An automated machine vision system for the histological grading of cervical intraepithelial neoplasia (CIN)

The histological grading of cervical intraepithelial neoplasia (CIN) remain s subjective, resulting in inter- and intra-observer variation and poor rep roducibility in the grading of cervical lesions. This study has attempted t o develop an objective grading system using automated machine vision. The a rchitectural features of cervical squamous epithelium are quantitatively an alysed using a combination of computerized digital. image processing and De launay triangulation analysis; 230 images digitally captured from cases pre viously classified by a gynaecological pathologist included normal cervical squamous epithelium (n=30), koilocytosis (n=46), CIN 1 (n=52), CIN 2 (n=56 ), and CIN 3 (n=46). Intra- and inter-observer variation had kappa values o f 0.502 and 0.415, respectively. A machine vision system was developed in K S400 macro programming language to segment and mark the centres of all nucl ei within the epithelium. By object-oriented analysis of image components, the positional information of nuclei was used to construct a Delaunay trian gulation mesh. Each mesh was analysed to compote triangle dimensions includ ing the mean triangle area, the mean triangle edge length, and the number o f triangles per unit area, giving an individual quantitative profile of mea surements for each case. Discriminant analysis of the geometric data reveal ed the significant discriminatory variables from which a classification sco re was derived. The scoring system distinguished between normal and CIN 3 i n 98.7% of cases and between koilocytosis and CIN 1 in 76.5% of cases, but only 62.3% of the CIN cases were classified into the correct group, with th e CM 2 group showing the highest rate of misclassification. Graphical plots of triangulation data demonstrated the continuum of morphological change f rom normal squamous epithelium to the highest grade of CIN, with overlappin g of the groups originally defined by the pathologists, This study shows th at automated location of nuclei in cervical biopsies using computerized ima ge analysis is possible. Analysis of positional information enables quantit ative evaluation of architectural features in CIN using Delaunay triangulat ion meshes, which is effective in the objective classification of CIN. This demonstrates the future potential of automated machine vision systems in d iagnostic histopathology. Copyright (C) 2000 John Wiley & Sons, Ltd.