search

matthieukomorowski / AI_Clinician

Reinforcement learning for medical decisions
95 stars 45 forks source link

Indices for item_id relabling in `AIClinician_sepsis3_def_160219.m`? #2

Open moberst opened 5 years ago

moberst commented 5 years ago

On lines 181-185, the item IDs (which can get very large, e.g., 227467) in third column of labU are replaced by the appropriate column number in reformat (which goes up to 68)

for i=10001:size(labU,1)
[~,locb]=ismember(Reflabs,labU(i,3));
labU(i,3)=find(max(locb')');
end
toc

However, this process does not start at the first row of labU, but instead starts at row 10001. Is there a reason for this, or is it a bug?

For context, when I run the script as written, the program crashes on lines 301-305

%LAB VALUES
    ii=temp2(:,2)==t(i);
    col=temp2(ii,3);
    value=temp2(ii,4);
reformat(irow,31+col)=value;

Because temp2 is copying portions of labU, and it ends up using numbers like 227467 as the column number of reformat, which then tries to allocate all those extra columns in memory. I checked, and this is only a problem in the first 10,000 rows, which got skipped by the itemid -> column script in lines 181-185

paulrich1234 commented 5 years ago

hi,moberst,thank you for your contributes for this ,and it works for me to this error ,and i have run a csv result after i run the AIClinician_mimic3_dataset_160219.m file and save the MIMICtable = reformat4t; as a csv file .which is differ from the data that the reporisity:

bloc icustayid charttime gender age elixhauser re_admission died_in_hosp mortality_90d Weight_kg ... mechvent SOFA SIRS median_dose_vaso max_dose_vaso input_total_tev input_4hourly_tev output_total output_4hourly cumulated_balance_tev 0 1 3 7245052800 0 17639.826435 0.0 0 0 1 77.5 ... 0 9 4 0.0 0.0 3500.0 2100.0 230.0 230.0 3270.0 1 2 3 7245067200 0 17639.826435 0.0 0 0 1 77.5 ... 0 8 3 0.0 0.0 5499.0 1999.0 697.0 467.0 4802.0 2 3 3 7245081600 0 17639.826435 0.0 0 0 1 77.5 ... 0 7 3 0.0 0.0 5604.0 105.0 2302.0 1605.0 3302.0 3 4 3 7245096000 0 17639.826435 0.0 0 0 1 77.5 ... 0 6 3 0.0 0.0 5664.0 60.0 2922.0 620.0 2742.0 4 5 3 7245110400 0 17639.826435 0.0 0 0 1 77.5 ... 0 5 3 0.0 0.0 5724.0 60.0 3352.0 430.0 2372.0

i have checked icustayid=3 and corresponding features there are some difference between the result i have got . is is normal ?