MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Quantification of Intrafraction Prostate Motion Using Detected Features in Sagittal 2D Cine-MR

B Strbac*, C Brouwer, S Both, J Langendijk, D Yakar, S Al-uwini,

Presentations

(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose:
To quantify intrafraction motion of different regions of prostate and seminal vesicles (P+SV) using cine-MR.


Methods:
For nine consecutive prostate cancer patients, 2D sagittal cine-MR was acquired for a median of 8 minutes (range 5-11 min) with a frame rate of 1.8 frames/sec. The characteristic features were recognized on the first frame by the Harris corner detector and tracked by Kanade-Lucas-Tomasi algorithm. The time-series were generated from features tracks. The spectral clustering of the time series was used to capture different motion patterns in the different regions of P+SV The centroids of clustered time series were used for motion description of apex, base, anterior, posterior, medium part of the prostate and seminal vesicles. The displacements probabilities on the last frame and maximum displacement probabilities during full cine-MR were calculated.


Results:
The patterns in time-series showed distinctive sets of movement in different parts of P+SV. The posterior part of the prostate and SV were more sensitive to rectum movement. Anterior part, adjacent to the pubic bone, base, and apex of the prostate were less affected by rectum movement and hence less prone to deviate from the initial position. The full bladder had non-significant interference on the position of P+SV.
The probability that displacement of P+SV is within 5 mm on the last frame was 100% for all patients. The probabilities of maximum displacements to be in 5mm during the whole cine MR time for different regions of P+SV over patients population were: posterior=87%, anterior=96%, base=93%, apex=99%, medium part=93% and SV=82 %

Conclusion:
This method describes probable distribution of displacements in different regions of P+SV using cine-MR. This enables more accurate calculation of margins needed for SBRT and proton therapy treatments, more sensitive to prostate motion induced by changes in rectum and bladder

Keywords

Computer Vision, Organ Motion, MRI

Taxonomy

IM- MRI : Machine learning, computer vision

Contact Email