gml16 / rl-medical Goto Github PK

Hi!
Thanks for sharing your code! RL it's a super interesting topic in which I'm still new, so my apologies in advance if these questions are too obvious.

When running the program, there is a hyperparameter “seed” (defalt: None). I found that when it is not set, the result of each evaluation will be different. How does the seed parameter affect the results of the evaluation? How should I set it when training and testing on an unseen dataset?

Thanks for the help!

I could not fined the website of the dataset which I mentioned .Could you tell me how to download it

could you please tell me how to generate the landmarks to txt files?
For example, in ADNI_002_S_0816_MR_MPR__GradWarp__B1_Correction__N3__Scaled_Br_20070217005829488_S18402_I40731.txt, how to explain the first line 86,86,84? Is this the corresponding to the physical coordinate in the CT image?

Hi, do you have any idea how to modify the coding on datareader.py so that the displayed image especially CT scans has higher contrast.
(before)
(after)

Hi, thank you for your great contribution.
I want to use the model to predict landmarks for images without landmark ground truth file. However, it seems both the train and eval model require the landmark file.
What should I do?
Looking forward to your reply.

After I applied for ADNI data set, I did not find the anatomical landmark label of corresponding MRI. Could you please tell me whether the dataset you used for training was marked by yourself ?

May I ask what tool doctors use to manually annotate your dataset?
I tried Dolphin and itk snap, but their effects were not very good.
This question is crucial to me, thank you

I am trying to train a model. Due to some version mismatch I run into some import errors. I am using python 3.8.2. Is there a specific python version that I should be using?

Traceback (most recent call last):
  File "DQN.py", line 8, in <module>
    from logger import Logger
  File "/scratch/sshanmug/rl-medical/src/logger.py", line 6, in <module>
    from torch.utils.tensorboard import SummaryWriter
  File "/scratch/sshanmug/rl-medical/env/lib/python3.8/site-packages/torch/utils/tensorboard/__init__.py", line 12, in <module>
    from .writer import FileWriter, SummaryWriter  # noqa: F401
  File "/scratch/sshanmug/rl-medical/env/lib/python3.8/site-packages/torch/utils/tensorboard/writer.py", line 9, in <module>
    from tensorboard.compat.proto.event_pb2 import SessionLog
  File "/scratch/sshanmug/rl-medical/env/lib/python3.8/site-packages/tensorboard/compat/proto/event_pb2.py", line 6, in <module>
    from google.protobuf import descriptor as _descriptor
  File "/scratch/sshanmug/rl-medical/env/lib/python3.8/site-packages/google/protobuf/descriptor.py", line 51, in <module>
    from google.protobuf.pyext import _message
ImportError: /scratch/sshanmug/rl-medical/env/lib/python3.8/site-packages/google/protobuf/pyext/_message.cpython-38-x86_64-linux-gnu.so: undefined symbol: _ZN4absl12lts_2023012512log_internal9kCharNullE

Hi,
May I ask how do you generate reproducible results as the environment randomly resets the state in each game? since I can't find how the random seeds can be manually set for each experiment.
Thank you in advance!

When I run your code：“python DQN.py --task eval --load 'data/models/BrainMRI/Network3d8agents.pt' --files 'data/filenames/image_files.txt' 'data/filenames/landmark_files.txt' --file_type brain --landmarks 13 14 0 1 2 3 4 5 --model_name "Network3d"

The result of my error is as follows:
Traceback (most recent call last):
File "DQN.py", line 234, in
evaluator.play_n_episodes(fixed_spawn=args.fixed_spawn)
File "C:\Users\86159\Desktop\rl-medical-master\src\evaluator.py", line 44, in play_n_episodes
score, start_dists, q_values, info = self.play_one_episode(render, fixed_spawn=fixed_spawn[j])
TypeError: 'NoneType' object is not subscriptable

Do you know why? This question may be very basic, because I am a novice.

hi, thanks for this code, it's amazing,
I'm having trouble training the model with my own data, i have some ct scanner with different size between them, for example, 512512250, 512256128...
So it seems that it is not possible to use this type of data? or what I should do?

thanks for your help

Oscar

Hi @gml16 , I'm trying to learn sth from this project.
I just cloned it to my PC and used the instruction of Evaluating with 8 C-MARL agents.
However, I met an error saying that'RuntimeError: Unable to find a valid cuDNN algorithm to run convolution'.
I'm sure that I'm using the same environment as the file 'environment.yml' offered.
Btw, I reinstalled cuDNN 7.6.5 but it didn't work.
Also, I've looked for several solutions saying that decreasing the batch size could be helpful.
I changed the batch size from 64 to 32 but the same problem occurred.
So where could the problem be at?
Maybe it's a very basic problem since this is my first time to apply a project on my own.I'm really desperate for the training.
Looking forward to your reply.

Hi @gml16, I was able to train a model for a set of ultrasound images. I was wondering how to change the view of the images in the medical player. I want to get something like what you have done for fetal head ultrasound.

When I evaluate I get the following,

but I would like to get something like this,

Hi @gml16 , may I know what is the criteria to use the 'play' task during testing. I mean can you show the example how to run the 'play' command for testing

Hey @gml16, I am working on adding data augmentation. Currently, I augment the data and place it inside the images folder. Is there any place that I could simply add the data augmentation logic within the code, so that it will be taken care of during runtime? Thanks

I'm sorry to disturb you.
I want to use the project for a large datset. I have a large dataset which contain more than 3000 data for training. The training log shows that distance is not decrease but fluctuate a lot.
I notice that the dataset you used in your article has fewer samples.
Do I need to do some special modification to the code to make it suitable for large dataset training, like using longer episodes or something?
Looking forward to your reply.