Encrypted login | home

Program Information

Prediction of Pelvic Nodal Coverage Using Mutual Information Between Cone-Beam and Planning CTs

no image available
S Jani

S Jani*, A Kishan , D O'Connell , C King , M Steinberg , D Low , J Lamb , University of California, Los Angeles, Los Angeles, CA

Presentations

SU-E-J-179 Sunday 3:00PM - 6:00PM Room: Exhibit Hall

Purpose: To investigate if pelvic nodal coverage for prostate patients undergoing intensity modulated radiotherapy (IMRT) can be predicted using mutual image information computed between planning and cone-beam CTs (CBCTs).

Methods: Four patients with high-risk prostate adenocarcinoma were treated with IMRT on a Varian TrueBeam. Plans were designed such that 95% of the nodal planning target volume (PTV) received the prescription dose of 45 Gy (N=1) or 50.4 Gy (N=3). Weekly CBCTs (N=25) were acquired and the nodal clinical target volumes and organs at risk were contoured by a physician. The percent nodal volume receiving prescription dose was recorded as a ground truth. Using the recorded shifts performed by the radiation therapists at the time of image acquisition, CBCTs were aligned with the planning kVCT. Mutual image information (MI) was calculated between the CBCT and the aligned planning CT within the contour of the nodal PTV. Due to variable CBCT fields-of-view, CBCT images covering less than 90% of the nodal volume were excluded from the analysis, resulting in the removal of eight CBCTs.

Results: A correlation coefficient of 0.40 was observed between the MI metric and the percent of the nodal target volume receiving the prescription dose. One patient’s CBCTs had clear outliers from the rest of the patients. Upon further investigation, we discovered image artifacts that were present only in that patient’s images. When those four images were excluded, the correlation improved to 0.81.

Conclusion: This pilot study shows the potential of predicting pelvic nodal dosimetry by computing the mutual image information between planning CTs and patient setup CBCTs. Importantly, this technique does not involve manual or automatic contouring of the CBCT images. Additional patients and more robust exclusion criteria will help validate our findings.




Contact Email: