Radiation and Cancer Physics

SS 29 - Physics 9 - Imaging for Treatment Planning

210 - Voxel-Based Texture Analysis of Multiparametric MRI for Intraprostatic Tumor Volume Delineation

Tuesday, October 23
5:25 PM - 5:35 PM
Location: Room 304

Voxel-Based Texture Analysis of Multiparametric MRI for Intraprostatic Tumor Volume Delineation
J. K. Lee, C. Liu, E. Carver, M. A. Elshaikh, and N. Wen; Henry Ford Health System, Detroit, MI

Purpose/Objective(s): Quantitative texture feature analysis of multiparametric magnetic resonance imaging (mpMRI) has the potential to increase the accuracy of tumor volume delineation in prostate cancer. This may help identify an optimal target volume for intraprostatic radiotherapy boost. We present data on the level of concordance between tumor volumes in T2-weighted (T2W) and apparent diffusion coefficient (ADC) images, including a novel approach of voxel-based calculations to assess the level of correlation between tumor foci in these MRI sequences.

Materials/Methods: Robust mpMRI (T2W and diffusion-weighted imaging [DWI]) data from 14 patients with MRI-guided biopsy-proven prostate cancer were obtained from the SPIE-AAPM-NCI Prostate MR Classification Challenge. All images were acquired using 3T MR scanners without an endorectal coil. T2W images were acquired using a turbo spin echo sequence, whereas the DWI series were acquired with a single-shot echo planar imaging sequence. The ADC map was calculated by the scanner software. Tumor foci were identified and contoured in the T2W and ADC images. The ADC map was resampled to a standardized voxel size of 0.5 mm x 0.5 mm x 3.0 mm using linear interpolation to match the T2W images. The two image sets were then co-registered for analysis. The Hausdorff distance (HD), mean distance-to-agreement (MDA), and Dice and Jaccard coefficients were calculated. A boolean sum volume (BSV) was defined as a combination of the contours from both image modalities for each patient. Fractional ranks were obtained for each voxel of the BSV and the Spearman correlation was calculated.

Results: Tumor foci were located in the peripheral zone, transition zone, and anterior fibromuscular stroma in 5 (36%), 5 (36%), and 4 (28%) patients, respectively. Data obtained for the tumor foci in the T2W and ADC images were: Mean tumor volume of 2.74 +/- 2.83 mL and 2.69 +/- 2.73 mL, respectively; mean HD and MDA of 4.58 +/- 1.61 mm and 1.11 +/- 0.53 mm, respectively; and Dice and Jaccard coefficients of 0.72 +/- 0.12 and 0.58 +/- 0.15, respectively. Mean voxel-based Spearman correlation between T2W and ADC was 0.19 +/- 0.17.

Conclusion: There was a high level of concordance between tumor foci in the T2W and ADC images, but a voxel-based Spearman correlation did not demonstrate a clear relationship between these two sequences. Further research incorporating a larger sample size may yield a better understanding of the voxel-based correlation between mpMRI in different zones and grades of tumor foci, and help define an optimal intraprostatic tumor volume for radiotherapy boost. This work was supported by a Research Scholar Grant, RSG-15-137-01-CCE from the American Cancer Society. The dataset was collected and curated for research in computer-aided diagnosis of prostate MR under supervision of Dr. Huisman as documented in: G. Litjens, O. Debats, J. Barentsz, N. Karssemeijer and H. Huisman. "Computer-aided detection of prostate cancer in MRI", IEEE Transactions on Medical Imaging 2014;33:1083-1092.

Author Disclosure: J.K. Lee: None. C. Liu: None. E. Carver: None. M.A. Elshaikh: None. N. Wen: Research Grant; American Cancer Society.

Joon Lee, MD

Disclosure:
No relationships to disclose.

Presentation(s):

Send Email for Joon Lee


Assets

210 - Voxel-Based Texture Analysis of Multiparametric MRI for Intraprostatic Tumor Volume Delineation



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 Voxel-Based Texture Analysis of Multiparametric MRI for Intraprostatic Tumor Volume Delineation