gms | German Medical Science

Indications for surgical interventions at the skull base are complications of inflammatory diseases in otorhinolaryngology, tumours and injuries. The individual and complex anatomy of this region and the rare indications make an accurate surgical planning before intervention necessary.

CT- and MRI-datasets are used as a standard in diagnostic of inflammations, injuries and especially malignant tumours, located at the skull base. 2-planare representation is still standard. This method of representation requires a high cognitive achievement from surgeons to translate the information into the patient. A loss of information for the patients disadvantage may occur.

Due to the growing resolution of image scanners radiological data increase rapidly and in future slice-by-slice inspection will be no longer feasible. At present time there are complex representations only on basis of simple litographic models. Computer-assistance that becomes more and more important in clinical work is hereby not possible. For example intraoperative navigation is said to be standard for difficult surgical interventions at the skull base in almost all hospitals. Some existing workstations have got the possibility to segment items in a manual way. Processing takes a long time and can only be done directly on the workstations. The resulting visualization is rather primitive and therefore not used very often. A transfer from postprocessed data at a personal computer to the navigation system is not possible.

Until now many studies have been done to generate 3-D-models from CT-data by developing automatic or semiautomatic algorithms of segmentation. The required time was reduced extremely in the last few years. Difficult anatomic structures like bony ethmoid sinuses can now be segmented. While segmentation methods are refined visualisation felt behind.

Aim of our study was to develop a software to enhance preoperative planning based on segmentation. Without preselection from the radiologists all acquired data should be used and analysis time should be reduced as much as possible. Based on the results an individual surgical planning has to be practicable and possible complications may be avoided. The emphasis is set on developing visualization options to get a greater benefit compared to 2-planar images and to evaluate the different results referring to their clinical use and acceptance.

Basic elements are CT-datasets (dicom) acquired for regular diagnostic. No special or fixed protocol was needed.

Anatomical and pathological structures of the skull base are segmented by using classic algorithms of image analysis. Altogether eleven CT-datasets were analyzed. For segmentation we used modified watershed-transform, region growing and live-wire. Bones vessels, nerves, tumours and metastases are separated.

All relevant structures could be segmented and visualized with a time requirement by 30 minutes to 2 hours depending mainly on the number structures that have to be processed and the extent of the primary tumour and his possible penetrations.

The integration of these segmented objects into a surface-diagram with visibility allows topographic attachment.

We designed a smart medical 3-D-visualization by using standardised object colours and transparency values. After several clinical examinations we found out, that surgical acceptance was high, when pale and well-known-colours from anatomical books are used. Fast recognition of anatomic structures was supported by this.

To enhance preoperative planning different viewing directions are helpful. We developed an operation planning software (InterventionPlannerENT), that is based on the ILab-Library of MeVis, Bremen with which turning and rotation of 3-D-images is practicable in all required directions. Additional to viewing directions dynamic interaction facilities, like clipping-planes and clip boxes, are integrated in this planning software. These tools are necessary especially for investigation of the rhinobasis.

Automated measurement tools compute distances and volumes. The extents of pathologic structures are presented automatically. Even more important is the possibility to show minimal distances between pathologic items and risk structures. Due to this assistance potential complications are realized before surgical intervention and may be prevented. Minimal distances are indicated by using either data in millimetre or colour-coding. A surrounding circle in yellow or red indicates a distance of less then four or less then two millimetres.

An occurring problem by generating three-dimensional objects is superposition. At the skull base most of the tumours are hidden behind bony structures. To improve visualization transparencies are added to make hidden objects able to be seen. Global and local transparencies can be differentiated. Local transparency is suitable to make familiar structures visible. In comparison to local transparencies global transparencies can be used to make different levels apparent. Surrounding bones are often visualised in this way. Different intensities of transparency can be generated depending on surgical demands.

The disadvantage of highly transparent displayed objects is poor separation. Silhouette generation algorithms enhance visualization of transparently rendered objects. The use of silhouettes improves demarcation and was pointed out as very useful for 3-dimensional orientation. Comprehension of the patients individual anatomy and pathology was accelerated.

Nevertheless a 3-D-object, in which all segmented anatomic structures are integrated is too difficult to understand. Careful choice of displayed structures is indispensable to get a clearly arranged 3-D-object that makes the patients individuality understood at once. Otherwise the viewer gets confused because of the high number of visualized structures.

A variety of emphasis techniques like shadow generation for selected objects are shown to be suitable for medical visualization. Shadows let surgeons get a 3-dimensional impression, which is necessary because the images that are described above as 3-dimensional are only pseudo-3-dimensional illustrations. By looking merely at 3-D-monitors real 3-D-objects are indicated. We have to admit that these monitors are rare, but fortunately the human brain is able to interpret images as 3-dimensional when visualization options like shadow, fading, reduce of object sizes in the background and superimpose of items are added.

Presentation was done by pictures, movies or by using the created planning software. For regarding movies we designed a viewer (MedicalMoviePlayer) that allows a change of speed and zoom of the rotating objects. The InterventionPlannerENT permits a selection of displayed structures and variable movements.

Visualization showed an increasing resonance if the visualization options that are mentioned above are employed. Most of the surgeons profited from the available pictures and movies in comparison to familiar two-planar representations but the need of interaction was expressed by most of them.

The clinical need to adapt 3-D-visualization options is shown by this work.

The results can be used as a basis for diagnostic, individual operative planning and simulation. We have to emphasize that 3-dimensional reconstructed data are very helpful and may lead to a great benefit but do not substitute the original 2-planar datasets. To check details a return to the original data have to be always practicable.

The new aspects regarding the representation possibilities have to be integrated into a suitable software with interaction facilities for the surgeon to extract maximum profit from 3-dimensional images. Pictures and movies are not sufficient at all for optimal operation-planning.

Because of the described digital results computer-assistance is available. Integration of the segmented and visualized objects in a navigation system helps to avoid complications during operation. In addition the outcome is basis for employment of navigated control^©. By using a dynamic co-registration of preoperative segmented data and intraoperative MRI- or ultrasound data it is possible to correct tissue shift and to reach high accuracy in relocalisation of relevant structures in connection with intraoperative navigation.

Besides the indisputable clinical profit that we pointed out before, there is a potential conflict between surgeons and radiologists. But this conflict may be solvable because of their dissimilar inquiries. Radiologists set their goals on diagnostic and surgeons on operation planning. Visualization may remain in the hand of the radiologists. However the surgeons have to take part in developing visualization methods for surgical planning.

Figure 1 [Fig. 1], Figure 2 [Fig. 2], Figure 3 [Fig. 3]