Skip to main content
. 2019 May 31;13:509. doi: 10.3389/fnins.2019.00509

Table 9.

Comparison of the performance of models trained from the CN vs. AD training set and the tasks’ own training set.

Task Training Set Testing Set B1
B2
ACC SEN SPE AUC ACC SEN SPE AUC
CN/pMCI CN/AD CN/pMCI 87.46% 90.73% 80.61% 87.61% 87.13% 87.80% 85.71% 90.31%
CN/pMCI CN/pMCI CN/pMCI 79.10% 87.80% 61.25% 76.84% 82.38% 87.20% 72.50% 81.64%
sMCI/pMCI CN/AD sMCI/pMCI 73.60% 66.67% 79.17% 75.59% 76.90% 68.15% 83.93% 79.61%
sMCI/pMCI sMCI/pMCI sMCI/pMCI 65.28% 54.69% 73.75% 66.82% 72.22% 73.44% 71.25% 77.49%

The Paired dataset was used.