Meeting Banner
Abstract #4133

B0-Correction and K-Means Clustering for Accurate and Automatic Identification of Regions with Reduced Apparent Diffusion Coefficient (ADC) in Advanced Cervical Cancer at the Time of Brachytherapy

Sren Haack1, Erik Morre Pedersen2, Mads Sloth Vinding3, Jesper Folsted Kallehauge4, Jacob Christian Lindegaard5, Kari Tanderup5, Sune Nrhj Jespersen6

1Dept. of Clinical Engineering, Aarhus University Hospital, Aarhus, Denmark; 2Dept. of Radiology, Aarhus University Hospital, Aarhus, Denmark; 3inSpin, iNANO, Aarhus University, Aarhus, Denmark; 4Dept. of Medical Physics, Aarhus University Hospital, Aarhus, Denmark; 5Dept. of Oncology, Aarhus University Hospital, Aarhus, Denmark; 6CFIN/Mindlab, Aarhus University, Aarhus, Denmark

Diffusion weighted MRI has shown great potential in diagnostic cancer imaging and may also have value for monitoring tumor response during radiotherapy. Uncertainties due to geometric distortions caused by B0-inhomogeneity and tumor delineation are major obstacles for implementing DWI for use in dose planning of radiotherapy. This study evaluates the use of k-means clustering for automatic user independent delineation of regions with reduced apparent diffusion coefficient (ADC) and the value of B0-correction of DW-MRI for reduction of geometrical distortions during dose planning of brachytherapy of advanced cervical cancer.

Abstract PDF E-Poster Video

Keywords

accurate acquisition advanced amplified apparent applicator automatic benefit better biased calculation cancer cervical clinical clustering coefficient collected combined computed contour contouring contours corrected correcting correction created cyan delineation delivery dept detection diagnostic diffusion displacement displacements distortions dose dual eight engineering ensure essential evaluated evaluates examinations feasible field geometric geometrical gradients great green gross hospital identification important improve improved improvement improves independent index informed inhomogeneity input inside intermediate leading locally mads making maps marked match materials medical mismatch monitoring objective obstacle oncology orange outlines overcome overlap paired patients pave percentage physics pixel pixels place planning plastic position potential promises pronounced radiology radiotherapy reduced reduces reduction respectively response risk sensitivity significantly similarity sloth solid spatial steep students symphony tandem target tissue towards treated tumor user volume volumes