Image Added

This dataset was collected by a collaboration of researchers from Children’s Wisconsin, Marquette University, Varian Medical Systems, Medical College of Wisconsin, and Stanford University as part of a project funded by the National Institute of Biomedical Imaging and Bioengineering (U01EB023822) to develop tools for rapid, patient-specific CT organ dose estimation. The collection consists of CT images in DICOM format of 359 pediatric chest-abdomen-pelvis or abdomen-pelvis exams acquired from three CT scanners.  The The datasets represent random pediatric cases based upon routine clinical indications. Each dataset contains expert contours of up to twentyof up to twenty-nine structures in DICOM RTSS format.  Some datasets are missing structures that are not in the scan range or that, in younger patients, could not be reliably identified.  Patient Patient ages range from 5 days to 16 to 16 years, with a mean age of 7 and with a near equal 7 and with a near equal distribution of male male (180)  and female and female (179)  patientspatients. The CT acquisition protocols and reconstruction methods vary across the scanner models and patient sizes, with specifications available in the DICOM headers. This data can be used to develop autosegmentation methods for radiation therapy, CT dosimetry, CT diagnostic algorithms, or other applications. The metadata of each CT image series contains the correct patient age and the height and weight data when available.

The native slice thickness for the acquired images was 0.625 mm for the GE scanners and 0.6 mm for the Siemens scanners.  Sixty-two datasets were manually contoured at this native slice thickness.  However, this process required extensive manual labor and was also challenged by high noise in the thin slices.  Therefore, the subsequent 297 datasets were reformatted to 2.0-mm slice thickness using a cubic spline interpolation algorithm prior to contouring. For some datasets, this interpolation caused artifacts in the most inferior or superior slices in the volume.  This is a known limitation of this dataset and users may need to disregard these corrupted slices, depending on the intended application.