Chair for Computer Aided Medical Procedures & Augmented Reality
Lehrstuhl für Informatikanwendungen in der Medizin & Augmented Reality

B. Röper, W. Wein, P. Kneschaurek, M. Molls, M. Dobritz, N. Navab
Image-based fusion of diagnostic ultrasound and planning CT for radiotherapy of head and neck cancer – a way towards improvement of regional target volume delineation
12. Jahreskongress der Deutschen Gesellschaft für Radioonkologie (DEGRO), May 2006, Dresden, Germany (bib)

To develop a workflow for integrating the diagnostic information of high-resolution 2D-ultrasonography (US) on neck lymph nodes into 3D-treatment planning on native CT for head and neck (H&N) cancer patients. The challenge is to develop an automatic image-based registration in spite of US and CT displaying different physical properties of the scanned tissue. Methods: A native planning CT (64-slice-CT, Siemens) of 0.6mm slice thickness is performed in supine position with the patient´s head stabilized posteriorly with a head pad and anteriorly with a mask. For US the mask is removed and the patient requested to keep his head steady. Several diagnostic sweeps with a high resolution US device (11MHz, 4cm linear probe, GE) covering all neck lymph node levels bilaterally are performed. Their video signals are digitally stored on a computer along with simultaneous information on probe positioning by optical tracking. Global surface registration: From the combination of 2D-US-video signal and tracking information 3D-data sets are created for all sweeps and a preregistration is performed by aligning the top of all US-pictures with the skin surface automatically extracted from CT. For automatic image based registration, an algorithm is developed based on several components contributing to a similarity metric, whose core are a modified Mutual Information on the image intensities, and alignment of edges extracted from CT. The algorithm is applied to central parts of the US frames only, omitting the upper 3.6mm of tissue depths where contact to the US-probe causes tissue compression, as well as automatically detected occlusion zones in deep tissue depths from where no further US-signal derives. For manual adjustment of individual sweeps the physician can choose any slice with identifiable anatomic landmarks in either modality and navigate it into the desired position with the stack of US-slices staying in the relative position as defined by tracking. For evaluation of the reliability of registration, meaningful targets were defined in each sweep. Results: This workflow was applied to five H&N cancer patients with cN2b-cN3 lymph node metastases. Patients´ acceptance was very high, their additional burden neglectable. Out of 10-15 sweeps per patient, 1-2 were chosen for evaluation, summing up to 9 sweeps representing all levels, both sides of the neck and displaying lymph nodes of all sizes. Two sweeps were already aligned correctly after global surface registration, four others after additional automatic image based registration. In three, manual adjustment was necessary in order to get into the capture range of correct image-based registration. The target registration error was 2.8 – 4.6mm for normal and metastatic lymph nodes, and 6.1 mm for bulky disease. Conclusion: This fundamental work brings target volume delineation based on examiner-independent US information within reach for H&N cancer and maybe the offspring for several other applications.
This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each authors copyright. In most cases, these works may not be reposted without the explicit permission of the copyright holder.

Edit | Attach | Refresh | Diffs | More | Revision r1.13 - 30 Jan 2019 - 15:16 - LeslieCasas

Lehrstuhl für Computer Aided Medical Procedures & Augmented Reality    rss.gif