top of page
Writer's pictureLetícia Kawano-Dourado

ALAT 2019 poster session, Panamá

Correlation Between Quantitative and Semi-Quantitative Computed Tomography Imaging on the evaluation of progression of Interstitial Lung Disease in Rheumatoid Arthritis


Authors: Leticia Kawano-Dourado1,2, Marcio V. Y. Sawamura3, Karina Bonfiglioli4, Jose Geraldo Alves Brito Neto3, Lorena Vaz Vieira3, Maria Laura Bertozo Sabbag5, Camila de Assis Molina5, Fábio E. Arimura1, Renato H. N. Santos2, Carlos R. R. Carvalho1, Ronaldo A. Kairalla1


1 Pulmonary Division, Heart Institute (InCor) - HCFMUSP, Medical School, University

of Sao Paulo, Sao Paulo, Brazil

2 Research Institute, Hospital do Coracao (HCor), Sao Paulo, Brazil

3 Radiology Division, Hospital das Clinicas - HCFMUSP, Medical School, University

of Sao Paulo, Sao Paulo, Brazil

4 Rheumatology Division, Hospital das Clinicas - HCFMUSP, Medical School,

University of Sao Paulo, Sao Paulo, Brazil

5 Centro Universitario Sao Camilo, Medical School, Sao Paulo Brazil;


Corresponding author: Leticia Kawano-Dourado

Email: leticiakawano@gmail.com




Background: The semi-quantitative HRCT analysis as per Goh 2008 has been demonstrated in Scleroderma and in rheumatoid arthritis (RA) to carry prognostic information [Sathi N, 2010]. Quantitative image analyses have been developed bringing more objectivity to ILD assessment. Recently it has been demonstrated that the visual semi-quantitative method proposed by Goh performs as well as automated CT texture analysis (CALIPER®) in RAILD and are associated with mortality [Jacob J et al, ERJ, 2019]. Our aim is to compare a densitometric* quantitative automated HRCT image analysis technique vs. Goh´s semi-quantitative analysis.

* densitometric CT analysis: high attenation areas, kurtosis and skewness


Methods: Single-center RA-ILD adults with at least two HRCTs available were retrospectively identified (n=39).

Visual semi-quantitative analysis as per Goh et al, 2008: First and last HRCTs of each patient were independently evaluated by two raters (M.S. and LKD), who were blinded to additional information. %ILD involvement ranged from 0% - 100%. The %ILD progression was obtained by the difference between last and first HRCT scores.

Goh´s analysis adapted:

The total extent of interstitial lung disease was estimated to the nearest five percent in each of the five sections, with global extent of disease on HRCT computed as the mean of the scores.


Five sections (anatomical markings):

The % ILD of the whole lung is the MEAN of the % of ILD of the 5 sections. Below you find images of the 5 levels:


Slice 1

Slice 2

Slice 3

Slice 4


Slice 5


Densitometric Quantitative analysis: A free software, SlicerCIP was operated by a radiologist (JGABN) blinded to any additional information. Kurtosis, skeweness and high attenuation areas (HAA) defined by % lung volume with attenuation between -600 and -250 Hounsfield units (HU) were recorded.

The higher the Kurtosis and the skewness, the more ILD there is. See below:


Graphic explanation of Kurtosis and Skewness in densitometric CT scan analyses


Statistical analysis: Correlation between %ILD and kurtosis, skewness and %HAA were performed using Pearson coefficient. The difference between last and first HRCT gave us delta values. Delta %HAA, delta kurtosis and delta skewness were compared across the %ILD progression categories (≤ 5%, >5% and ≤10%, and >10%) using Kruskal-Wallis test.


Results: Mean age 66±9.6, 75% females, 44% ever smokers, articular disease duration 10.5±10.6.

Baseline %ILD involvement was 6% [ 3 – 13], %HAA was 6% [ 5 – 9.5], kurtosis 9.2 [6.8 – 13.4], skewness 2.8 [2.5 – 3.3].

Time between first and last HRCT: 4.4±2.3y. Cases were categorized as per semi-quantitative Goh´s analyses as:


1. Less than 5% ILD progression: 69% of cases (n = 27)

2. %ILD progression >5% and ≤10%: 20% of cases (n = 8)

3. %ILD progression > 10%: 11% of cases (n = 4)


There was a strong correlation between %ILD and %HAA (r=0.74, p<0.001 and 0.85 after removal of 3 outliers), kurtosis (r= -0.64, p<0.01) and skewness (r= -0.72, p<0.001).

Note the correlation between absolute densitometric variables depicted in the Y-axis with the % of ILD involvement in the X-axis. Of note: the correlation between the variation (delta) densitometric variables and the delta %ILD in 4.4 years is depicted in the next figure

It was also observed that the greater the %ILD progression, the greater the delta %HAA, delta kurtosis and delta skewness, figure below:

On the Y-axis the densitometric variables: High attenuation areas (HAA), Kurtosis and Skewness. On the x-axis the % of interstitial lung disease (ILD) progression in +- 4.4 years assessed by the semi-quantitative method adapted from Goh et al. Note the correlation between the variation (delta) densitometric variables and categories of %ILD progression

Conclusions: We observed a strong correlation between the quantitative HRCT HAA and the semi-quantitative analysis (% of ILD involvement as per Goh´s adapted method). The intra-patient variation in % ILD in 4 years was also consistently associated with variations in % HAA, kurtosis and skewness. This is preliminary evidence in favor of quantitative methods to assess disease progression in RAILD. And demonstrates good correlation of densitometric quantitative CT analysis with semi-quantitative visual methods, already associated before with outcomes such as mortality.



This work is to be presented at ALAT 2019 in a poster session. The poster for this work follows the new design proposed by Michael Morrison and you can see it down below:






Comments


bottom of page