Question A growing body of research suggest that aberrant interactions among cortical regions are crucially linked to motor rehabilitation after stroke (Guggisberg et al. 2019). Recent studies have shown the potential of Neurofeedback (NF) for stroke rehabilitation, however the effect of NF training upon functional interactions between motor areas is poorly understood. In a previous work (Lioi et al., 2020), we have tested bimodal EEG-fMRI NF for stroke rehabilitation: here we investigate the effect of NF training on motor networks. Methods Four right-handed chronic stroke patients (54 - 76 years, 2 females) with left hemiparesis took part to the study. The experimental protocol included 2 bimodal EEG-fMRI and 3 unimodal EEG NF sessions within a week. During each session, patients underwent 3 training runs alternating blocks of rest and motor imagery of the affected upper limb with NF. NF was displayed as a ball moving on a gauge proportionally to EEG and fMRI activities from regions of interest (ROIs) identified in the ipsilesional motor cortex. Representative ROIs time-series were extracted by selecting voxels that exceeded the NF contrast statistical threshold (p=0.05) within bilateral premotor, supplementary and primary motor cortices (PMC, SMA, M1). We used dynamic causal modeling (Zeidman et al., 2019) to assess causal influences between motor areas. The models were defined apriori on the base previous results (Grefkes et al., 2010) and included, respectively, 5 and 6 ROIs (Figure 1). We tested the effect of NF training on connection strengths with a Parametric Empirical Bayes second level analysis (Friston et al., 2016). Results In the model best explaining the difference between the first and the last training session (Figure 2) an increase in connectivity between ipsilesional motor ROIs was observed, which did not necessarily correspond to an increase in contralesional connectivity (Figure 2 B.). A general decrease in the strength of connection between hemispheres for PMC and M1 was also observed. This is of particular interest as an increase in ipsilesional connectivity between premotor and motor areas and a decrease in pathological transcallosal connections have been associated with improved motor performances in stroke patients (Grefkes & Fink, 2011) Conclusion These preliminary results on a small sample of stroke patients suggest that NF training of ipsilesional motor areas is associated to a reorganization of motor effective connectivity.