ü 
a N 1485 7 % 
\ | ^A 
= / 
2&3 7& 
»- = NV 3s 
beam-line 
chair 
  
  
  
  
Fig. 4 Vertical section along beam line. 
Of the eight cameras, only three, on adjacent 
stations, will be used at any one time. This 
arrangement provides acceptable geometry and target 
visibility, as well as redundancies for system 
reliability. More than three simultaneously active 
cameras would demand additional  frame-grabbing 
capabilities and complicate operator procedures 
unnecessarily. 
CONTROL FRAME 
The CCD cameras need to be regularly calibrated in 
the beam coordinate system using control targets of 
known position. For this purpose a total of twelve to 
fourteen control points, visible on each image, was 
considered suitable. This number of points in an 
appropriate configuration is sufficient to obtain the 
required precision and reliability. To guarantee a 
minimum of twelve visible points per camera station, 
a total of 40 suitably placed points had to be 
mounted on a removable, but stable frame of cube- 
shape. 
Permanent mounts are incorporated into the treatment 
room floor so that, for each calibration procedure, 
the frame can be placed in an identical position with 
its centre approximately at the lesion point. 
The targets were coordinated to sub-millimetre 
precision using theodolite observations and bundle 
adjustment algorithms (Brown, 1985). 
TARGET DETECTION AND CENTRE DETERMINATION 
To enable automatic target detection it was necessary 
to introduce point markers which are readily 
distinguishable from the background in the controlled 
environment of the treatment room. Retroreflective 
tape, the use of which is widely reported in the 
literature (Brown 1982, Fraser 1988), proved 
appropriate for this application. This material 
requires suitable lighting originating from the 
direction of each camera, best achieved by a 
ringshaped light source around the camera lens. 
The target detection algorithm developed for the 
system relies on binary images, necessitating a 
thresholding procedure to separate targets and 
background. A computer-aided routine allows the 
operator to choose a suitable threshold value by 
inspection of the output images, resulting in a 
binary image in which targets are represented by bit 
value one (white) and the background by value zero 
(black). This binary image is then easily searched 
for targets and the target boundaries are determined 
for centre determination on the original grey images. 
After detailed investigation of target centre 
determination algorithms (Rubinstein and Riither, 
1991) the weighted centre of gravity model in 
conjunction with background-reduced circular targets 
emerged as the most efficient method for automatic 
centre determination. For the system, circular 
targets of approximately eight millimetres in 
diameter, equivalent to about seven by six pixels on 
the image given the geometric parameters, proved 
ideal as control point markers. 
    
   
   
     
     
    
   
  
     
    
     
   
    
    
   
  
    
     
    
   
      
    
  
  
  
  
    
   
   
   
   
     
      
    
     
     
     
   
  
   
   
     
    
    
     
   
   
    
THE DIGITAL PHOTOGRAMMETRIC PROCEDURE 
The photogrammetric procedure can be divided into 
four stages: 
1 Camera calibration - the position as 
well as the interior and exterior 
orientation of each camera is found 
using the control points. 
2 System check - the camera calibrations 
and the chair position are checked. 
3 Patient positioning - the patient is 
moved into the beam-line. 
4 Patient monitoring - possible patient 
movements are monitored to ensure that 
correct alignment is maintained. 
  
C Pre-treatment Stage ) 
| 
Automated camera 
callibration and chair 
initialisation 
  
  
  
  
Lesion is located relative 
to targets in CT or MRI | 
  
  
  
  
  
  
  
  
  
  
scans 
Cameras and chair are 
checked relatively 
| 
C Patient Positioning Stage E 
  
Patient is seated in chair 
and approximately moved 
into treatment position 
Image capture and 
location, coordination 
and identification 
of reference targets 
  
  
  
  
  
Transformation of lesion 
and entry point scan 
coordinates into beam 
coordinates 
A | 
Calculation of necessary 
movements to position 
patient into beam 
Patient automatically 
—] moved into beam and 
position checked 
| 
C Patient Monitoring Stage E 
| 
Beam activated and 
automatic monitoring of 
target begins. Automatic 
or manual beam cut-off 
  
  
  
  
  
  
  
  
  
  
  
  
  
Fig.5 Flowchart of positioning stages. 
Figure 5 depicts a flow chart of these 
photogrammetric stages including the lesion location 
stage, which is carried out separately as described 
above. This diagram emphasises the patient 
positioning stage of the project.