Image Processing and Cancer Detection

The Biology Project: Mathematics and Biology

Image Processing and Cancer Detection

Classification of Nuclei

The final phase in this process of quantitative analysis is to incorporate all measurable features of a sample into a single cell profile. Each of the 99 features is analyzed according to its standardized distance from a corresponding normal value. The "Average Standardized Distance from Normal," then, is an average of each of these 99 values, where distance is measured in standard deviations. Distance from "normal" of nuclei in terms of different progression graph

Since even "normal" cells are never identical, some small variation is seen - up to two standard deviations. Different stages of tumor development can now be identified based on this standardized difference from the profile of a normal tissue sample.

Once a profile has been generated, lesions can be classified according to degree of abnormality and grade of progression. Highly specific determinations can be made at this point regarding optimum patient management.

Specific Classification of Individual Patient Lesions classification of lesions

Although a pathologist would diagnose both of these lesions as cancerous, only through the processes of digital and statistical analysis can this marked difference in severity be illustrated.

Whereas traditional cancer survival statistics are based primarily on life span averages for patients with similar lesions, application of image processing technology could make available a lesion-specific prediction that is more accurate than has ever before been possible.

The Biology Project
The University of Arizona
March 26, 1998
denicew@u.arizona.edu