The function train_bck_Complex_RM_CNN is not found ,and how to cite your work? Thanks!

[guyucowboy]

Hi, bck888 I have two questions to ask you. The function train_bck_Complex_RM_CNN is not found in train_model.py. How to solve that? How to cite your work, Multi-scale Gradual Integration Convolutional Neural Network for false positive reduction in pulmonary nodule detection in CT scan? Thank you very much! Best regards Gu Yu

[guyucowboy]

Hi,bck888 The function test_patch_extraction is also not found. What is the difference between test_patch_extraction and nodule_patch_extraction? Thanks for your help! Best regards Gu Yu

[bck8888]

Hi, Gu Yu!

I'm using cross-validation with LUNA16 dataset.

The dataset has 'candidate nodule list'.csv file.

My train and test patch extraction has the same process but uses difference .csv file.

Thanks for your question!

Best regards Bum-Chae Kim (bck8888)

[guyucowboy]

Hi, Bum-Chae Kim I read you excellent paper. Thanks for sharing your great work! I have some questions to ask you.

1. Do you mean that the function "test_patch_extraction" can be got by changing the "candidates_V2.csv" in the function "dodule_patch_extraction" to "candidates.csv"?
2. The function "mk_patch_origial_CNN" is not found in train_mode.py. How to extract the 3D patch especially when the patchs are near the boundaries which would cause out of the range of the CT scans?
3. I found that three scales ,"40 × 40 × 26","30 × 30 × 10",and "20 × 20 × 6" are extracted in the paper. Considering the different image spacings of the different scans, do you tried the following method? First, make CT scans isotropic with the resolution 1mm×1mm×1mm, or 0.5mm×0.5mm×0.5mm. Then, extract three kind of cubes with different three scales. But I do not know the extractly suitable scales.

Thanks for your help! Best regards Gu Yu

[bck8888]

Hi, Gu Yu

I’m focus the false positive reduction part in LUNA16.

So, my code is not include candidate nodule detection.

I’m separate patients using 5-fold cross-validation and then extract candidate nodules in "candidates_V2.csv”.

[1] So, training and test patients dataset use ‘nodule_patch_extraction’. ‘total_patch_extraction’ def is used for LUNA16 results file. Because, The ‘nodule_patch_extraction’ def has data augmentation code.

[2] I’m apologize the confusion code. The code is not complete version. I’m finish the code checking as soon as possible! and for check boundary, I use ‘TP_voxel_extraction’.

[3] Medical image dataset have a spacing issue. But, in my case, we did not change the spacing. Because that method changes the CT slice width and height size. I did not know how to prevent size change. So, I checked the LUNA dataset’s spacing values. The whole dataset’s spacing value ratio is x:y:z = 0.75:0.75:1. Then, I just extracted 3D patches using x:y:z = 1:1:0.7 ratio.

I considering recommended method for other project.

This is my answer to your question. I hope this helps you.

Best regards Bum-Chae Kim

2018. 8. 8. 12:56, guyucowboy notifications@github.com 작성:

Hi, Bum-Chae Kim I read you excellent paper. Thanks for sharing your great work! I have some questions to ask you.

Do you mean that the function "test_patch_extraction" can be got by changing the "candidates_V2.csv" in the function "dodule_patch_extraction" to "candidates.csv"? The function "mk_patch_origial_CNN" is not found in train_mode.py. How to extract the 3D patch especially when the patchs are near the boundaries which would cause out of the range of the CT scans? I found that three scales ,"40 × 40 × 26","30 × 30 × 10",and "20 × 20 × 6" are extracted in the paper. Considering the different image spacings of the different scans, do you tried the following method? First, make CT scans isotropic with the resolution 1mm×1mm×1mm, or 0.5mm×0.5mm×0.5mm. Then, extract three kind of cubes with different three scales. But I do not know the extractly suitable scales. Thanks for your help! Best regards Gu Yu

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/bck8888/MGICNN/issues/1#issuecomment-411276544, or mute the thread https://github.com/notifications/unsubscribe-auth/AMim8hSOOsSYU5ZkGkGPGwPz_rs6g8GQks5uOmFUgaJpZM4Vx0BM.

[guyucowboy]

Hi, Bum-Chae Kim Thanks for your help! I do appreciate all your great work. I am looking forward to your reply about the code checking at your earliest convenience. Thank you very much. Best regards Gu Yu

[guyucowboy]

Hi, Bum-Chae Kim.
How are you these days? Could you help me about code checking? Thanks you very much! Best regards Gu Yu