Deborah Joy Walter PhD, Presenter: Nothing to Disclose

Colin Craig McCulloch PhD, Abstract Co-Author: Nothing to Disclose

Claudia Ingrid Henschke MD, Abstract Co-Author: Nothing to Disclose

Ella Annabelle Kazerooni MD, Abstract Co-Author: Nothing to Disclose

David F. DON'T USE THIS Yankelevitz MD, Abstract Co-Author: Nothing to Disclose

Smita Patel MBBS, Abstract Co-Author: Nothing to Disclose

To measure radiologist detection performance for nodules in low dose CT chest exams and to measure the differences in how radiologists characterize detected nodules.

In a multi-reader, multi-institution study the variability of radiological readings of low-dose chest CT exams was measured. Readers were asked to note all nodules and rate them according to their size, calcification, density type (solid, part-solid, and non-solid), and severity based on the radiologist’s assessment of the likelihood of malignancy on a 5 point scale (definitely not malignant, probably not malignant, intermediate, probably malignant, or definitely malignant). Three readers at each institution reviewed 45 cases; scans were taken by a GE Lightspeed. The data were enriched such that 75% of the cases had at least 1 non-calcified nodule (NCN) >5mm. The nodule length and width were measured by the radiologist using a digital caliper tool and recorded electronically using GE’s Advance Lung Analysis program on the Advantage Workstation.

The number of NCNs >5mm in diameter, rated as severity of 2 (probably not malignant) or higher recorded by each radiologist ranged from 25-36. The institutional average of NCNs recorded was 31.3 and 26.7%. An individual reader’s sensitivity was calculated as the percent a radiologist noted an NCN >5mm, with severity 2 or higher, that was also noted by a majority of readers (at least 3 out of 6). A total of 24 NCNs were in the majority set. Reader sensitivity ranged from 67-100%; the institutional averages were 78 and 90%, respectively. Focusing on nodule characteristics, we analyzed the percent of nodules that each radiologist determined were solid vs. sub-solid. The number of NCNs classified as solid ranged from 46-81%; the institutional averages were 69.7 and 64.0%.

Differences exist between institutions in the number and type of NCNs identified. More variability was seen between readers than between institutions. A significant number of disagreements existed among radiologists when asked to classify nodule density type. Automated tools for nodule detection and classification may be able to reduce the variability among readers.

E.A.K.,C.I.H.,D.D.Y.,S.P.: Authors participated in a study funded by General ElectricD.J.W.,C.C.M.: Authors work for General Electric a manufactor of medical equipment

Walter, D, McCulloch, C, Henschke, C, Kazerooni, E, Yankelevitz, D, Patel, S, Inter-Institution Variability in the Detection of Noncalcified Nodules in Low-dose CT Lung Screening Exams.  Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL. http://archive.rsna.org/2004/4403668.html