shuchaopangunsw / radiomicscnns Goto Github PK

---------------------------------------------Read Me-----------------------------------------------------

1. All trained models with testing codes and extracted data with
different modalities are released here, based on our recently published paper:

   [1] Pang, Shuchao, Matthew Field, Jason Dowling, Shalini Vinod, Lois Holloway, and Arcot Sowmya. 
   "Training radiomics-based CNNs for clinical outcome prediction: Challenges, strategies and findings." 
   Artificial Intelligence in Medicine 123 (2022): 102230.

   So, please refer to our paper in your manuscript if you use our codes, trained models and extracted data.
   Thank you.

2. Implementation settings:
   All experiments in this study were implemented on two NVIDIA GPUs of GeForce RTX 2080TI and we coded all experiments 
   using Python 3.6.9 and Tensorflow 1.14.0. Here, we also provided the dependencies.txt file for readers, where you can
   find more detailed information (e.g., versions) of used libraries in our environment for this repository. 

3. About contents in this repository, details about them are introduced below:

   --> folder: trained_Models: 
               a) on_CT_data: here we used the released central slice CT data [3] to train and test our models.
               b) on_CT+GTV_data: here we used the extracted data by ourselves to train and test our models based on 
               raw data from [2].
   
   --> folder: extrated_Data: 
               a) CT_data: here please refer to the published reference [3] in Acknowledgements below.
               b) CT+GTV_data: here we extracted the CT+GTV data from the raw data from [2].
   
   --> folder: codes: 
               a) on_CT_data: 
                  1) DM: please used our trained models (DM) to test restuls on testing sets. 
                  2) LRF: please used our trained models (LRF) to test restuls on testing sets. 
                  3) OS: please used our trained models (OS) to test restuls on testing sets. 
               b) on_CT+GTV_data:
                  1) DM: please used our trained models (DM) to test restuls on testing sets. 
                  2) LRF: please used our trained models (LRF) to test restuls on testing sets. 
                  3) OS: please used our trained models (OS) to test restuls on testing sets. 

Acknowledgements:
   [2] Vallieres M, Kay-Rivest E, Perrin LJ, Liem X, Furstoss C, Aerts HJ, Khaouam N,
   Nguyen-Tan PF, Wang CS, Sultanem K, Seuntjens J. Radiomics strategies for risk
   assessment of tumour failure in head-and-neck cancer. Sci Rep 2017;7(1):1–14.
   [3] Diamant A, Chatterjee A, Valli`eres M, Shenouda G, Seuntjens J. Deep learning in
   head & neck cancer outcome prediction. Sci Rep 2019;9(1):1–10.