Haili Chui PhD, Presenter: Nothing to Disclose

Feng Ma, Abstract Co-Author: Nothing to Disclose

Alex Schneider, Abstract Co-Author: Nothing to Disclose

Philip F. Judy PhD, Abstract Co-Author: Nothing to Disclose

Susan Alyson Wood PhD, Abstract Co-Author: Nothing to Disclose

To test and validate the performance of a CAD system that automatically detects abnormal growth patterns of nodular structures in temporal MDCT scans.

Abnormal growth of nodular structures provides important information for lung nodule diagnosis. A CAD system (R2 Technology, Sunnyvale, CA) is designed to automatically compare multiple temporal MDCT chest scans and detect such abnormal growth patterns for nodular structures. The system was evaluated on a set of 40 current-prior pairs of MDCT chest examinations (acquired 1-18 months apart) using 1.0-2.5 mm collimations. Double reading by radiologists on each of the 80 (40x2) exams identified a set of 97 nodules that required intervention or surveillance. The CAD system was deployed to detect nodule candidates showing abnormal growth (30 days < doubling time < 450 days) in this data set. Provided with the additional nodule growth information, a radiologist reviewed each pair of the detected nodule candidates (one in prior scan and one in current scan) to evaluate their clinical significance. The evaluation results were compared to the set of 97 nodules initially identified by radiologists without any growth information.

The CAD growth detection system reported a total of 57 nodule candidates with abnormal growth. 20 out of these 57 nodule candidates were found included in the radiologist-identified nodule set of 97 nodules. 37 of the 57 candidates were not included in that set of 97 nodules found by radiologists. Further reviewing by radiologist of these 37 candidates and their growth patterns concluded that 15 out of 37 were highly likely nodules and worth further surveillance. One common characteristic of these 15 newly found nodules was that they were relatively smaller in size (around or = 4mm).

The addition of nodule growth information is valuable to the nodule reviewing/diagnosis process. The growth detection CAD system can accurately track nodular structures across temporal scans and detect potential abnormal growth patterns. This system can potentially make the lung nodule reviewing process more efficient and more accurate.

H.C.,F.M.,S.A.W.,A.S.: H. Chui, F. Ma, A. Schneider and S. Wood are employees at R2 Technology, Inc.

Chui, H, Ma, F, Schneider, A, Judy , P, Wood, S, An Automatic Growth Detection CAD System for Analyzing Lung Nodules in Temporal MDCT Scans.  Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL. http://archive.rsna.org/2004/4415126.html