Purpose: Evaluation of the effectiveness of computer-aided diagnosis (CAD) in the identification of pulmonary nodules.
Materials and methods: Two observers (A1, A2) with different levels of experience independently evaluated 20 chest MSCT studies with and without the aid of a CAD system (LungCheck, R2 Technology, Inc). The study parameters were as follows: 140 kVs, 40 mAs, collimation 4 x 1 mm, slice thickness 1.25 mm, reconstruction interval 1.0 mm. The observers analysed the images with and without CAD and evaluated: 1) nodule size (longer axis); 2) number and location of nodules; 3) reading time for each observer. The gold standard was represented by the evaluation of both readers in consensus with the aid of the CAD system.
Results: Without CAD support the two readers identified 77 (A1) and 79 (A2) nodules and with CAD 81 (A1) and 82 (A2) nodules. Working in consensus the two observers identified 81 nodules without the aid of the CAD and 84 nodules with the aid of CAD. Total number of nodules identified by CAD was 104, 25 of which were false positive and 5 false negative. The average reading time with the aid of the CAD decreased by as much as 40% for both the observers.
Conclusions: The preliminary results of our study suggest that the CAD technique is an accurate automatic support tool in the identification of pulmonary nodules. It reduces reading time and automatically supplies the size, volume, density and number of nodules, thus being useful both in screening programmes and in the follow-up of cancer patients, in whom comparison of the images is particularly difficult.