HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Graf, B. Articles by Fiedler, V. Search for Related Content PubMed PubMed Citation Articles by Graf, B. Articles by Fiedler, V. Social Bookmarking                      What's this?

IK Versus 2K Monitor

A Clinical Alternative Free-Response Receiver Operating Characteristic Study of Observer Performance Using Pulmonary Nodules

¹ Klinikum Krefeld, Institute of Diagnostic Radiology, Lutherplatz 40, 47805 Krefeld, Germany.
² Present address: Siemens AG, Medical Engineering, Völklinger Str. 2, 40219 Düsseldorf, Germany
³ German National Research Center for Information Technology (GMD), Institute for Applied Information Technology, Schloss Birlinghoven, 53754 Sankt Augustin, Germany.

View larger version (32K): [in a new window]   Fig. 1. —Bar chart shows prevalence of indications for CT examinations of patients with and without pulmonary nodules. Note that because the difference between sample size of both patient groups was approximately 75% statistical testing of similarity between two prevalence distributions for samples was not conducted. Light shading represents number of patients with nodules. Dark shading represents number of patients without nodules.

View larger version (14K): [in a new window]   Fig. 2. —Histogram shows diameter of pulmonary nodules revealed by CT. Note that in total 68 pulmonary nodules were evident on CT scans with diameter of 3-35 mm, mean diameter of 12 mm, and median value of 10 mm. SD = 7.5 mm.

View larger version (20K): [in a new window]   Fig. 3. —Graph shows alternative free-response receiver operating characteristic (ROC) curves of every monitor condition averaged for all six observers. Az = area under alternative free-response ROC curve. a and b represent vertical intercept and slope, respectively, of alternative free-response ROC curve when it is plotted on normal—deviate axes. Note that with regard to area under average alternative free-response ROC curve, 1K monitor with postprocessing (0.73 ± 0.081) and 2K monitor with (0.75 ± 0.101) and without (0.75 ± 0.109) postprocessing have equivalent results in detecting pulmonary nodules in digital chest radiographs in clinical interpretation environment. 1K monitor without postprocessing performed poorly (0.69 ± 0.086). 1K overview, a = 0.65, b = 0.81, Az = 0.6931. 2K overview, a = 0.86, b = 0.76, Az = 0.7524. 1K with postprocessing, a = 0.81, b = 0.84, Az = 0.7330. 2K with postprocessing, a = 0.85, b = 0.75, Az = 0.7504.

View larger version (26K): [in a new window]   Fig. 4. —Graph shows alternative free-response receiver operating characteristic curves for all six observers on 1K monitor overview without postprocessing. Note high interobserver variability but consistency of accuracy in performance of each observer under different viewing conditions. Observer 2 (minimum area under curve, 0.58) and observer 6 (maximum area under curve, 0.82) stood out among six observers. Observer 1, a = 0.38, b = 1.00, Az = 0.6050. Observer 2, a = 0.24, b = 0.52, Az = 0.5839. Observer 3, a = 0.76, b = 0.81, Az = 0.7211. Observer 4, a = 0.67, b = 1.06, Az = 0.6768. Observer 5, a = 0.70, b = 0.70, Az = 0.7159. Observer 6, a = 1.17, b = 0.80, Az = 0.8190.

View larger version (27K): [in a new window]   Fig. 5. —Graph shows alternative free-response receiver operating characteristic curves for all six observers on 2K monitor overview without postprocessing. We determined ranking for observers according to calculated area under curve values for every monitor condition for which 2K monitor overview is representative example. On average, observer 6 performed best followed by observer 4, observer 5, observer 3, observer 1, and observer 2. Observer 1, a = 0.57, b = 0.96, Az = 0.6597. Observer 2, a = 0.27, b = 0.28, Az = 0.6024. Observer 3, a = 0.56, b = 0.56, Az = 0.6884. Observer 4, a = 1.15, b = 1.03, Az = 0.7882. Observer 5, a = 0.76, b = 0.55, Az = 0.7457. Observer 6, a = 1.83, b = 1.18, Az = 0.8821.

View larger version (24K): [in a new window]   Fig. 6. —Graph shows alternative free-response receiver operating characteristic curves for all six observers on 1K monitor with voluntary postprocessing. Note that again observer 6 performed best with area under curve of 0.82 and observer 2 performed the worst with area under curve of 0.60. Observer 1, a = 0.71, b = 1.13, Az = 0.6816. Observer 2, a = 0.29, b = 0.60, Az = 0.5971. Observer 3, a = 1.07, b = 0.82, Az = 0.7963. Observer 4, a = 0.83, b = 0.93, Az = 0.7281. Observer 5, a = 0.69, b = 0.54, Az = 0.7278. Observer 6, a = 1.27, b = 0.96, Az = 0.8190.

View larger version (26K): [in a new window]   Fig. 7. —Graph shows alternative free-response ROC curves for all six observers on 2K monitor with voluntary postprocessing. Note that on 2K monitor with voluntary postprocessing, observer 6 achieved highest area under curve value (0.89) in this alternative free-response ROC analysis. Again observer 2 showed inferior performance compared with other observers (area under curve, 0.62). Observer 1, a = 0.97, b = 1.21, Az = 0.7311. Observer 2, a = 0.30, b = 0.30, Az = 0.6130. Observer 3, a = 0.30, b = 0.21, Az = 0.6150. Observer 4, a = 1.12, b = 1.20, Az = 0.7645. Observer 5, a = 0.56, b = 0.48, Az = 0.6923. Observer 6, a = 1.83, b = 1.12, Az = 0.8882.

View larger version (26K): [in a new window]   Fig. 8. —Bar chart shows application of postprocessing options averaged for all observers and patients. Note that in just over half of cases observers applied both options, and only in few cases (1K: 8%, 2K: 13%) did observers not apply postprocessing. For all 48 patients, application frequency of window setting and magnification was investigated for all observers. Student's t test did not indicate statistically significant differences in use of window setting (p = 0.14) and magnification (p = 0.24) between 1K and 2K monitors. Light shading = 1K monitor with postprocessing. Dark shading = 2K monitor with postprocessing.

CiteULike

Complore

Connotea

Del.icio.us

Digg

Reddit

Technorati