Objective By collecting the anisotropy fraction FA in the diffusion tensor imaging data of 60 young smokers and 60 young non-smokers who were matched in terms of gender and education level,using the fiber bundle-based spatial statistical analysis method And a classification method based on support vector machine,in the 50 regions of the white mat-ter of the brain,the classification prediction of the two groups of subjects at the voxel level was provided to detect the smok-ing status of the brain and to distinguish between addicted patients and healthy groups'biomarkers.The classification had an average accuracy of 87.50%and an area under the curve of 0.92.The most important influence on the classification results was on both sides of the lower cerebellar peduncle,the right side of the corticospinal tract,the right side of the cerebral pe-duncle,both sides of the cingulate(hippocampus),the left side of the uncinate bundle,the fornix and the right side of the upper cerebellar peduncle.Our results demonstrated that the anisotropy score was fully functional as a discriminative bio-marker in detecting smoking status and had great potential in predictive classification and provides new insights into ma-chine learning studies of smoking-related neurophysiology perspective.
NETL
NSTL
万方数据
