Portal de Periódicos da CAPES

TB or Not TB: Interreader and Intrareader Variability in Screening Diagnosis on an iPad versus a Traditional Display

2013; Elsevier BV; Volume: 10; Issue: 1 Linguagem: Inglês

10.1016/j.jacr.2012.07.019

1558-349X

Samir Abboud, Frederick Weiss, Eliot L. Siegel, Jean Jeudy,

Reliability and Agreement in Measurement

Mobile displays may have the ability to increase the flexibility of consulting radiologists if they can be shown not to negatively influence a reader's performance in clinical tasks. The authors report a comparison of a mobile display with a larger liquid crystal display for the task of making a binary decision for the diagnosis of tuberculosis on chest radiography. Deidentified DICOM images of 240 chest x-rays were transferred from a PACS to an iPad 2 running OsiriX HD software. The images were reviewed independently by 5 radiologists of varying experience and were graded as positive or negative for tuberculosis on both the liquid crystal display monitor and the iPad. The reading sessions were a minimum of 2 weeks apart to minimize recall bias. Agreement was measured in terms of κ statistics. Overall, multirater generalized κ was 0.9694. These results suggest that there is no detectable effect of monitor type (liquid crystal display or iPad 2) on the reader's decision for the task of tuberculosis diagnosis. Mobile displays may have the ability to increase the flexibility of consulting radiologists if they can be shown not to negatively influence a reader's performance in clinical tasks. The authors report a comparison of a mobile display with a larger liquid crystal display for the task of making a binary decision for the diagnosis of tuberculosis on chest radiography. Deidentified DICOM images of 240 chest x-rays were transferred from a PACS to an iPad 2 running OsiriX HD software. The images were reviewed independently by 5 radiologists of varying experience and were graded as positive or negative for tuberculosis on both the liquid crystal display monitor and the iPad. The reading sessions were a minimum of 2 weeks apart to minimize recall bias. Agreement was measured in terms of κ statistics. Overall, multirater generalized κ was 0.9694. These results suggest that there is no detectable effect of monitor type (liquid crystal display or iPad 2) on the reader's decision for the task of tuberculosis diagnosis.