MONAI Label App - Radiology - Transforms - Bug

Hi everyone,

I have started researching active learning and I am trying to use MONAI Label to develop my pipeline. I have started from the radiology sample app as a baseline app to develop it to be scriptable and I am having issues running it since I encountered this bug.

Describe the bug The reordering of the labels generates that happens here generates the wrong indexed order that in turn creates issues when its index is used here.

To Reproduce Use the main.py in radiology app to train a network.

Expected behavior I would expect this method to take as input something like:

{
    "background":0,
    "target_class_0":2,
    "target_class_1":1,
}

and transform it into:

{
    "background":0,
    "target_class_0":1,
    "target_class_1":2,
}

but instead, it transforms it into:

{
    "background":0,
    "target_class_0":2,
    "target_class_1":3,
}

And now the code here tries to compensate for this +1 shift and it ends up indexing outside of the number of labels since “target_class_1” with index 3 should be at index 2 actually.

Environment

Ensuring you use the relevant python executable, please paste the output of:

================================
Printing MONAI config...
================================
MONAI version: 1.0.1rc4
Numpy version: 1.23.4
Pytorch version: 1.12.1+cu102
MONAI flags: HAS_EXT = False, USE_COMPILED = False, USE_META_DICT = False
MONAI rev id: 52a7fde440123460f2ce13a515a0f3d023292f95
MONAI __file__: /home/danieltudosiu/projects/MONAILabel/venv/lib/python3.9/site-packages/monai/__init__.py

Optional dependencies:
Pytorch Ignite version: 0.4.10
Nibabel version: 4.0.2
scikit-image version: 0.19.3
Pillow version: 9.2.0
Tensorboard version: 2.10.1
gdown version: 4.5.3
TorchVision version: 0.13.1+cu102
tqdm version: 4.64.1
lmdb version: 1.3.0
psutil version: 5.9.3
pandas version: NOT INSTALLED or UNKNOWN VERSION.
einops version: 0.5.0
transformers version: NOT INSTALLED or UNKNOWN VERSION.
mlflow version: NOT INSTALLED or UNKNOWN VERSION.
pynrrd version: 0.4.3

For details about installing the optional dependencies, please visit:
    https://docs.monai.io/en/latest/installation.html#installing-the-recommended-dependencies


================================
Printing system config...
================================
System: Linux
Linux version: Ubuntu 18.04.6 LTS
Platform: Linux-5.4.0-131-generic-x86_64-with-glibc2.27
Processor: x86_64
Machine: x86_64
Python version: 3.9.15
Process name: python
Command: ['python', '-c', 'import monai; monai.config.print_debug_info()']
Open files: []
Num physical CPUs: 16
Num logical CPUs: 32
Num usable CPUs: 32
CPU usage (%): [16.8, 7.1, 5.1, 5.1, 5.8, 6.4, 5.8, 7.2, 5.8, 7.2, 5.1, 5.1, 7.1, 6.5, 8.0, 47.9, 6.4, 6.4, 5.8, 5.8, 6.5, 5.0, 6.4, 5.0, 5.8, 5.8, 5.8, 5.8, 5.9, 5.8, 6.4, 46.0]
CPU freq. (MHz): 2246
Load avg. in last 1, 5, 15 mins (%): [2.5, 3.0, 2.9]
Disk usage (%): 64.8
Avg. sensor temp. (Celsius): UNKNOWN for given OS
Total physical memory (GB): 62.7
Available memory (GB): 42.5
Used memory (GB): 19.1

================================
Printing GPU config...
================================
Num GPUs: 2
Has CUDA: True
CUDA version: 10.2
cuDNN enabled: True
cuDNN version: 7605
Current device: 0
Library compiled for CUDA architectures: ['sm_37', 'sm_50', 'sm_60', 'sm_70']
GPU 0 Name: TITAN V
GPU 0 Is integrated: False
GPU 0 Is multi GPU board: False
GPU 0 Multi processor count: 80
GPU 0 Total memory (GB): 11.8
GPU 0 CUDA capability (maj.min): 7.0
GPU 1 Name: GeForce GT 1030
GPU 1 Is integrated: False
GPU 1 Is multi GPU board: False
GPU 1 Multi processor count: 3
GPU 1 Total memory (GB): 2.0
GPU 1 CUDA capability (maj.min): 6.1

[danieltudosiu]

I had a chat with @diazandr3s today and explained the issue in detail. The buggy code is this for loop. I am proposing to change it to:

            idx=1
            for key_label, val_label in self.label_names.items():
                if key_label != "background":
                    new_label_names[key_label] = idx
                    label[d[key] == val_label] = idx
                    idx+=1
                if key_label == "background":
                    new_label_names["background"] = 0

@diazandr3s agreed with the surgical change.

@SachidanandAlle regarding the Radiology App I am mainly talking about the main.py which is peculiarly underdeveloped. What I would propose is to expose all the variables in the main.py and create a TODO list for users. Furthermore, I see some unused code (mainly looking at the self.planner variable) that could be removed to clarify of code. I am more than happy to try and do a PR for this as I have been doing it for my research.

[SachidanandAlle]

I think that the current Radiology app is somewhat underdeveloped.

@danieltudosiu it is true… it’s more of a reference implementation with couple of different types of model (to segment xyz labels) and these examples will not generalize to fit in every other use-case to train your own model.

Either you have to fix data preparation, transforms etc… depending on your dataset. But if you find better ways to make it more generic, please feel free to chime in and raise a PR to fix/add those improvements for everyone to use.