Exporting two segmentation nodes in a row in the same output directory overwrite the prevous one

[jcfr]

Since the DICOM file name is initialized using the current date and time, exporting in the same directory overwrites the previous file.

QuantitativeReporting/DICOMPlugins/DICOMSegmentationPlugin.py

Line 351 in 87852c8

segFileName = "subject_hierarchy_export.SEG" + exporter.currentDateTime + ".dcm"

QuantitativeReporting/DICOMPlugins/DICOMSegmentationPlugin.py

Lines 437 to 440 in 87852c8

	@property
	def currentDateTime(self, outputFormat='%Y-%m-%d_%H%M%S'):
	from datetime import datetime
	return datetime.now().strftime(outputFormat)

[fedorov]

Good catch! We should probably change the exporter to initialize the name of the file to the UID of the instance.

[lassoan]

Instance UID can be too long to be used as filename. We often had DICOM import failures when used DICOM UIDs directly for folder and file names due to the complete file path becoming longer than maximum path length.

When we create temporary filenames based on timestamp in Slicer, we typically add a 3-4 digit random number to the timestamp to keep the filename relatively short but prevent name clashes.

For cases when we want to separate DICOM files based on UIDs, we use an md5 hash, as it provides good separation but keeps the file path length in control. It is usable even when we build a folder structure from patient, study, series, instance ID.

Note that DICOM export plugins generally expect to get an empty output folder. For example, if I remember correctly, the DICOM RT export plugin uses image0001, image0002,... output filenames.