El taller como potenciador de la lectura y la escritura

A mixed-reality based guidance system usually consists of the following major components:

pre-operative images or models, intra-operative images, tool tracking system, image processing

programs, and visualization programs (see Figure 1.11).

1.8.1

Pre-operative Images

Pre-operative images and models are mostly used for diagnosis and surgery planning. They

can also be integrated into the intra-operative guidance system to provide high quality im-

ages as enhancement to the intra-operative images. Pre-operatively defined models can also

provide useful information that may be difficult to obtain or to produce during the operation, such as planned surgical path, clearly defined target position, annotations of surrounding tis-

sues, and customized fusion and visualization of multiple images. However, the limitation of

pre-operative images is that they may not present the true intra-operative patient state. Many

factors differentiate pre-operative images from their intra-operative counterparts, such as dif- ferent imaging modalities, different scanning positions or poses, deformed tissues, surgical tool presence, and changes in lesions or surrounding tissues. Image registration techniques are

commonly required for using pre-operative images during the surgery.

1.8.2

Intra-operative Images

Intra-operative images show the true state of a patient during the surgery. They can be the

same or different imaging modalities as the pre-operative images. The importance of having intra-operative images is that they can show not only the anatomy and motion of the patient, but

also the surgical tools and the interaction between the tools and the tissues. The latter cannot

be found in the pre-operative images. However, it is a general requirement that the procedure

of obtaining an intra-operative image should not significantly interfere with the OR workflow.

Waiting half an hour or maybe even ten minutes to get an image is usually not acceptable in the

OR. Furthermore, in beating heart interventions, it is also important to show the real time heart

motion. Inside a beating heart, some tissues, such as valve leaflets, can move rapidly, requiring

high temporal resolution of the imaging modality, which means the scanning process for one

1.8.3

Tool Tracking Systems

Tool tracking systems continuously track the position and orientation of the surgical tools and

send the updated information to the guidance system. Common choices of tracking systems are

optical tracking, visual tracking, magnetic tracking, and image based tracking. Optical tracking

systems usually use two cameras to locate the positions of infrared-emitting or retro-reflective

markers, usually spheres, and compute the positions and orientations of the tracked tool based

on the markers positions. The accuracy is generally excellent, in many cases better than one

millimeter. However, the line-of-sight is a common limitation for optical tracking, implying

that the markers should not be blocked from the camera by any objects at any time for the

tracker to be able to work. Visual tracking is similar to optical tracking, but it uses geometric

patterns attached to the tools for tracking. It is cheaper and can be more scalable than optical

tracking, but it also has the occlusion issue and the general accuracy is not determined yet.

Magnetic tracking systems employ a field generator to generate a magnetic field and sensor

coils placed in the tracking area induce different voltages according to the varying magnetic fields, which can be used to calculate the position and orientation of the sensors. A major ad-

vantage of magnetic tracking is that it does not have the “line-of-sight” constraint. Sensor coils

can be placed anywhere inside a human body and will operate as long as they are within the

sensitive volume of the field generator. However, metal tools may greatly affect the tracking ac- curacy, although medical grade stainless steel and titanium are claimed to be non-problematic.

Also, the general accuracy of magnetic tracking is worse than optical tracking. Image based

tracking doesn’t rely on any external devices other than the imaging devices themselves, and

the position and orientation of the tool is calculated based on the tools appearance in the im-

ages. It is a good option when the other three techniques are not applicable, but accuracy of

1.8.4

Image Processing Techniques

Image processing programs act as the linkage between pre-operative images/models, the intra- operative image, and visualization programs. They are used to bring all the useful information

together from available images, either pre-operative or intra-operative, and make them more

salient and easier to interpret for the surgeons. For example, locating and dynamically updating

the position of the surgical target, if possible, often receive the most attention in any image

guidance systems; marking the regions that may cause potential danger if the tool accidently

gets into is helpful for improving the overall safety of the procedure; and incorporating high

quality pre-operative image to the system can help the surgeon to better interpret the anatomy,

when intra-operative images have relatively low quality.

In terms of processing techniques, segmentation, registration, and image filtering including

regularization, enhancing, feature extraction, and thresholding are the most commonly used.

Segmentationusually forms the basis for generating anatomical and/or functional models, locating surgical targets, and quantitative measurements within the area of interest. For med-

ical purposes, the segmentation results are usually expected to be highly accurate (≤1mm).

However, this level of accuracy is often very hard to achieve. Many segmentation approaches

that give satisfactory results are then very computationally intensive, which limits the oppor-

tunity of using them intra-operatively. Thus, most of the segmentation work is still performed

pre-operatively, which also allows review and correction by the clinician.

Registration is the key component for linking multiple images together, especially for

aligning pre-operative images to intra-operative images. Many factors can lead to variations

between pre-operative and intra-operative images, such as tissue deformation, changes in le-

sions, inter-modality intensity distribution differences, and different coordinate system used by the imaging systems. In most cases, these factors are combined in various ways, so the choice

of using either rigid or non-rigid registration and the selection of a similarity metric depend

on applications. The factors to be considered are the differences in image presentation, com- putational efficiency of the approach versus requirement of the efficiency of the application,

general accuracy of the approach versus tolerance of registration error of the application, ex-

tra time and efforts required for implementing a certain approach versus estimated benefits of using this approach.

Image filtersare commonly used in the pre-processing step before segmentation or regis-

tration. They can smooth or sharpen images, enhance contrast, and detect geometrical features,

etc. In some cases, a well chosen filter can be the difference between success and failure. His- togram matching, for example, may greatly improve the performance of mono-modality regis-

tration. Such steps are often referred to as engineering tricks as they are not always reported in

publications.

1.8.5

Visualization

Visualization platforms obtain data from the image processing programs and the tool tracking

systems and display them on the screen. There are many different ways of visualizing an image or a virtual model, such as maximum intensity projection, volume rendering of the whole

image, showing orthogonal slices, rendering a model as polygon or surfaces with shading, and

overlaying/fusing images together. It is not always necessary to show everything that the image processing programs can provide on the screen. Actually, in many situation, showing too much

is as bad as showing too little. The optimal choice of what to visualize at each certain stage

during the procedure needs careful studies. A general criterion is that the visualization should

ease the mental workload of the surgeons but not add to workload.

1.9

Image Registration Techniques for Guidance of MI Off-