Airway Measurements Overview

The bronchial lumen size and bronchial wall thickness can provides valuable information regarding diseased states of the respiratory system. For instance, changes in the bronchial lumen of an airway can be used to evaluate airway reactivity which could be useful in evaluation of conditions such as asthma. Measuring the enlargement of terminal airways might provide helpful information about the severity of enlargement of airways of patients with bronchiectasis. Since cystic fibrosis involves thickening of the airway walls, measuring the lumenal to wall thickness ratio may provide means of estimating the severity of disease. Last, in its 3-D implementation this can serve as the starting point for digital bronchoscopy.

In order to accurately quantitate the luminal area of a CT image of the airway, the edge between the soft tissue and the airway needs to be objectively defined. The algorithm used in VIDA to detect airway edge is based upon the "full-width-half-maximum" principle commonly used for radiologic images, meaning that measurements are made at a level equal to one half the maximal range of densities which are present in a given area of interest. To measure a structure's diameter with this algorithm,a line is drawn across the structure extending into the tissue surrounding it. A profile of the image brightness along this line represents varying x-ray densities (Hounsfield Units) in the structures the line crosses. The change in slope of the curve at the boundaries of components of different densities (or brightness) reflects both the physics of the imaging system and "partial volume effect" of structures involved. In "partial volume effect", pixels encompassing more than one structure take on the average density value proportionate to the amount of each tissue of different density present. By setting the level at half way between the baseline and the peak of the density curve, the structure`s diameter is measured as the distance between inflections of the curve at that setting. The edge finding algorithm thus employs a regionally set boundary level in its semi-automated program to determine an objective margin of a structure, in this case the airway lumen.

The user draws a first, rough estimate of the margin of the airway lumen (the image can be magnified without effecting the calculations) which is then smoothed using a three point running average, low pass filter. This filter is applied to the rough polygon until the user determines that the polygon is visually smooth (usually about 4 passes). Points defining the polygon are then added to the smoothed contour to assure a minimum number of points. Finally, at each point on the polygon, a ray perpendicular to the local contour is cast with half its length inside of the polygon and half outside. The length is adjusted for the size of the airway of interest and for the proximity of other unwanted, nearby edges. In this study, we used a length of 6 pixels for airways 4mm diameter or smaller and a length of 16 for larger airways. The algorithm automatically finds the baseline density of air inside of the airway and the plateau density of soft tissue clearly outside of the airway lumen and the spatial location of the brightness half way between the base and the plateau is defined as the true airway edge; the point on the polygon is moved to this new location. After all points on the polygon have been adjusted, the polygon is again smoothed using the low pass filter and the area calculated. The smoothing of the initial drawing is of importance since, if the originally drawn edge is jagged, the rays cast in the next step could yield an erroneous path for brightness evaluation.This technique yields consistent edge estimates even when confronted with widely varying initial drawings.

To compensate for potential through plane motion of the airway location from scan to scan, one scan in front of and one scan behind the "matched" level were also used for airway luminal measurements. Airway area is measured 5 times at each level for the 3 levels; thus, each airway cross sectional area value reflects an average of fifteen measurements.

Last modified: Fri Jun 4 13:54:26 CDT