DESIGN CYCLE U: EARLY AND CONTINUOUS TESTING
6.2 A 1 Statistical error analysis
Statistical errors are due to sample selection. The author examined the image perception variance between operators when they use the 3D display or the microscope eyepieces. Figure 6.15 shows the mean values of the difference in binoculars and 3D free viewing (y-axis) plotted against the difference in autostereoscopic 3D display and 3D free viewing (x-axis). Each square plotted refers to a different observer. Ideally, all squares should lie very close to the origin.
De s ig n Cy c l e II: Ea r l ya n d Co n t in u o u s Te s t in g
O p e ra to rs ' 3D im age v isu a l p e rc e p tio n (16x)
> ' -1 --- 1 - 0.5 -
M um
1 -0.5 g ( -0.5 - !--- -4-^ W ' ' ) 0.5 X (mm) O p e ra to rs ' 3 0 im age v is u a l d e p th p e rc e p tio n (10x) 0.5 - -0.5 0.5 -0.5 - X (mm)Figure 6.15: Depth perception variation for all operators.
Figure 6.15 clearly illustrates the way all operators view depth information from both visualisation media. Note that position mean differences are very small with all values lying very close to zero, thus implying only small discrepancies between both the Sharp 3D autostereoscopic display and the binocular stereo microscope with respect to the actual position for each operator. The line along which the points are clustered demonstrates that there is a very small difference indeed between the 3D autostereoscopic display and the binoculars depth perception for each operator in spite of variations in operator bias. Ideally, all points should lie at the origin. The data pattern confirms the fact that different surgeons tend to interpret the image differently. This finding suggests future research into the quality of depth percept when using the autostereoscopic display.
De s ig n Cy c l e H : Ea r l ya n d Co n t in u o u s Te s t in g
6.2.4.2 Systematic error analysis
The active window o f our prototype Micro-optic twin-LCD monitor was placed at a distance o f 270 mm from the back LCD panel. Every operator was positioned at 270 mm from the window plane (see Figure 6.16). At this position, the specific prototype offers optimum three-dimensional visual information. The average interpupillary distance of the operators is assumed to be at the normal average o f 60 mm. The display’s depth (or longitudinal) resolution can be determined from the binocular viewing geometry using the divergences of the LCD light rays through the active window. For the display system discussed in this thesis, this was found to be 0.16 mm. This error is comparable to the 0.12 mm error in the operator-response graph o f depth perception.
5z = 0.16 mm
2 7 0 m m
A ctive W indow (V iew from top) 27 0 m m
Figure 6.16: Depth variations in the viewer-display space.
6.3 Discussion
6.3.1 Early testing
The early testing period of the second cycle in the proposed design process involved the development o f a prototype stereo camera system that can extract microscopic stereo video images from a conventional surgical microscope and display them in an autostereoscopic 3D display. That was done by designing and manufacturing a mechanical coupling component that mounts a video camera to the body o f commercially available video objective adapter with bayonet mounting. The optical arrangement o f the TV objective-coupler (mounted directly into each side of the Zeiss beam splitter, as shown in Figure 6.17) favours a parailel-axes stereo video camera system o f binocular imaging. Furthermore, this allows the user to use the binocular
De s ig n Cy c l e II: Ea r l ya n d Co n t in u o u s Te s t in g
eyepieces if there is a need to switch back to the conventional viewing technique. This architecture ensures that the prototype has very good flexibility, and can be easily integrated with any standard surgical microscope apparatus.
Figure 6.17: The stereo camera system can be installed quickly at any Zeiss ENT microscope. For other commercial ENT microscopes, a different beam splitter must be used.
The ability of the coupler to move the camera head in the x-y plane as well as in the z direction allows good alignment and focusing of stereoscopic images from different cameras. Camera translation could also be useful for controlling the amount of depth in stereoscopic image pairs - for example, through image cropping - a technique that is explained during the third cycle of the design process.
6.3.2 Continuous testing
The evaluation results showed that there’s a correlation between the viewing properties of the Sharp Micro-optic twin-LCD autostereoscopic display and the eyepieces of the surgical microscope. Great emphasis is given to the perception of depth, as this is very important to the acting ENT surgeon while s/he performs a surgical operation. Equally, the results establish the accuracy and precision in viewing 3D objects using the new ‘heads up’ display. A paired t-test sample confirmed that there was no significant difference between the expected and mean measured values for both lOx and 16x magnification factors (see Table 4.2). Depth recognition measurements of a microscopic model result in deviations not more than 1 mm, the average surgeon’s hand movement precision.
De s ig n Cy c l e II: Ea r l ya n d Co n t in u o u s Te s t in g
C onditional Param eters for a sam ple o f n=35
M agniG cation -xlO - (m m )
M agnification -x l6 - (m m )
x-direction actual error y-direction actual error z-direction actual error Statistical t for x-direction Statistical t for y-direction Statistical t for z-direction
-0.015 -0.082 -0.062 0.40 1.04 0.59 -0.061 0.050 0.022 0.95 0.68 0.27
Table 6.2: Variation of statistical results for magnification factors 10x& 16x.
The combined effect of good accuracy and precision was also noticed during the time study, where no significant differences were noticed between the execution of tasks using the 3D display and the conventional microscope. This implies good throughput.
Moreover, the study revealed a significant difference between each of the two methods and the 2D display alone. This effect was expected as depth perception was eliminated, but its demonstration verifies the necessity of stereoscopic viewing in the practice of microsurgery.
The attitude of the clinical subjects who tried the experiment was positive to the use of this new 3D viewing technology. The device was found to be easily adaptable, accommodating to the eyes while offering natural viewing conditions and substantial freedom of movement of the user’s head. As a result, users were able to maintain a general sense of their position within the surgical workspace through their peripheral vision, as opposed to the immersive viewing conditions of conventional microscopes. The latter was proved a great obstacle in the microsurgical operations studied during the first cycle of the design process, and was described analytically in §5.1.1.
Similar results are also noted by other research groups as in laparoscopic [104] and endoscopic [105] surgery. It is envisaged that autostereoscopic 3D displays offer opportunities for lengthy surgeries. The additional benefit of the ‘electronic eye’ is that the image can also be routed to other display monitors located outside the operating theatre. This would introduce educational benefits for training purposes, as it will be explained in Chapter 8.
De s ig n Cy c l e III: In t e r f a c e Op t im is a t io n