[+] update nnUNet.md

Author: Kirscher

nnUNet/nnUNet.md

@@ -4,6 +4,28 @@
 
 ### Preprocessing
 
+The preprocessing pipeline follows the official nnU-Net v2 workflow
+(reorientation, resampling, intensity normalisation, etc.).
+
 ### Augmentations
 
+PSAT introduces a contraction-based augmentation strategy to expose the network 
+to stronger anatomical scaling.  This is achieved by modifying the
+`SpatialTransform` in nnU-Net so that random scaling factors are sampled from a
+wider range `(0.7, 2.5)` rather than the default `(0.7, 1.4)`.  Mirroring is disabled
+for the experiments.
+
 ### Training
+
+Several custom trainers are provided to support different experimental setups.
+`nnUNetTrainer_400epochs_NoMirroring_Finetune_lr_1e4` and
+`nnUNetTrainer_200epochs_NoMirroring_Finetune_lr_1e4` perform fine‑tuning with a
+lower learning rate and without mirroring.  The
+`nnUNetTrainer_1000epochs_NoMirroring_contraction` variants train from scratch
+for 1000 epochs while applying the contraction augmentation described above.
+
+All trainers inherit from `nnUNetTrainer` and thus retain full compatibility
+with the nnU-Net v2 command line tools.
+
+
+You have to place the scripts in your nnUNetv2 library corresponding folders.