It was brought to our attention that there is an issue with the intensity values in the test set imaging. In particular, each scan is offset from its true range of HU values, in some cases by several thousand units.
If you'd like to apply the offset on the data downloaded previously, you can download a CSV file [here](https://kits21-training.s3.us-east-2.amazonaws.com/offset_sheet.csv) with the offset values to **add** in order to achieve the correct values. If you'd like to simply download the corrected images, please use [this corrected download link](https://kits21-training.s3.us-east-2.amazonaws.com/knight_test_data_corrected.tar).
I am truly sorry for the inconvenience this may have caused.
Created by Nicholas Heller helle246 Hello,
How are you reading the files? I don't see this problem either with the the corrected images or original ones and offset from the CSV.
The offset in the original images for these cases is indeed large, but the corresponding value in the CSV is also large (with negative sign), so adding it results in a reasonable value.
I read the images in Python like this:
```
import SimpleITK as sitk
data_itk = sitk.ReadImage(f) # read image
data_npy = sitk.GetArrayFromImage(data_itk).astype(np.float32) # convert to numpy array
```
and then just add the corresponding offset value to data_npy
Hi
In the corrected dataset, we feel images corresponding to case_00419 and case_00425 have not been corrected yet and have huge offset. We tried to correct it with the offset value from the csv file, but still the image fails to load when the code runs.
Please help us out with this.
Regards,
Varsha
Drop files to upload
IMPORTANT: Issue with image intensity values page is loading…