ParticipantsMaximiliano Klug, MD, Ramat Gan, Israel (Presenter) Nothing to Disclose
Michael Green, MSc, Tel Aviv, Israel (Abstract Co-Author) Nothing to Disclose
Arnaldo Mayer, PhD, Los Gatos, CA (Abstract Co-Author) Nothing to Disclose
Eli Konen, MD, Tel Hashomer, Israel (Abstract Co-Author) Nothing to Disclose
Edith M. Marom, MD, Tel Aviv, Israel (Abstract Co-Author) Speaker, Bristol-Myers Squibb Company; Speaker, Boehringer Ingelheim GmbH; Speaker, Merck & Co, Inc; Officer, Voxellence ; ; ; ;
maxiklug@hotmail.com
PURPOSETo assess the effect of a denoising method (D) for ultra-low-dose CT (ULDCT) on infection assessment in immunocompromised patients.
METHOD AND MATERIALS25 consented adult immunocompromised patients referred for chest CT, had 2 scans: a normal dose CT (NDCT), 120 kVp and automatic current modulation, immediately followed by an ULDCT, 100 kVp and fixed current at 10 mA. Each ULDCT was denoised using a locally consistent non-local-mean (LCNLM) algorithm to obtain a high signal to noise ratio version of the ULDCT. Blinded to all clinical information, a radiology resident and a chest radiologist assessed the ULDCT, and denoised ULDCT(D), images as compared to the NDCT, documented findings, and compared between them. Our IRB approved this study.
RESULTSThe average ULDCT radiation dose was 2% of NDCT, with an average effective radiation dose of 0.13 mSv. 3/25 patients were correctly classified as normal lungs by ULDCT(D) whereas only 2/25 by ULDCT. Both methods identified pneumonia and features consistent with fungal pneumonia. ULDCT(D) was better in differentiating the consistency of micronodules<1cm compared to ULDCT, with an increased accuracy (acc.) for solid (92% vs. 80%) and subsolid (80% vs. 72%) micronodules. ULDCT(D) was also more accurate for detecting focal GGO (acc. 96% vs. 80%), and in determining the main pattern of distribution of GGO: central (acc. 96% vs. 88%) and peripheral (acc. 96% vs. 84%). Fine details were better appreciated with ULDCT(D): tree-in-bud opacities (acc. 100% vs. 80%), interlobular septal thickening (acc. 84% vs. 72%), intralobular thickening (acc. 80% vs. 56%) and pericardial effusion (acc. 96% vs. 84%).
CONCLUSIONULDCT is good for identifying pneumonia and normal lungs in the immunocompromised patient and in suggesting fungal infection. However, ULDCT(D) outperforms ULDCT in the identification of fine details. Performing ULDCT(D) in lieu of NDCT in young patients expected to undergo repetitive CT scans for infection evaluation should be considered to reduce cumulative radiation dose while preserving diagnostic accuracy.
CLINICAL RELEVANCE/APPLICATIONDenoising ULDCT with the LCNLM algorithm correctly identifies pneumonia in immunocompromised patients with dose reductions greater than 95%.