Semi-automatic delineation using weighted CT-MRI registered images for radiotherapy of nasopharyngeal cancer

• Published in: Medical physics (Lancaster) Vol. 38; no. 8; p. 4662
• Main Authors: Fitton, I, Cornelissen, S A P, Duppen, J C, Steenbakkers, R J H M, Peeters, S T H, Hoebers, F J P, Kaanders, J H A M, Nowak, P J C M, Rasch, C R N, van Herk, M
• Format: Journal Article
• Language: English
• Published: United States 01.08.2011
• Subjects: Aged; Algorithms; Humans; Magnetic Resonance Imaging - statistics & numerical data; Middle Aged; Nasopharyngeal Neoplasms - diagnostic imaging; Nasopharyngeal Neoplasms - pathology; Nasopharyngeal Neoplasms - radiotherapy; Radiotherapy Planning, Computer-Assisted - statistics & numerical data; Radiotherapy, Intensity-Modulated - statistics & numerical data; Software; Tomography, X-Ray Computed - statistics & numerical data
• ISSN: 0094-2405
• DOI: 10.1118/1.3611045

To develop a delineation tool that refines physician-drawn contours of the gross tumor volume (GTV) in nasopharynx cancer, using combined pixel value information from x-ray computed tomography (CT) and magnetic resonance imaging (MRI) during delineation. Operator-guided delineation assisted by a so-called "snake" algorithm was applied on weighted CT-MRI registered images. The physician delineates a rough tumor contour that is continuously adjusted by the snake algorithm using the underlying image characteristics. The algorithm was evaluated on five nasopharyngeal cancer patients. Different linear weightings CT and MRI were tested as input for the snake algorithm and compared according to contrast and tumor to noise ratio (TNR). The semi-automatic delineation was compared with manual contouring by seven experienced radiation oncologists. A good compromise for TNR and contrast was obtained by weighing CT twice as strong as MRI. The new algorithm did not notably reduce interobserver variability, it did however, reduce the average delineation time by 6 min per case. The authors developed a user-driven tool for delineation and correction based a snake algorithm and registered weighted CT image and MRI. The algorithm adds morphological information from CT during the delineation on MRI and accelerates the delineation task.