Opinions

Reply to: Free DICOM de-identification tools in clinical research: functioning and safety of patient privacy

by KYE Aryanto and PMA van Ooijen (k.y.e.aryanto@umcg.nl)

Aryanto KYE, Oudkerk M, van Ooijen PMA (2015) Free DICOM de-identification tools in clinical research: functioning and safety of patient privacy. Eur Radiol 25(12):3685–95.

Dear Editor,

First of all, we would like to thank Dr. Clunie – being a well-known expert in the area of DICOM – for his critical appraisal of our paper.

We fully agree that the de-identification of certain information in the DICOM header depends on the requirements as dictated by the research project. It could therefore indeed be necessary to
retain certain sensitive information in the DICOM header. However, we also noticed that it leads to extensive requests by sponsors to keep almost all information in case of the possibilities of their use in the future. Therefore, we would like to advocate the implementation of de-identification of the DICOM data as strict as possible.

We also fully agree with Dr. Clunie that overzealous de-identification rendering unusable data is something that could happen and this issue could have also been discussed in more detail in our paper to provide the readers with the full picture. The overzealous de-identification could be tackled, by the fact that the information is – in most clinical studies – also kept in a clinical research database. This database can vary – depending on the size of the study – from a simple spreadsheet to a full sized clinical research registration system. By only keeping the information in such a system and not also in the DICOM header makes the DICOM images ‘less sensitive’ since both sources are then required to get the full information. This could be a motivation to de-identify the imaging data as much as possible.

Dr. Clunie also raises the issue of the ability to change the configuration. Indeed we did not look into the ease with which configurations can be changed. Partly because we found when talking to clinical users that they rarely used any configuration, even when it was easy, because of the uncertainty of what would happen and the fact that they rely on the software to do a good job. This issue was not discussed in our paper since we did not investigate this formally and thus could not draw any sound conclusion on this. It would be an interesting addition study to look into this issue also with different tools and ask un-experienced clinical users to get to a certain setup with every tool and then determine if they can do that without much effort. We fully agree that the design of the software tools is a big challenge that should take into account both the experienced and in-experienced users.

The remark of the misquotation of the PixelMedDicomCleaner as not having a GUI is well taken as it does indeed come with a JAVA GUI.

We would like to thank Dr. Clunie again for his extensive response to our paper. It has provided us with extra insights and will also provide the reader with additional feedback from an expert in the field.

Kind regards, KYEAryanto and PM van Ooijen

References