Rianne Wittenberg MD, Presenter: Research grant, Koninklijke Philips Electronics NV

Joost Peters, Abstract Co-Author: Employee, Koninklijke Philips Electronics NV

Jeroen J. Sonnemans, Abstract Co-Author: Employee, Koninklijke Philips Electronics NV

Mathias Prokop MD, Abstract Co-Author: Nothing to Disclose

Cornelia Maria Schaefer-Prokop MD, Abstract Co-Author: Nothing to Disclose

To assess the impact of a computer assisted detection (CAD) prototype on observer performance for detection of acute pulmonary embolism in pulmonary CT angiographies (CTPA) performed off-hours.

Six observers of varying experience evaluated 209 CTPA scans (158 negative, 51 positive) that had been consecutively obtained during weekends and nights in a university hospital. Observers were asked to rate their diagnostic confidence without CAD and with CAD as second reader using a 9 point scale ranging from 1 (definitely no PE), 5 (inconclusive scan) to 9 (definite PE). Reading times without and with CAD were recorded. Reader data were compared with an independent standard of reference that had been established by two independent readers and a third chest radiologist in case of discordant results. The same standard was used to classify the CAD lesions as true positive (TP) or false positive (FP). Statistical evaluation was performed on a per-patient basis and a per-reader basis.

Detection rates were high: the area under the ROC curve, averaged over all readers, increased from 0.95 without to 0.97 with CAD (P> 0.05). Sensitivity with CAD increased for 3 of the 6 readers (87% to 91%, 89% to 94% and 90% to 94%) and remained constant for three readers (91%, 92% and 89%, respectively). Out of a total of 6x209 evaluations CAD was beneficial in 24%, with a correction of diagnosis in 29 cases and an increase of confidence in the correct diagnosis in 270 cases. CAD was detrimental in 2%, with a change towards the wrong diagnosis in 11 cases and a decrease of confidence in the correct diagnosis in 18 cases. The number of inconclusive ratings remained constant (38 vs. 36 cases). Reading time was on average extended by 22% using CAD as second reader.

Reader performance was not significantly improved with CAD but reader confidence in the correct diagnosis increased.

For CTPA scans obtained in an off-hour setting CAD can increase reader confidence in the correct diagnosis.

Wittenberg, R, Peters, J, Sonnemans, J, Prokop, M, Schaefer-Prokop, C, Impact of a CAD Prototype on the Detection of Acute Pulmonary Embolism Used as Second Reader for Off-hours CTPA Studies.  Radiological Society of North America 2009 Scientific Assembly and Annual Meeting, November 29 - December 4, 2009 ,Chicago IL. http://archive.rsna.org/2009/8014765.html