Program Information
Evaluation of Image Registration Metrics to Quantify Mid-Treatment Anatomic Changes for Locally Advanced Lung Cancer Patients
J Kavanaugh*, Z Ji , H Li , Washington University School of Medicine, Saint Louis, MO
Presentations
WE-RAM2-GePD-JT-5 (Wednesday, August 2, 2017) 10:00 AM - 10:30 AM Room: Joint Imaging-Therapy ePoster Theater
Purpose: A preliminary study to evaluate the efficacy of using intensity-based image registration similarity metrics to quantify mid-treatment anatomic changes for locally advanced lung cancer.
Methods: Eight patients treated with IMRT for stage III locally advanced lung cancer were included in the retrospective study and were evenly divided into two groups: those exhibiting significant mid-treatment anatomic changes and those with minimal anatomic changes. Significance of anatomic changes was determined by a radiation oncologist during treatment and resulted in an adaptive IMRT plan in all 4 cases. Weekly cone beam CT (CBCT) imaging was exported from the record and verify system (Mosaiq, Elekta Inc.) and processed, registered, and evaluated using an in-house developed program. The clinical PTV was used to delineate voxels in each CBCT to be included in calculating the normalized mean squared error (NMSE) similarity metric, with independent values quantified on each axial slice. The maximum NMSE across all axial slices for each weekly CBCT was recorded for each patient and the average weekly NMSE for each group was calculated. Significance between the average NMSE for each patient group was determined using a two-tailed equal variance Student’s t-test (p = 0.05).
Results: The NMSE was significantly larger over the first 4 weeks for the group exhibiting large anatomic changes, with a maximum p-value of 0.0051 occurring during week 3. No comparison was conducted for weeks 5 and 6 as all 4 patients with anatomic changes were resimulated.
Conclusion: Intensity-based image registration similarity metrics provide the potential to easily quantify the anatomic changes observed in the 3D CBCT imaging for locally advanced stage III lung cancer. Daily automated tracking of anatomic changes using these metrics may provide quantifiable data to generate machine learning models that predict when a patient may benefit from adaptive radiation therapy.
Funding Support, Disclosures, and Conflict of Interest: This work is funded in part by a Varian Medical System Master Research Agreement grant.
Contact Email: