The effect of computer-aided detection on radiologist performance in the detection of lung cancers previously missed on a chest radiograph.

PURPOSE: The purpose of the study was to determine whether computer-aided detection (CAD) can improve a radiologist's ability to detect lung cancers previously missed on a chest radiograph (CXR).

MATERIALS AND METHODS: Eighty-one cases of lung cancer previously missed on CXR were collected, along with the CXRs of 215 age-matched and emphysema-matched controls without lung cancer. Tumor subtlety was scored from 1 (very subtle) to 5 (very obvious) by expert thoracic radiologists. All 297 CXRs were processed using a CAD system (OnGuard, Version 5.1 Riverain Medical, Miamisburg, OH) to create a set of 2 images for each patient, 1 with and 1 without CAD. Eleven general radiologists took part in a reader study. Each radiologist viewed the CXR without CAD and then the one with CAD for each patient sequentially. Areas of concern, if present, were marked. The degree of confidence in diagnosis was scored on a scale of 0 (no cancer) to 100 (definite) in succession for CXRs without CAD and then for those with CAD. Localization receiver operating characteristic analysis was used for evaluation of the observers' performance.

RESULTS: Of the 81 cancer cases, OnGuard correctly detected 40 tumors with a sensitivity of 49.4%. In the reader study, there was a significant increase in the area under the localization receiver operating characteristic curve with the aid of OnGuard, which increased from 0.38 to 0.43. Aggregate reader sensitivity improved significantly from 0.44 to 0.5 with the use of OnGuard.

CONCLUSION: The use of OnGuard improves reader accuracy and sensitivity for the detection of lung cancers previously missed on CXR.