The outcomes of Transcranial Magnetic Stimulation (TMS) depend critically on the momentary intrinsic cortical state, resulting in high variability that compromises clinical efficacy. To enable reliable, personalised TMS, closed-loop systems are essential for predicting and targeting favourable brain states in real-time.

This study proposes a novel TMS-EEG approach based on a Hybrid Fuzzy Logic System (FLS) to predict single-trial brain responsiveness accurately. Inputs extracted from the pre-stimulus TEP included Power Spectral Densities across canonical bands (delta to gamma) and the non-linear measure Higuchi Fractal Dimension (HFD), which reflects network complexity. The post-stimulus response was quantified using the Area Under the Curve (AUC), with trials labelled as ‘low’ or ‘high’ responders.

The FLS is uniquely suited to model the non-linear relationships of biological data, providing transparency and interpretability absent in 'black box' models. It utilises a hybrid rule-based inference system integrating Expert-defined rules (neurophysiology-based) and Data-driven rules from a Random Forest classifier to generate understandable, linguistic mappings.

Across 1560 trials, the model achieved classification accuracy of 73% and a Cohen's Kappa score of 0.46. Rule inspection confirmed that brain states characterised by reduced HFD and reduced beta and gamma power were associated with 'high' responsiveness. This research validates the FLS as a transparent, high-performance computational engine, representing a substantial step toward practical adaptive TMS protocols guided by real-time brain-state prediction to maximise therapeutic efficacy.