Histogram based Analysis of Lung Perfusion of Children after Congenital Diaphramatic Hernia Repair
Nora Kassner1, Meike Weis2, Katrin Zahn3, Thomas Schaible4, Stefan O Schoenberg2, Lothar R Schad1, K Wolfgang Neff2, and Frank G Zöllner1

1Computer Assisted Clinical Medicine, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany, 2Institute of Clinical Radiology and Nuclear Medicine, University Medical Center Mannheim, Heidelberg University, Mannheim, Germany, 3Department of Pediatric Surgery, University Medical Center Mannheim, Heidelberg University, Mannheim, Germany, 4Department of Neonatology, University Medical Center Mannheim, Heidelberg University, Mannheim, Germany

Synopsis

Reported measured lung perfusion data of 2-year old children after congenital diaphragmatic hernia repair was evaluated by regions of interest (ROI) within the acquired 3D volume. In this work a histogram based approach is used to characterize the distribution of perfusion in the whole left and right lung, and suitable quantities to characterize the distribution are extracted.

Introduction

Congenital Diaphragmatic Hernia (CDH) is a malformation of the diaphragm, allowing organs of the abdominal cavity to enter the chest and damage the lung. In cases of isolated CDH, the severity of lung hypoplasia is the main predictor of outcome concerning both mortality and lung morbidity for example represented by the development of a chronic lung disease1. With MRI, lung perfusion can be measured without the need for ionizing radiation and additional three-dimensional morphological information is available during one examination. Previous studies have shown reduced quantitative perfusion values on the ipsilateral side in comparison to the contralateral side2, 3.

In these studies, however, the measured lung perfusion data was evaluated by cylindrical regions of interest (ROI) within the acquired 3D volume3. In this work, a histogram based approach is used to characterize the distribution of perfusion in the whole left and right lung.

Materials and Methods

29 children (age of 24.2 ±1.7 months; 1 right sided hernia) were imaged according to the local follow up program which was approved by the local IRB. Written consent of the parents was obtained. The examination was performed at a 3T system (Magnetom TimTrio,Siemens Healthcare Inc., Erlangen, Germany) and a time-resolved angiography with stochastic trajectories (TWIST) sequence was used with parameters TR/TE/FA=2.3ms/0.8ms/15°, matrix=192x162x56, voxel resolution of 2.0x2.0x2.0 mm³, TWIST view sharing with outer/inner sampling density of 15%/20%, and PAT 3 resulting in a temporal resolution of 1.5s. After recording 5 volumes the contrast agent (Dotarem, Guerber, France) and a sodium solution (both 0.05 mmol/kg body weight) were injected followed by a 10 ml saline flush. Perfusion values were calculated by a pixel-by-pixel deconvolution approach2.

Based on manual segmentations, histograms were created for the left and right lung, respectively. To fi nd the right trade-off between individual characteristics of each histogram for every patient, and to ensure comparability, an optimal bin size was estimated using the Freedman-Diaconis rule4. Furthermore, as some subjects show high perfusion, values higher than 450ml/min/100ml are cut off .

For each histogram mean, median, standard deviation, geometric mean, geometric standard deviation, skewness, and kurtosis are calculated. The diff erence in distribution of the ipsilateral and contralateral lung are evaluated by the Kolmogorow-Smirnow (KS) test. Previous studies document a lognormal distribution of pulmonary blood flow5. To estimate whether the distribution of lung perfusion values of children with CDH also follows such distribution, the data is fi tted to a lognormal probability density function.

Results

The histogram of perfusion values including data of all children, separating only the left and the right lung, shows impaired perfusion inside the ipsilateral lung by a smaller mean, median and geometric mean. This histogram is presented in Figure 1 and an overview of all computed quantities is given in Table 1. The KS test denotes a signi cant di fference in distribution between left and right lung (p=0.002). A lognormal distribution fi ts the histogram perfusion data in both sides well (see Fig. 2). A positive skewness in both lungs indicates a left tilted distribution, meaning the tail on the right side is longer and more data lies on the left side of the mean. This tilt becomes even stronger in the ipsilateral lung with a skewness of 1.9 in comparison to 1.6. Furthermore, the kurtosis of the data distribution is 6.0 in the contralateral and 7.6 in the ipsilateral lung indicating a sharper curve compared to a Gaussian. This sharpness is also more pronounced in the left side. Similar observations can be made in case of the right sided hernia (see Fig.3), however, here the hypolastic lung is on the right side, respectively. Visual differences between left and right lung are depicted in Figure 5.

Discussion

A histogram based analysis of whole lung is an appropriate approach to characterize lung perfusion of CDH patients. In total, impaired lung perfusion is observed in the ipsilateral lung, indicated by a lower mean, median and geometric mean in most subjects. This is in concordance with previous reports using ROI-based analysis2,3.

Mean and standard deviation are very sensitive to extreme values. Instead of using mean and standard deviation to characterize perfusion, more reliable quantities are the median or geometric mean and the geometric standard deviation, particularly as the perfusion data follows a lognormal distribution.

In conclusion, we showed that histogram analysis is a valuable tool to characterize and visualize whole lung perfusion of children after CDH repair. In future we plan to use this tool to investigate and track changes in lung perfusion of our patients within our follow up program.

Acknowledgements

No acknowledgement found.

References

[1] Stege G, et al., Pediatrics, 2003;112(3 Pt 1):532-5.

[2] Zöllner FG, et al., Eur Radiol 2012;22:2743-2749

[3] Weidner M et al., Eur Radiol, 2014;24(10):2427-2434

[4] Freedman, D and Diaconis, P, Probability Theory and Related Fields, 1981; 57(4):453–476

[5] Young R, et al. J Magn Reson Imaging, 2007;26:1053-1063

Figures

Figure 1: Histogram of PBF of all children. top: red bars indicate the right lung, blue dots indicate the left lung. bottom:blue bars indicate the left lung, red dots indicate the right lung.

Table 1: Histogram quantities derived from the PBF of all children.

Figure 2: Fit of the lognormal distribution to the histogram data of all children. Top: right lung side, blue dots depict the left lung histogram, solid red line depicts the lognormal fit. Bottom: left lung side, red dots show the right lung histogram, solid blue line the lognormal fit, respectively.

Figure 3: Example for a patient with a right sides hernia. Top: red bars indicate the right lung, blue dots the respective histogram of the contralateral lung. Bottom: blue color represent the left lung, red dots the ipsilateral lung, respectively. Compared to a left side hernia, the distribution are switched but similarily, the hypolastic lung can be distinguish from the contralateral lung.

Figure 5. Example of a maximum intensity projection (MIP) (A) and a PBF map (B) and a PBV map (C) calculated in one slice of a patients's DCE-MRI. The MIP was calculated inline with the exam while the PBF and PBV map was calculated using in house developed perfusion tool as described in (2,3). In all images a clear difference between left and right lung can be detected.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
2915