Shailaja Sajja MS, Presenter: Nothing to Disclose

Samuel Richard PhD, BSC, Abstract Co-Author: Employee, Carestream Health, Inc

David Foos MS, Abstract Co-Author: Employee, Carestream Health, Inc

Scott Boerner MD, FRCR, Abstract Co-Author: Nothing to Disclose

Narinder S. Paul MD, Abstract Co-Author: Research funded, Toshiba Corporation

To discriminate bloody from non-bloody pleural effusions using multi-spectral portable digital radiography (DR).   

Thirty seven (13 bloody and 24 non-bloody) pleural aspirates were classified by an experienced cytopathologist as bloody or non-bloody and the classification was blinded to the research team. The samples were imaged with a portable DR unit (Carestream Health DRX-1C retrofit with GE AMX4) at 5 x-ray energies (60, 70, 80, 110 and 120 kVp) - with adjusted mAs to ensure fixed entrance surface dose (50 uGy). The mean signal attenuation (μx) for each pleural sample was calculated across kVp by selecting a corresponding ROI of the sample in the image and computing the mean-log signal. Values of μx were plotted vs. kVp and the behaviour of the curves for a sub-sample of 5 pleural samples was analyzed from which a set of classifying metrics (K) were derived based on the ratio of the signal attenuations at different energies. Metric K1 was the ratio of the difference of μx at (120, 60) kVp and (70, 60) kVp - namely [(μ120x-μ60x)/(μ70x-μ60x)=(μ120-μ60)/(μ70-μ60) - independent of sample thickness x], similarly for K2=(μ110-μ60)/(μ70-μ60) and K3=(μ120-μ80)/(μ70-μ60). The 37 pleural samples were analyzed to determine the accuracy of using the metrics to discriminate between bloody or non-bloody fluids and the false positive rate (FPR) and true positive rate (TPR) were calculated for various thresholds to yield the receiver operating characteristic curves ROC.

The ROC was plotted for the 37 blinded pleural samples. The area under the operating curve (AUC) was 0.66, 0.67, and 0.56 for K1, K2, and K3, respectively. Based on these results, K2 was found to be most suitable for differentiating between bloody and non-bloody effusions. This was mainly due to the greater difference in x-ray energies used for the measurements.

In this work, a novel metric based on the multispectral x-ray signal was defined and tested on 37 pleural effusion samples. This metric may provide a reasonable differentiation between bloody and non-bloody pleural effusions. Further work is required to evaluate clinical feasibility.

We report initial results of a novel multi-spectral portable DR system, which may have the potential for discriminating between pleural effusions types (i.e., bloody from a non- bloody effusion).  

Sajja, S, Richard, S, Foos, D, Boerner, S, Paul, N, Tissue Characterization using Multispectral X-ray: Description of a Novel Metric for Categorization of Pleural Effusions .  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14009644.html