Radiation and Cancer Physics

PD 13 - Physics 8 - Poster Discussion - Outcome Analysis and Response Imaging

1113 - Radio-Morphology: Parametric Shape-Based Features for Outcome Prediction in Radiation Therapy

Tuesday, October 23
5:15 PM - 5:21 PM
Location: Room 217 C/D

Radio-Morphology: Parametric Shape-Based Features for Outcome Prediction in Radiation Therapy
P. Lakshminarayanan1, W. Jiang2, S. P. Robertson3, Z. Cheng4, P. Han1, M. R. Bowers1, J. Moore1, H. Quon1, R. Taylor2, and T. R. McNutt1; 1Department of Radiation Oncology and Molecular Radiation Sciences, Johns Hopkins University School of Medicine, Baltimore, MD, 2Johns Hopkins University, Baltimore, MD, 3WellSpan Health York Cancer Center, York, PA, 4Johns Hopkins Medicine, Baltimore, MD

Purpose/Objective(s): Current methods of characterizing dose distributions are limited by the assumption of physiologic homogeneity within a region of interest (ROI) and do not provide a spatially aware description of dose. A practice termed radio-morphology (RM) is proposed as a method to apply anatomical knowledge to parametrically derive new shapes and substructures from a normalized set of anatomy, ensuring consistently identifiable features are produced. The goal of this study was to design a method that can identify characteristics of a dose distribution, at a higher resolution than organ-level dose-volume histograms (DVHs), that are predictive of post-treatment outcomes.

Materials/Methods: The RM feature generation pipeline consists of three steps: anatomy normalization, image transformation, and dose feature extraction. Anatomy normalization is the process where anatomical structures are deformed to a standard coordinate frame, allowing ROIs to be compared across a patient population. Examples of methods include coherent point drift, when images are available as point sets, and optical flow, for gray level images. Next, geometric transformations, such as scaling and partitioning, are applied to derive new sub-structures. Finally, radiation dose is mapped onto the new shapes, and dose statistics are extracted. Using data from a learning health system database, a variety of RM features were generated using this pipeline at multiple anatomical sites. For the prostate, RM shape transformations were used with 115 patients to approximate the shapes containing the neuro-vasculature surrounding the prostate. Dose statistics and demographic information of these regions were used to understand the effect of the spatial distribution of radiation on patient reported sexual function. In the head and neck region, salivary glands in 427 patients were examined using both voxel-based analysis and shape-based substructures to identify regions that are predictive of the development of high grade xerostomia (grade 2+ at 3-6 months post-treatment).

Results: The RM pipeline produced feature sets that help uncover the importance of the spatial distribution of dose to the anatomy. High dose to the inferior region of the prostate and the lower urethral sphincter was linked to loss of sexual potency. High dose to the superior-anterior region of the contralateral parotid gland was found to have the greatest effect on the development of high-grade post-treatment xerostomia.

Conclusion: The results of these experiments support the use of the RM pipeline to model the effect of radiation dose on post-treatment clinical outcomes. By parameterizing feature generation, it is possible to iteratively optimize the features and methodically identify key regions of the anatomy. With modern databases storing full dose distributions, patient images, and clinical assessments, the proposed parametric feature generation pipeline can rapidly lead to new insight about the quality of radiotherapy plans.

Author Disclosure: P. Lakshminarayanan: None. W. Jiang: None. S.P. Robertson: None. P. Han: None. M.R. Bowers: None. H. Quon: None. R. Taylor: None. T.R. McNutt: Research Grant; Toshiba, Philips Radiation Oncology Systems, Elekta Oncology Systems. Patent/License Fees/Copyright; Sun Nuclear, Accuray-Tomotherapy. President Elect; AAPM-MAC.

Pranav Lakshminarayanan, MS, BS

Johns Hopkins Hospital: Senior Programmer Analyst: Employee


Send Email for Pranav Lakshminarayanan


1113 - Radio-Morphology: Parametric Shape-Based Features for Outcome Prediction in Radiation Therapy

Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Radio-Morphology: Parametric Shape-Based Features for Outcome Prediction in Radiation Therapy