LungCTAnalyzer and fibrosis quantification

May be I am wrong but it seems LungCTAnalyzer cannot be used to quantify lung fibrosis. Perhaps this feature will be added to the extension in the future. Fibrosis quantification will be very useful for many. I have read about some software for this purpose but either they are old or unavailable or not deployable on todays computers, like LUFIT and YACTA, or commercial and pricey, like IMBIO.

What do you want to see, which patterns are of special interest for you , what would you want to quantify in ml. Please make an anonymized dataset available so we could have a look at it in Slicer. You could also post some screenshots here where you think LCTA fails.

@rbumm

Dr. Bumm, thanks for responding. I am a novice and actually don’t know myself but was wondering if some expert had assessed LungCTAnalyzer for the purpose of quantifying fibrosis in ILD, IPF, etc. by measuring GGO, honeycombing, reticular density, etc. (CT pattern or texture analysis). E.g., is ‘infiltration’ identified by LungCTAnalyzer a good measure of GGO?

Below, I have copied some text about the method used by the CALIPER CT pattern analysis software that purports to quantify fibrosis.

Following anatomic segmentation, the parenchymal detection and classification is performed by using a similarity metric to match parenchymal histogram features within 15x15x15 pixel volumes of interest to validated histogram signatures of characteristic voxels and morphological assessment of classified voxels in order to label each pixel in the dataset with the radiographic characteristics. These labels specify a pixel as belonging to normal lung parenchyma or to features of ILD such as ground glass opacity, reticular densities, honeycombing, or low attenuation areas (with sub-classes of mild, moderate and severe). The technique involves a sliding window supervised classification scheme described by Maldonado et al. [7] The total volume, measured in liters, and the percentage of normal parenchyma, low attenuation areas (mild, moderate, and severe sub-classes), and interstitial abnormalities (GGO, RD, and HC) were automatically generated by the CALIPER software. The total volume was comprised of low attenuation areas (mild, moderate, severe), ground glass opacities, honeycombing, reticular densities, normal areas, and vessel volumes. Total interstitial abnormalities were defined as the summation of GGO, RD, and HC. The remainder of the lung volume (vessel volume, normal and low attenuation areas) was summed as non-involved lung areas.

I have no experience with CALIPER, nor have I analyzed ILD patients yet.
Pattern recognition is something you can not do with Lung CT Segmenter, which was made to determine thresholded areas of special lung density. i.e. pre-setting a threshold value of Hounsfield Units and obtaining a segmentation of both lungs and a quantitative evaluation of emphysema, healthy residual lung parenchyma, GGO, and consolidation (like in this paper)

1 Like

Here is a Lung CT Analyzer result in a patient with lung fibrosis and a good visual correlation of the fibrotic lung changes and their detected volumes. I do not see why you could not use LCTA this way in IPF - if you need more threshold ranges please let us know.

1 Like