystal junctions and 30 fT/√Hz for step‐edge junctions. The third magnetometer involved a flux transformer with a 10×10 mm2 pickup coil connected to a 16-turn input coil which was inductively coupled to a bicrystal SQUID. This device achieved a white noise of 16.2 fT/√Hz. High quality magnetocardiograms were obtained without signal averaging.

Abstract:

At the current state of technology, multichannel simultaneous recording of combined electric potentials and magnetic fields should constitute the most powerful tool for separation and localization of focal brain activity. We performed an explorative study of multichannel simultaneous electric SEPs and magnetically recorded SEFs. MEG only sees tangentially oriented sources, while EEG signals include the entire activity of the brain. These characteristics were found to be very useful in separating multiple sources with overlap of activity in time. The electrically recorded SEPs were adequately modelled by three equivalent dipoles located: (1) in the region of the brainstem, modelling the P14 peak at the scalp, (2) a tangentially oriented dipole, modelling the N20-P20 and N30-P30 peaks, and part of the P45, and (3) a radially oriented dipole, modelling the P22 peak and part of the P45, both located in the region of the somatosensory cortex. Magnetically recorded SEFs were adequately modelled by a single equivalent dipole, modelling the N20-P20 and N30-P30 peaks, located close to the posterior bank of the central sulcus, in area 3b (mean deviation: 3 mm). The tangential sources in the electrical data were located 6 mm on average from the area 3b. MEG and EEG was able to locate the sources of finger stimulated SEFs in accordance with the somatotopic arrangement along the central fissure. A combined analysis demonstrated that MEG can provide constraints to the orientation and location of sources and helps to stabilize the inverse solution in a multiple-source model of the EEG.

Abstract:

Stereoelectroencephalography is a minimally invasive neurosurgical technique used to lo-calize the epileptogenic zone (region responsible for seizure generation) in patients withdrug-resistant epilepsy. The procedure requires the placement of multiple intracranial elec-trodes along carefully planned trajectories that reach predefined brain targets while avoidingcritical structures such as blood vessels. Planning these trajectories is complex and timeconsuming, relying heavily on manual inspection of medical images and surgical expertise.Consequently, there is increasing interest in computer-assisted approaches that can supportclinicians during the planning process.This thesis investigates the feasibility of using deep reinforcement learning to automatethe trajectory generation step of the SEEG planning workflow. The proposed frameworkintegrates medical imaging, vascular modeling, and a reinforcement learning agent into amodular pipeline. The trajectory planning task is then formulated as a sequential decision-making problem in which an agent iteratively adjusts electrode entry and target points. Inorder to approximate the action-value function and guide the search through the anatomicalspace, a Deep Q-Network is used, together with a reward formulation that incorporates safetyand geometric constraints, particularly vessel avoidance.Experimental evaluation using patient imaging data demonstrates that the RL agent cangenerate candidate electrode trajectories that reach predefined targets while maintaining safedistances from vascular structures. In this work, 2000 candidate trajectories were generatedand evaluated, with feasibility rates ranging from about 26% to 52% depending on anatomicalcomplexity. Generation of trajectories requires only a few seconds per electrode, indicatingthe potential for substantial reductions in planning time.Overall, the results demonstrate that reinforcement learning can effectively automate thegeneration of candidate SEEG trajectories and provide a foundation for future computer-assisted surgical planning systems.