Characterization of tissue using non-invasive in vivo imaging techniques is used for detection and measurement of disease burden in oncology. Researchers have developed numerous algorithms, such as neural networks, and classification techniques to improve the characterization (morphological changes) of tissue. Unfortunately, to obtain statistical significance, large datasets are a requirement in this research endeavor due to tumor heterogeneity within the same histologic classification. Pre-clinical patient derived xenograft animal models can be a significant resource by providing collections with a more homogenous tumor genome across the collection with companion genomic and pathologic characterization available (https://pdmr.cancer.gov/), allowing determination of the variability of imaging characteristics.

This dataset of a patient derived xenograft model (below table) can be used for training algorithms for evaluating variations in tissue texture with respect to tumor growth and cancer model.

PDX Model Characterizations and Biweekly imaging sessions 

In this study we performed non-contrast non-gated T2w MRI (SOP50101_MRI), initiated 2 weeks post implantation, and continued biweekly imaging sessions until their tumors reached a size requiring humane termination (ACUC guidance > 2 cm in any linear dimension by caliper or MRI measurement) or their clinical status required euthanasia. Fragments (2x2x2 mm³) from the NCI/DCTD PDMR repository were implanted into 5-10 donor mice (NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG)). When tumors reached enrollment criteria (100 – 300 mm³), tumors were excised, cut into 2x2x2 mm³ fragments and implanted with Matrigel (per PDMR SOP50101_Tumor Implantation) into NSG study mice. The multi-mouse non-gated DICOM dataset was split according to the method published in Tomography and retained their individual mouse DICOM header information.  Structured Reports (SR) were added to the dataset to include fragment implant date, CTEP description, mouse strain (NSG) and model.

The genomic and pathologic characteristics of these models, which is available from the National Cancer Institute Patient-Derived Models Repository (https://pdmr.cancer.gov/), can be used in conjunction with this publicly available dataset to guide the development of algorithms for enhanced characterization of tissue for precision medicine.

Summary:

This collection contains serial non-contrast T2w MRI of 18 patient derived xenograft cancer models (514 images) for researchers to develop various algorithms using neural networks, and classification techniques to improve tissue characterization (morphological changes). The genomic and pathologic characteristics of these models, which is available from the National Cancer Institute Patient-Derived Models Repository (https://pdmr.cancer.gov/), can be used in conjunction with this publicly available dataset to guide their development of algorithms for enhanced characterization of tissue for precision medicine.