The image quality and localization accuracy for C-arm tomosynthesis and cone-beam computed tomography(CBCT) guidance of head and neck surgery were investigated. A continuum in image acquisition was explored, ranging from a single exposure (radiograph) to multiple projections acquired over a limited arc (tomosynthesis) to a full semicircular trajectory (CBCT). Experiments were performed using a prototype mobile C-arm modified to perform 3D image acquisition (a modified Siemens PowerMobil). The tradeoffs in image quality associated with the extent of the source-detector arc , the number of projection views, and the total imaging dose were evaluated in phantom and cadaver studies. Surgical localization performance was evaluated using three cadaver heads imaged as a function of . Six localization tasks were considered, ranging from high-contrast feature identification (e.g., tip of a K-wire pointer) to more challenging soft-tissue delineation (e.g., junction of the hard and soft palate). Five head and neck surgeons and one radiologist participated as observers. For each localization task, the 3D coordinates of landmarks pinpointed by each observer were analyzed as a function of . For all tomosynthesis angles, image quality was highest in the coronal plane, whereas sagittal and axial planes exhibited a substantial decrease in spatial resolution associated with out-of-plane blur and distortion. Tasks involving complex, lower-contrast features demonstrated steeper degradation with smaller tomosynthetic arc. Localization accuracy in the coronal plane was correspondingly high, maintained to down to , whereas sagittal and axial localization degraded rapidly below . Similarly, localization precision was better than within the coronal plane, compared to out-of-plane for tomosynthesis angles below . An overall 3D localization accuracy of was achieved with for most tasks. The high in-plane spatial resolution, short scanning time, and low radiation dose characteristic of tomosynthesis may enable the surgeon to collect near real-time images throughout the procedure with minimal interference to surgical workflow. Therefore, tomosynthesis could provide a useful addition to the image-guided surgery arsenal, providing on-demand, high quality image updates, complemented by CBCT at critical milestones in the surgical procedure.
The C-arm prototype was developed in collaboration with scientists at Siemens Medical Solutions, Special Products Division (Erlangen, Germany), including Dr. K. Hermann, Dr. D. Ritter, Dr. M. Mitschke, and Dr. R. Graumann. The authors extend their gratitude to Dr. M. Wiley (University of Toronto) as well as L. Satterthwaite, D. Rego, and M. Romanova (Surgical Skills Centre, Mt. Sinai Hospital, Toronto, Ontario) for assistance with the cadavers. Special thanks to the physicians who participated in the observer study, including Dr. D. Goldstein, Dr. A. Moshaver, Dr. J. White, and Dr. E. Yu. Ms. S. Chhabra (University of Toronto) assisted with the phantom measurements. Ongoing collaboration with surgeons and scientists in the Guided Therapeutics (GTx) Program at the University Health Network are gratefully acknowledged, including Dr. M. Fehlings, Dr. M. Jewett, Dr. S. Keshavjee, Dr. W. Kucharczyk, Dr. D. McCready, Dr. C. Menard, Dr. J. Trachtenberg, Dr. R. Weersink, and Dr. B. Wilson. The research was supported in part by the Princess Margaret Hospital Foundation, the Toronto General Hospital Otolaryngology Associates, and the National Institutes of Health (Grant Nos. R01-CA112163-02 and R01-CA127444-01).
II. MATERIALS AND METHODS
II.A. Mobile C-arm for volumetric image guidance
II.B. Tomosynthesisimage quality
II.B.1. Depth resolution
II.B.2. Imagenoise, artifact, and out-of-plane “clutter”
II.C. Accuracy of surgical localization
II.C.1. Experimental setup
II.C.2. Surgical localization tasks
II.C.3. Observer study
III.A. Depth resolution
III.B. Tomosynthesisimage quality: Noise, artifact, and out-of-plane clutter
III.C. Surgical localization
III.D. Image acquisition time, reconstruction time, and dose
Data & Media loading...
Article metrics loading...
Full text loading...