Zhenyu Jin MD, Abstract Co-Author: Nothing to Disclose

Da-Qing Ma MD, Abstract Co-Author: Nothing to Disclose

Wei Song MD,PHD, Abstract Co-Author: Nothing to Disclose

Li Fan PhD, Presenter: Nothing to Disclose

Guo-Qing Wei PhD, Abstract Co-Author: Nothing to Disclose

Jian-Zhong Qian PhD, Abstract Co-Author: Nothing to Disclose

To assess the ability of a real-time interactive pulmonary nodule analysis system for improving radiological interpretation of chest Digital Radiograph (DR) images cross multiple centers.

A group of hospitals performing DR lung cancer screening participated in this study. A total of about 200 chest DR PA studies were collected from screening patients. The DR images were generated from different DR hardware, but all with corresponding CT images. A real-time interactive pulmonary nodule analysis system (IQQA-Chest, EDDA Technology, Inc.) was developed to aid radiologists in chest DR image softcopy reading. Two radiologists from each hospital independently read the studies collected therein and marked small actionable nodules in two passes: first without using the computer assistance; then using the assistant, concurrent reading tools of the computer system. Using the corresponding CT images as “Gold Standard”, nodule marking and characterization results of radiologists before and after using the computer assistance were compared. The automatic nodule suggestion function of the computer system was also assessed.

Preliminary results showed that before using computer assistance, individual radiologist’s detection rate was low, from about 37% to 60%. In addition, all radiologists detected only about 35% true nodules in common, which indicated substantial inter-observer variations. After using the real-time interactive computer assistance, individual radiologist’s detection rate increased to as high as over 80%, and the inter-observer variation decreased significantly. The automatic nodule suggestion function of the computer system achieved a detection rate of 71% with an average of 1.8 FP per study.

The multi-center study showed that real-time interactive pulmonary nodule analysis and concurrent reading system can effectively help to increase radiologists’ nodule detection rates and to decrease inter-observer variations.

Jin, Z, Ma, D, Song, W, Fan, L, Wei, G, Qian, J, Improving Radiological Interpretation of Chest Digital Radiograph Images Using a Real-time Interactive Pulmonary Nodule Analysis System: A Cross-center Study.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4418829.html