September 1997 - September 2000
B.C. Stoel H.A. Vrooman
Pulmonary emphysema is the main cause of chronic obstruction of the airways and leads to more than 100.00 deaths per year in the United States only. At the moment, there is no adequate diagnostic method to detect emphysema at an early stage. For the development of therapeutic procedures for emphysema it is essential that the stage of the disease is determined accurately. In particular for patients with minor or moderate emphysema an accurate diagnosis is important, since in these cases it is sensible to slow down the development of emphysema, for example with protease inhibitors.
The lung function parameters that have been used until now, such as CO diffusion and FEV1, are not sensitive enough to detect small changes in the disease process. Visual (qualitative) assessment of conventional X-ray images or even advanced diagnostic techniques (such as nuclear medicine or computed tomography (CT)) are still not reliable and accurate enough to detect emphysema in an early stage, if these images are assessed visually. It is therefore our belief and that of others, that the diagnosis can be improved in the sense that more objective and reproducible data can be extracted from these images, if modern image processing techniques are applied.
Innovative research have been carried out by Wegener in the field of quantifying emphysema using CT, who indicated in the late seventies that CT quantification offers the opportunity to perform follow-up studies in patients with lung diseases in general. Since this point in time, the quality of CT scanners have been improved tremendously. This has also brought a strong improvement in the reliability of the density measurements. Especially the group of McNee (Edinburgh, GB) has disseminated the use of CT quantification and has proven its efficacy in a number of clinical studies. The CT technology now seems to be reliable enough to be used for the diagnosis of emphysema and for the evaluation of potential treatment of emphysema.
In a former project sponsored by the Netherlands Asthma Foundation (NAF 92.70) over a period of 3 years (January 1993 - January 1996), a method has been developed to detect emphysema using CT densitometry. This lead to a software program that detects the lungs slice for slice using 2-D segmentation techniques. From this study it became clear that in order to accurately and reproducibly segment the lungs in CT images, some major adjustments should be made including 3-D segmentation techniques.
The main objective of this project was to determine clinically relevant parameters that can be obtained from CT images for diagnosis, progression and regression of pulmonary emphysema and other density related lung diseases.
A second objective was to develop the corresponding algorithms that can be used for the assessment of target areas for lung volume reduction surgery.
In order to facilitate the above applications, the segmented lungs should be visualized in 3-D.
The software package "Pulmo-CMS" detects the lungs in CT scans and automatically analyses the density distribution of the lungs. This software package has been developed for a standard PC and runs under the Windows-NT and Windows-2000 operating system. The software tool can therefore communicate with other Windows-based programs, such as MS-Office's Word, Excel, Access etc. Pulmo-CMS is able to read CT images obtained with scanners from different manufacturers through the image standards ACR-NEMA 1.0, ACR-NEMA 2.0 and DICOM 3.0. Images can be read from several storage media, like hard disks, CDs or a connected network.
Basically, the software package consists of four parts: 1) calibration 2) (semi-)automatic selection of the lungs in the three-dimensional image data; 3) calculation and subsequent analysis of the density distributions; and 4) presentation of the measurement results.
First, the densities in the images can be calibrated by rescaling these densities depending upon the measured density of the blood in the patient. By this calibration procedure, changes in the spectrum of the X-ray tube, due to tube ageing, and changes in the patient's blood density, due the emphysema, are taken into account. The measurement of the mean blood density is carried out by a semi-automatic detection of the largest artery, the aorta, and by calculating the mean density value within this region.
In the second step, the lungs are selected automatically by having the user define a location within the trachea. Subsequently, the software carries out a so-called region growing, which lets this initial location expand (like a balloon) until it reaches the borders of the lungs. The septum (the border between the left and right lung), the trachea and the carina are detected separately. The lung parenchyma is defined by excluding the septa, trachea and large vessels from the initial segmentation result. By doing so, the lungs are selected with a high degree of reproducibility: two CT scans of the same subject will give the same analysis results. A number of editing tools have been developed to facilitate user-friendly procedures for correcting the contours; after editing the program checks the resulting contours for inconsistencies.
In the third step, the histograms of the left and right lung parenchyma are calculated. From these histograms fiveparameters are derived: the total lung volume, mean lung density, the lung weight (which is equal to the product of volume and density), the 15th percentile point (as described above) and the area of the lungs (in %) below a certain density value, called the relative area. Finally, these results are presented to the user, and can be saved on disk to facilitate further statistical analysis by spreadsheet programs or statistical packages. The contours of the left and right parenchyma can be saved to disk and, if the users wants to analyze the histogram data by a separate computer program, the raw histogram data can be saved to disk in text (tab delimited) format.
For the planning of lung volume reduction surgery, the program can show regions in the segmented lungs with the highest degree of emphysema, displayed in a purple overlay.
The project finished in 2000.
For further information, please contact:
Berend C. Stoel, PhD.
Division of Image Processing
Department of Radiology, 1-C2S
Leiden University Medical Center
P.O. Box 9600
2300 RC Leiden
Tel. +31 (0)71 526 1911
Fax. +31 (0)71 526 6801