Computer Determination of Alveolar Number per Unit Area of Lung Tissue

Abstract

The total nuber of alveoli in the human lung has long been a matter of interest and of some dispute for many years. This measurement, as well as the size of the individual alveoli are of fundamental biological importance in terms of the growth and development of the lung, as well as in its aging. Further, the number of alveoli is of interest and significance to pulmonary physiologists. 

One method of calculation presently in use is based upon a random sample of lung tissue from which can be determined the average number of alveoli per unit area of tissue. The latter is obtained by visually examining the sample under magnification and counting the number of alveoli which can be related through a known formula to the internal surface area. However, the technique is an extremely tedious one for a human and has been shown not to yield repeatable results. In order to overcome these problems, a mehtod of computerized image processing which would automatically perform the computation was invoked. 

The computer analysis of the lung tissue first involves the digitization of the image on the histological slide by the McGill Scanner (McScan) System. Next a binary matrix representation of the image is calculated using a normalization procedure which takes into account image variability. The alveolar walls are thinned and topological considerations are then used to determine the number of alveoli per unit area.

The application of the method to lung samples previously processed visually has yielded comparable results.