L A TENTATIVA INACABADA Y ACABADA DE ESTAFA

Predictably, radiation dose significantly decreased with increasing SID (p<0.05) with also a decrease in visual image quality found with increasing SID (p<0.05). This was slightly surprising considering the studies conducted in this area from Tugwell et al. (2014), Heath et al. (2011), Woods and Messer (2009) where they found increasing SID to significantly reduce dose but with no significant difference in visual image quality. Nevertheless, these studies were carried out using conventional imaging on an x-ray tabletop and therefore may not be directly comparable to trolley imaging when considering the differences between these two scenarios as discussed in section 3.1.4 on page 20. These previous studies on SID did not have an increased OID as for trolley imaging; there was less

variables to consider. This thesis demonstrates a decrease in image quality with increasing SID which is not surprising when considering that image quality should theoretically decrease when SID is increased according to the inverse square law due to reduced beam intensity (Reid-Paul, 2011).

Another reasonable explanation for the decrease in visual image quality with increasing SID may possibly be related to the method used to assess image quality. The visual image quality assessment for this thesis used bespoke software and a newly developed validated psychometric scale. This scale may be more sensitive to changes in image quality than the scales used in previous literature such as that of the CEC quality criteria which are based on film/screen imaging (Heath et al., 2012; Woods & Messer, 2009; Brennan, McDonnell & O'Leary, 2004; Grondin et al., 2004). As previously stated, the CEC guidelines were developed in 1996 for film/screen imaging and therefore many of the benchmarks do not apply in the digital environment, plus other important aspects of image quality relating to digital imaging are missing (Knight, 2014). In comparison to the CEC image quality criteria, the psychometric scale used for this thesis has gone through robust systematic testing and has validation data to support it. This newly developed psychometric scale may therefore be more sensitive to changes in visual image quality.

The last item on the 2AFC task brought about interesting results as it required the observers to decide whether the experimental image in question was acceptable or

unacceptable for diagnostic purposes. This item was included in the image quality scale since an image with a lower score than the reference image does not necessarily signify that it is not acceptable for diagnostic purpose thus requiring a repeat. There were 48 experimental images acquired for this thesis and five observers evaluating image quality. Twenty images where deemed to be of unacceptable image quality by one or more observers, however, only two of these twenty images were deemed unacceptable by a unanimous decision (i.e. three or more observers). This demonstrates a large variation in opinion amongst the observers as to the diagnostic quality of the images. Even though this type of decision (deeming images acceptable or unacceptable for diagnostic purposes) is an important and everyday responsibility of radiographers, in clinical practice the clinical indication for the examination is known to the radiographer. As previously mentioned, this demonstrates a major downfall to the last item on the visual image quality criteria because observers were asked to decide whether image quality was acceptable or not without any indication as to why the examination was undertaken. This is important because the clinical indication determines the required quality of an image (Chan & Fung, 2014; Harding et al., 2014). The AP pelvis examination is predominantly performed on a trolley in trauma situations (Whitley et al., 2015; Carver & Carver, 2012) and therefore image quality needs to be high when searching for fractures (Chan & Fung, 2014; Uffman & Schaefer-Prokop, 2009).

On the other hand, if observer number two was eliminated from the experiment, only seven images would have been deemed diagnostic by more than one observer instead of twenty. This means that observer number two was much more critical of image quality in

comparison to the other observers. Allen and Triantaphillidou (2011) commented that the experience of the observer may cause variation in the assessment of image quality which is interesting since observer number two was more senior in comparison to the other four. Perhaps stricter observer inclusion criteria should be made for visual evaluation rather than it being a radiographer with more than 5 years experience as this could be a radiographer with 6 years experience or a radiographer with 30 years experience.

On reflection to the above limitation regarding the last item on the visual image quality scale, it would have benefited from some validation work to determine whether the item achieved its aims and discriminated between good, adequate and unacceptable images rather than it being binary. In addition, this item may have achieved its goal if a more

specific question such as “is this image of diagnostic quality to detect fractures for trauma AP pelvis?’ would have been asked. Perhaps observers would have been more critical of image quality since according to Mraity et al. (2014a), it is commonly accepted that image quality can be described with regards to its acceptability for achieving the main clinical question. Uffman and Schaefer-Prokop (2009) and Busch and Faulkner (2005) also stated that the interpretation of image quality should be considered in groups/class where fracture detection requires the highest possible image quality in comparison to an image post hip replacement which requires lower image quality.

For this thesis, three (platform position, mattress thickness and SID) of the four

independent variables caused inconsistency in magnification level. This was because the OID varied for the different imaging conditions which meant that magnification level increased for all experimental images compared with the reference image. This can be worrying from a clinical perspective since the AP radiograph, even after trauma, can be used for surgical planning where the measurements of the patient’s prosthesis are determined. Femoral offset is a common measurement taken from the AP pelvis, were surgeons measure the distance from the centre of rotation of the femoral head to a line bisecting the long axis of the femur (Lecerf et al., 2009). According to Merle et al. (2013), femoral offset is frequently underestimated on AP pelvis radiographs as a result of

imprecise magnification. Paul, Docquier, Cartiaux and Banse (2008) went on to state that magnification on radiographs is a well known predicament in preoperative planning of orthopaedic surgery where software are used to presume magnification of approx 110% due to buttocks. When magnification levels vary, it can impact upon the correct selection of prosthesis size; this is why the use of a calibration ball is regularly used in clinical practice (Conn, Clarke & Hallett, 2002). To overcome these issues, Clohisy et al. (2008) stated that when evaluating a radiograph, whether it is AP pelvis or other projections, diagnostic accuracy and disease classification is improved when there are standardised imaging protocols in place where the same acquisitions parameters are used for every projection. This would not only improve surgical planning but also improve the

interpretation of images as they would be more comparable to previous and future images of the same area in the same patient.

6.3.4 Comparison of image quality and radiation dose for various mAs