Mingzhu Liang MD, Presenter: Nothing to Disclose

Wei Tang MD, Abstract Co-Author: Nothing to Disclose

Dongming Xu MD, PhD, Abstract Co-Author: Nothing to Disclose

Rowena Yip MPH, Abstract Co-Author: Nothing to Disclose

Artit C. Jirapatnakul PhD, Abstract Co-Author: Nothing to Disclose

Yu Htwe MD, Abstract Co-Author: Nothing to Disclose

David F. Yankelevitz MD, Abstract Co-Author: Research Grant, AstraZeneca PLC Royalties, General Electric Company

Claudia I. Henschke MD, PhD, Abstract Co-Author: Nothing to Disclose

To evaluate the usefulness of CAD in detecting lung cancer in low-dose CT scans that was missed by the radiologist in earlier rounds.

Lung cancers manifesting as a solid nodule and diagnosed in annual rounds of screening were reviewed to determine whether the cancer could be identified in the prior round of screening. All CT images were obtained at 1.25 mm slice thickness or less. Three software packages (Lung VCAR (GE healthcare), Image Checker CT (LN-1000, R2 Technology) and Syngovia(Siemens Medical Solutions)) were used to evaluate the scan when the cancer was first identified by the radiologists and the frequency with which the CAD identified the cancer was calculated.  The CAD was also used to determine whether it could identify the cancer on the earlier CT scans, when it was missed by the radiologist. The false positive rate of any nodule detection by the CAD was calculated using the consensus of two radiologists.

50 cases of lung cancer were identified (median age of 63 years), where in retrospect the cancer could be seen but was not reported. The average diameter was 11.4 mm (SD of 10.1 mm) when the cancer was identified by the radiologist for workup. The detection rates for GE, R2, and Siemens system were 74%, 82%, and 82%, respectively.  On the earlier CT scans (when missed by the radiologist), the average diameter was 4.8 mm (SD of 1.6 mm). The detection rates for each CAD were 56%, 70% and 68%, respectively. The false positive rate for any nodule was 7.4, 1.7 and 0.6 per study.

The CAD detected 56% to 70% of cancers on the earlier CT scans, all of which had been missed by the radiologists.  However, the CAD missed 18% to 26% of cancers when it was identified by the radiologist. This suggests that the current CADs may be useful as a second reader in CT screening programs as the lung cancers may be identified one year earlier although the false positive rate was highly variable.   

CAD has the potential to detect the majority of cancers missed by the radiologists in earlier round of screening when the cancer is smaller and more curable. 

Liang, M, Tang, W, Xu, D, Yip, R, Jirapatnakul, A, Htwe, Y, Yankelevitz, D, Henschke, C, Detection of "Missed" Lung Cancers using Computer-aided Detection Systems (CAD) in CT Screening for Lung Cancer.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14045614.html