Accuracy doesn't change regardless of different models

Hi all! I’m trying to do binary segmentation with the slice of CT data. Whenever I computed accuracy I got the same value for the average of each epoch. I used the sigmoid function for the output of the model. any helps would be appreciated!

check and see if your model can learn any thing with small subset of your dataset, try to over train on them.
10 sample should be enough for test.
see if your loss decease.

I checked it. but train and validation acc doesn’t make sense anytime. pls see the plots

sorry, i don’t know