What can i do to improve my model

Hello, I'm trying to implement a 3D multi-class segmentation of light sheet scans using monai.
This is the result I have, and I have no idea of how i can improve this. It looks like it segments the outside of the area I'm interested in, does anyone have an idea of how to fix this ?

I'll also put here the transforms that I used and the training process.


Thank you in advance for your help

[Training and validation processes](https://preview.redd.it/ppejqpsxonyc1.png?width=926&format=png&auto=webp&s=83a2a46e441c4e8117df4e8ab60f2bd49720586c) …

class computervision hello light monai scans segmentation