The "tibiaSeq_deploy" folder contains the directory structure
and pre-compiled Windows executables for Pytorch-based Attention-UNET
segmentation of mouse tibia from GRE-MRI (MT-off) acquistions
(Kushwaha, et al. Tomography 2023: 10.3390/tomography9020048).

The corresponding images used for training and resulting segmentation masks
are available from https://github.com/dumichgh/MFJK1_Segmentation_MHDs
Trained models are provided in 'train_model_bak' folder of this distribution

%%% 8/2023 Deployment Author: Humera Tariq (thumera@umich.edu) %%%

Follow the steps below for deployment: 

1- Create a new folder 'myTibiaSeg' on your C:Drive (within Win-exe-Path, and permissions)
2- Copy the folder contents from 'tibiaSeg_deploy'. 
3- Check that at least one 'pth' model file exists in folder 'local_model'
FYI: 'model66.pth' is trained on 107 scans, 23 mice, 32 test-retest pairs annotated by a single expert 
    (see Tomography 2023: 10.3390/tomography9020048 )
4- Open 'input_mt_off_model.csv' to edit (i) mhd paths (ii) model name (iii) choice 'Y' or 'N'
FYI: current example path and naming convention are for published data 
     (https://github.com/dumichgh/MFJK1_Segmentation_MHDs)
NOTE: update the local "path" to input MHD data in CSV and set at least one "choice" to 'Y' to run.
5- Open the batch file 'run_me_MTseg.bat' with notepad to edit csv file name and paths if needed 
6- Close all setup files, especially the .csv
7- Double click on batch file  'run_me_MTseg.bat' to run the application
8- Watch console window for messages 
9- upon completion, the log file is saved in folder 'logs/log.txt' for your record.