Skip to main content
. Author manuscript; available in PMC: 2022 Jun 17.
Published in final edited form as: Med Image Comput Comput Assist Interv. 2021 Sep 21;12902:80–89. doi: 10.1007/978-3-030-87196-3_8

Table 1:

Supervised downstream tasks in frozen or fine-tune scenarios. Left: Age regression on healthy subjects with R2 as an evaluation metric. Right: classification on ADNI dataset with BACC as the metric.

Methods Health Aging (R2)
ADNI (BACC)
Age
NC vs AD
sMCI vs pMCI
Frozen Fine-tune Frozen Fine-tune Frozen Fine-tune
No pretrain - 0.72 - 79.4 - 69.3
AE 0.53 0.69 72.2 80.7 62.6 69.5
VAE [12] 0.51 0.69 66.7 77.0 61.3 63.8
SimCLR [6] 0.56 0.73 72.9 82.4 63.3 69.5
LSSL [24] 0.59 0.74 74.2 82.1 69.4 71.2
Ours (LNE) 0.62 0.74 81.9 83.6 70.6 73.4