Skip to main content
. Author manuscript; available in PMC: 2021 Nov 19.
Published in final edited form as: Neuroimage. 2021 Aug 24;243:118514. doi: 10.1016/j.neuroimage.2021.118514

Table 2.

Average accuracy of the DeepAtrophy PAIIR measure and the ALOHA (Das et al., 2012) hippocampal atrophy rate measure in inferring the scan temporal order (STO) of same-subject scan pairs input in arbitrary order (STO accuracy). For the ALOHA measure, we consider it to be “correct” if the sign of hippocampal atrophy is negative for scans input in chronological order, and positive for scans in reverse chronological order. Accuracy is pooled across all five cross-validation folds. Accuracy is expected to be lower for less impaired groups because there is less underlying biological change for the same time interval than in more impaired groups. Abbreviations: ALOHA = Automatic Longitudinal Hippocampal Atrophy software/package; A+/A− = beta-amyloid positive/negative; NC = cognitively normal adults; eMCI = early mild cognitive impairment; lMCI = late mild cognitive impair.

A− NC A+ NC A+ eMCI A+ lMCI All Groups
ALOHA 69.7% 74.3% 75.1% 85.2% 75.5%
DeepAtrophy 85.4% 89.5% 88.6% 92.4% 88.5%