Automatically tracking and detecting significant nodal mass shrinkage during head-and-neck radiation treatment using image saliency Conference Paper
| Authors: | Hu, Y. C.; Polvorosa, C.; Tsai, C. J.; Hunt, M. |
| Title: | Automatically tracking and detecting significant nodal mass shrinkage during head-and-neck radiation treatment using image saliency |
| Conference Title: | 1st International Workshop - Artificial Intelligence in Radiation Therapy |
| Abstract: | Large nodal masses shrink during head-and-neck radiation treatment. If the shrinkage is dramatic, nearby organs at risk (OARs) may receive potentially harmful radiation dose. In an institutional IRB-approved protocol, patients were monitored with weekly T2-weighted MRIs. Gross tumor volumes (GTV) from pre-treatment MRI were propagated to weekly MRIs via deformable image registrations (DIR) for tracking the change of GTV nodal volume and detection of significant shrinkage. This detection method, however, becomes problematic when a significant amount of the nodal mass dissolves during treatment, invalidating the assumption of correspondence between images for accurate deformable registration. We presented a novel method using image saliency to detect whether a involved nodal volume becomes significantly small during the treatment. We adapted a multi-resolution pyramid method and introduced symmetry in calculating image saliency of MRI images. The ratio of mean saliency value (RSal) from the propagated nodal volume on a weekly image to the mean saliency value of the pre-treatment nodal volume was calculated to assess whether the nodal volume shrank significantly. We evaluated our method using 94 MRI scans from 19 patients enrolled in the protocol. We achieved AUC of 0.97 in detection of significant shrinkage (smaller than 30% of the original volume) and the optimal RSal is 0.698. © Springer Nature Switzerland AG 2019. |
| Keywords: | magnetic resonance imaging; radiation; radiotherapy; medical imaging; tumors; artificial intelligence; deformation; radiation treatments; gross tumor volume; deformable registration; shrinkage; adaptive radiation therapy (art); deformable image registration; adaptive radiation therapies; image saliency; nodal tumor shrinkage; automatically tracking; image saliencies; resolution pyramids |
| Journal Title | Lecture Notes in Computer Science |
| Volume: | 11850 |
| Conference Dates: | 2019 Oct 17 |
| Conference Location: | Shenzhen, China |
| ISBN: | 0302-9743 |
| Publisher: | Springer |
| Date Published: | 2019-01-01 |
| Start Page: | 18 |
| End Page: | 25 |
| Language: | English |
| DOI: | 10.1007/978-3-030-32486-5_3 |
| PROVIDER: | scopus |
| DOI/URL: | |
| Notes: | Part of the MICCAI 2019 Conference -- Source: Scopus |
Altmetric
Citation Impact
BMJ Impact Analytics
Related MSK Work