Office of Technology Transfer – University of Michigan

Subvolume Identification from DCE MRI for Prediction of Treament Response

Technology #5352

Questions about this technology? Ask a Technology Manager

Download Printable PDF

Categories
Researchers
Yue Cao
Managed By
Drew Bennett
Associate Director - Software Licensing 734-615-4004
Patent Protection
US Patent 9,370,304

Prediction of therapy on cancer

The impact of therapy on cancer as well as normal organs is hard to predict by conventional means, including tumor size, mean value of physiological image parameters in the tumor, or even voxel-by-voxel analysis of physiological images taken before and during treatment. Additional information for tumor response to therapy might be obtained through physiological parameters, which however have to be quantified from dynamic images using pharmacokinetic models (e.g., fitting dynamic contrast enhanced (DCE) MRI to a Toft model) and thus are time-consuming and require expert knowledge.

Automatic principal sub-volume identification

A research group in the Radiology and Medical School of University of Michigan has reported a technology that finds aggressive subvolumes of tumors from DCE MRI without pharmacokinetic modeling and via automated processing, the presence of which predicts for local treatment outcome (failure or response). The technology conducts the principal component analysis (PCA) directly on the DCE MRI data to extract and analyze most important and predictive physiological parameters for prediction of tumor response to therapy. Then it applies probabilistic classification and clustering method to calculate sub-volumes of volume-weighted physiological parameters. Such an idea has been evaluated on the DCE MRI data from a group of patients with various cancers. As a result, the technology is far more robust than existing 2-step analysis of DCE MRI to derive physiological parameter and to apply point-by-point (“voxel level”) analysis. Also, it improves the robustness of analysis for imaging of patients with various underlying physiology and imaged on MRI and other scanners.

Applications

  • Early prediction of systemic as well as focal cancer therapies
  • Calculation of aggressive sub-volumes of tumors from DCE MRI without pharmacokinetic modeling and via automated processing

Advantages

  • Automatic principal physiological parameter finding without pharmacokinetic modeling
  • More robust and reliable than existing voxel-level physiological analyses
  • Improved robustness of analysis with various underlying physiology and scanners