The reproducibility o f frame fixation was assessed by measuring the displacement o f points o f interest in the head on AP and lateral radiographs with fixed x-ray sources and plates. Radiographs were used rather than CT or other imaging modalities because the spatial resolution o f radiographs is extremely high (i.e., the grain size on the film). By contrast the spatial resolution o f CT is relatively poor being depaidant on the voxel size (eg .6 x .6 x 1.5mm). Although the SGV localiser was used to provide external referrace points from which to measure frame repositioning error, the SGV software was not used because this averages out the co-ordinate positions to the nearest millimetre which was felt to be o f insufficient accuracy.
One o f the main difficulties with this radiographic technique if one does not have access to a dedicated teleradiographic set-up, is the requirement to exactly reproduce the x-ray source to frame to film relationship each time a patient is re-imaged. In particular, errors in the realignment o f the central axis o f the x-ray beam relative to the stereotactic frame between images can result in the apparent displacement o f fixed reference points. Significant errors may therefore be introduced into comparative measurements made between the films.
Alignment o f the central axis of the x-ray source for the phantom studies was o f less consequence than for the patient studies, because after the initial set-up the frame remained fixed relative to the AP and lateral x-ray sources for all repositionings o f the phantom head.
For the Group 1 patients at Maida Vale Hospital the ceiling mounted x-ray sources were at fixed distances from the BRW floor stand. The alignment o f the central ray for each x-ray source was achieved when the light beam from each source struck the central reference markers on the opposing AP or lateral plates of the angiographic localiser. The alignment was checked on AP and lateral films and compared with previous radiographs prior to the patient being repositioned in the frame.
F or the Group 2 patients, malalignment o f the central beam may have played a more significant role in introducing error. Simulator couch movements were used to align the central beam o f the treatment simulator x-ray source with the centre o f the BRW co-ordinate system o f the frame. This was achieved by the alignment o f a laser light on the x-ray source to crosses engraved on three faces o f a Perspex localiser box indicating the central cardinal axes o f the co-ordinate system. The errors introduced by this method of alignment are difficult to assess but may be as much as 0.3mm.
Points o f interest were marked in the phantom skull by positioning 1mm metallic ball bearings which could be identified on the AP and lateral films. These were in the frontal, parietal, occipital and thalamic regions. In the patient studies, reference points in the mid-sagittal plane were defined by pinholes in the
ACCURACY OF RELOCATION - Disscussion 9 0 lateral reference film at the Bregma (frontal point), the Lambda (occipital point) and at the midpoint between the frontal and occipital poles (thalamic point). The pinhole reference points were then transposed to the mid-sagittal plane on the AP film taking into account the relevant magnification factors. The advantage of using pinholes to identify points in the skull rather than visible bony landmarks alone, is that it is extremely difficult to reliably identify the same bony points on AP and lateral radiographs. In addition, if one is attempting to measure sub-millimetre displacement o f bony points on a series o f films, then one could not rely on an individual to repeatedly identify the same point on a series o f radiographs with an accuracy better than 1 to 2 millimetres.
In contrast, by using the ’pinhole’ method described, points identified on a reference film can be transposed to subsequent films o f the same patient with an accuracy in the order o f 0.1 millimetres. When a radiograph and a reference mask film are superimposed then any minute degree o f mis-matching at any of the many thousands of potential reference points on the skull will be seen as an interference pattern of black and white lines on a grey background. When the films are perfectly matched the images will completely subtract. The pinhole reference points made in the reference mask film can now be made through the matched film with great precision.
When the superimposed films are realigned so that the SGV reference markers completely subtract, then the displacement o f each pinhole from its counterpart in the reference film is measured to the nearest 0 .1mm. It did not seem reasonable to measure displacements with greater precision than 0 . 1mm (the width o f a human hair), because this was assumed to be the approximate accuracy with which one could match superinqx)sed films.
One shortcoming of using the pinhole method of transposing reference points from one film to another is apparent when reference points at different depths along the mid-sagittal plane are transposed from one AP film to another. If these were actual points in the brain visible on an AP radiograph and lateral displacement o f the head occurred between the first and second AP film, then the differential displacement o f the frontal reference point would appear to be greater than that of the occipital point. This is because the magnification factor at the position of the frontal reference point is greater than at the occipital as it is furthest away from the film. Using the ’pinhole’ method described, the mid-sagittal points are transposed from the reference AP film to the test film by superimposing the skulls. The films are then realigned so that the SGV markers subtract. This obviously makes all the points on the mid-sagittal plane together irrespective o f their different magnification factors.
To determine the actual lateral displacement of each point, the assumption is made that in matching the skulls on superimposed AP films, the ’best fit’ is made at the mid-coronal plane. If there is lateral displacement o f the head between frame fittings, then the measured lateral displacement of each point along
ACCURACY OF RELOCATION - Disscussion 91 the mid-sagittal plane can be corrected to its actual displacement by dividing its value by the magnification factor at the mid-coronal plane.