(A) The connected network with significant differences among the groups. Note that the FA-weighted white matter network was constructed using the automated anatomical labelling template. The network primarily involved the frontal and thalamus regions. (B) Differences in the mean strength of the components among the groups. Post-hoc tests revealed decreased values in all patient subgroups compared with those in HC. The decreased pattern was more prominent in NTSZ and CTSZ patients, indicated by significant lower value. (C) The structural connectivity within the NBS network significantly mediated the difference between the 2 treated groups regarding their cognitive ability level. Path c represents the variance in treatment choice associated with global cognitive ability, and Path c’ represents the association between treatment choice and cognition after taking into account the structural connectivity of the NBS network as a mediator. Path c-c’ is the mediation effect and is significant at P < .05 based on confidence intervals from bias-corrected bootstrapping of 5000 samples. CTSZ, clozapine-treated schizophrenia patients; CUN, cuneus; HC, healthy controls; HIP, hippocampus; IFGtriang, triangular part of inferior frontal gyrus; L, left; MFG, middle frontal gyrus; NTSZ, never-treated schizophrenia patients; ORBinf, orbital part of inferior frontal gyrus; ORBmid, orbital part of middle frontal gyrus; PreCG, precentral gyrus; R, right; RTSZ, risperidone-treated schizophrenia patients; SOG, superior occipital gyrus; THA, thalamus. **P < .01; ***P < .001; ****P < .0001. For detailed information regarding the white matter connections in the significant NBS components, see supplementary Table 2.