chl8856 / dynamic-deephit Goto Github PK

Hello,

I was wondering if this is normal that the likelihood for censored and uncensored data is computed the same way, shouldn't tmp1 and tmp2 (lines 233 and 237) be different ?
I think the current way is correct for uncensored data but I might have missed something ?

Dear Dr. Lee,

I would like to confirm that "get_main_cf.py" was not used/deployed during the modeling process (when I replicate the package with different dataset, seems like this python script was not called). Thanks!

Hi Dr. Lee,

Per your research article, you have shown the work on providing post-processing statistics that generates clinical insights. I will appreciate if you can kindly post the codes here (output individual patients' top influencing variables based on trained model and risks).

Regards.

In the implementation and experiments, data normalization/standardization is done before splitting into train and test set, that could result in test set information leaked into training set...

Hi,
I was trying to figure out how I could use this code on a project of mine. I think I figured out most of it but couldn't find where in the code you define T_max (the maximum age person can attain). Is it the num_Category parameter? This seems to be closest to the interpretation of T_max in the paper but I'm not sure. The assumption then is that adding 20% to the max observed age of the oldest person in the dataset is the maximum age that any future person can attain? In the paper you say that you set T_max to 1200 (100 years in months) but I couldn't find this anywhere in the code.
thanks in advance.

Dear Changhee Lee, we have an issue with the code you provide as it was written with tf1.13 and tf after 2.0 does not support the tensorflow.contrib library anymore. Do you have an updated code to try your algorithm?
Thanks in advance

Dear author/To whom it may concern,

I hope my message finds you well. I ran into problem when I wanted to apply the code in this repo on a "survival" dataset where each patient has "only one" follow-up that contained patient baseline characteristics - essentially, what I wanted to do was to test if the DDH model could be used to predict patient survival time with patient baseline characteristics, like conventional survival models (eg. Cox) did. The code was terminated with the following error:

TensorArray has size zero, but element shape [?,x] is not fully defined. Currently only static shapes are supported when packing zero-size TensorArrays.

(x is a positive integer and probably has something to do with my dataset)

This error ONLY occured when the training dataset contained exactly one and only one follow-up for each patient. Adding even just one repetitive measurement to the training set would resolve the error. I think this might be something to do with how the training set was structured in the Tensorflow Array, but I have very limited experience with Tensorflow-based deep learning and there was no much information about this error on the internet.

Any help regarding my question would be greatly appreciated.

Greetings, Dr. Lee,

Thank you so much for sharing the dev code here. I was looking into the "tutorial.ipynb" notebook, and in section 5. Test the Trained Network; I was wondering, what's the correct syntax to print/output the forecasted results (individual patients' survival rates)?

Thank you.

Dear authors,

The results in the research paper include a ranking for the covariates in the study concerning the CF dataset. However, in this git, the method/code to obtain said results is not in any of the scripts.

I'm currently conducting a research project for the University of Navarra and I'm interested in applying this method to obtain covariate incites. If possible, can the code you have implemented to obtain these results be uploaded?

Kind regards.

Dear Dr. Lee,

Please refer to "tutorial.ipynb" notebook, section "1. Import Dataset", the following pred_time and eval_time were defined.

# This must be changed depending on the datasets, prediction/evaliation times of interest
pred_time = [52, 352, 552] # prediction time (in months)
eval_time = [12, 36, 60, 120] # months evaluation time (for C-index and Brier-Score)

I would like to confirm the logic on providing these two windows.
(1) pred_time outputs 52th, 352th, and 552th month risk for each patient
(2) eval_time scores the risk of each patient at 12th, 36th, 60th, and 120th month
(3) the C-index is computed as a 3-by-4 matrix, can you illustrate what does (pred_time == 52 & eval_time==12), the C-index imply?

Thank you so much!