Commissioning and validation of commercial deformable image registration software for adaptive contouring
| Autor: | Deepak D. Deshpande, Kishore Joshi, Swamidas V. Jamema, Siji Nojin Paul, Reena Phurailatpam |
|---|---|
| Rok vydání: | 2017 |
| Předmět: |
Similarity (geometry)
Shape change Computer science Biophysics General Physics and Astronomy Image registration Imaging phantom 030218 nuclear medicine & medical imaging 03 medical and health sciences 0302 clinical medicine Software Image Processing Computer-Assisted Humans Radiology Nuclear Medicine and imaging Computer vision Ground truth Contouring 3d slicer business.industry Phantoms Imaging General Medicine 030220 oncology & carcinogenesis Artificial intelligence business Tomography X-Ray Computed |
| Zdroj: | Physica medica : PM : an international journal devoted to the applications of physics to medicine and biology : official journal of the Italian Association of Biomedical Physics (AIFB). 47 |
| ISSN: | 1724-191X |
| Popis: | Purpose To report the commissioning and validation of deformable image registration(DIR) software for adaptive contouring. Methods DIR (SmartAdapt®v13.6) was validated using two methods namely contour propagation accuracy and landmark tracking, using physical phantoms and clinical images of various disease sites. Five in-house made phantoms with various known deformations and a set of 10 virtual phantoms were used. Displacement in lateral, anterio-posterior (AP) and superior-inferior (SI) direction were evaluated for various organs and compared with the ground truth. Four clinical sites namely, brain (n = 5), HN (n = 9), cervix (n = 18) and prostate (n = 23) were used. Organs were manually delineated by a radiation oncologist, compared with the deformable image registration (DIR) generated contours. 3D slicer v4.5.0.1 was used to analyze Dice Similarity Co-efficient (DSC), shift in centre of mass (COM) and Hausdorff distances Hf95%/avg. Results Mean (SD) DSC, Hf95% (mm), Hfavg (mm) and COM of all the phantoms 1–5 were 0.84 (0.2) mm, 5.1 (7.4) mm, 1.6 (2.2) mm, and 1.6 (0.2) mm respectively. Phantom-5 had the largest deformation as compared to phantoms 1–4, and hence had suboptimal indices. The virtual phantom resulted in consistent results for all the ROIs investigated. Contours propagated for brain patients were better with a high DSC score (0.91 (0.04)) as compared to other sites (HN: 0.84, prostate: 0.81 and cervix 0.77). A similar trend was seen in other indices too. The accuracy of propagated contours is limited for complex deformations that include large volume and shape change of bladder and rectum respectively. Visual validation of the propagated contours is recommended for clinical implementation. Conclusion The DIR algorithm was commissioned and validated for adaptive contouring. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect