Observer Performance Studies

Performance on Film vs. Monitor - Bone Radiography

The goal of this study [2] was to assess the suitability of using an image console monitor for interpretation of computed radiography bone radiographs using ROC analysis and subjective evaluation techniques.

Methods

Images

  • 27 CR bone trauma cases
  • 9 no trauma signs (normal)
  • 18 subtle to moderate signs of swelling
  • 12 also with single subtle fracture
  • 4 also with dislocation or subluxation
  • 10 also with ligamentous injury
  • 8 also with joint effusions

Display

CR films on a viewbox Comparable position screen-film images on a viewbox (acquired at same time as CR films) 1152 x 1568 10-bit resolution display monitor

Procedure

3 radiologists & 3 orthopedic surgeons Read each image 3 times (once on CR film, once on screen-film, once on the monitor) Report diagnostic decision for each of the possible trauma signs & give confidence using a 6-level scale Subjectively rated the quality of film vs monitor

Eye-Position Recording

Eye-position was also recorded in this study. To find these results link to "Film vs. Monitor Search" in the Eye-Position Recording section of this web site.

Results

Performance Analysis

Figure 1. Percent true-positives for monitor (CRT) vs. screen-film (SF) vs. CR film (CR) for radiologists and orthopedic surgeons

Figure 1. Percent true-positives for monitor (CRT) vs. screen-film (SF) vs. CR film (CR) for radiologists and orthopedic surgeons

Figure 2. Percent false-positives for monitor (CRT) vs. screen-film (SF) vs. CR film (CR) for radiologists and orthopedic surgeons

Figure 2. Percent false-positives for monitor (CRT) vs. screen-film (SF) vs. CR film (CR) for radiologists and orthopedic surgeons

Image Quality

Figure 3. Image quality rating for CRT, SF & CR. 1 = poor, 2 = fair, 3 = good, 4 = excellent

Figure 3. Image quality rating for CRT, SF & CR. 1 = poor, 2 = fair, 3 = good, 4 = excellent

Image Processing Use

Figure 4. Percent of times each image processing function was used.

Figure 4. Percent of times each image processing function was used.

Conclusions

  • Viewing computed radiography chest images on a monitor does not affect diagnostic accuracy
  • There were no significant differences between CR, SR and monitor performance
  • Radiologists performed slightly better overall than the orthopedic surgeons
  • Image quality (contrast, sharpness & overall quality) were equivalent for the 3 viewing conditions, although CR film was rated
    lowest on all 3 measures
  • The most used image processing functions were the normal (film-like) setting and the bone enhance setting

Return to Observer Performance Studies page