Jaydev Dave, Presenter: Nothing to Disclose

Eric Laurence Gingold PhD, Abstract Co-Author: Nothing to Disclose

An algorithm to identify and extract radiation dose metadata from DICOM image-based dose sheets was developed.  Dose index information can be mined from a PACS database for quality monitoring and trend analysis

The ability to monitor and track radiation dose from CT procedures is an area of high current interest. Efforts are being directed towards standardized dose index reporting.   Currently available methods include optical character recognition (OCR) applied to bitmapped dose reports, and extraction of DICOM metadata.  OCR methods are prone to error.

CT files acquired with Philips (Cleveland, OH) Brilliance 64, iCT 256, and Big Bore  scanners and a Siemens (Malvern, PA) Biograph 6 PET/CT scanner were used. An algorithm was developed using Matlab (Mathworks, Natick, MA, USA) to automatically select DICOM image-based dose sheets from amongst all other DICOM image files and reports present in an image archive. For Philips and Siemens scanners this was accomplished by searching for the DICOM field ‘CommentsOnRadiationDose’ and for series ‘501’, respectively. For the identified dose information files from the Philips scanners, the ‘ExposureDoseSequence’ field in the metadata contained a structure within a structure for each exposure sequence containing many acquisition parameters,  such as kilovoltage peak, exposure time and current, filter type and material, scan length, collimation width and CTDI – this data was extracted using a recursive loop. For these files from the Siemens scanners, the dose information was found to be present in a field ‘Private_0029_1110’ in the form of integers used for character coding in a single array – this data was also extracted using a recursive loop after decoding. The algorithm then sorted the data based on individual series for each accession number and made this data available in the form of a Microsoft Excel sheet for data interpretation and analysis.

DICOM dose data was successfully extracted from all investigated dose files. Data mining of the metadata within DICOM image-based bitmap dose sheets provides a direct access to the required dose data, eliminating potential error involved in OCR.

Dave, J, Gingold, E, Extraction of CT Dose Information from DICOM Metadata.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11016606.html