No abstract

SUMMARY A review of aspects relating to microcomputer video image analysis is presented. The technical limitations of early systems have been surpassed by high performance video digitizers and microcomputers which are now widely marketed. Commercial systems based on these devices are increasingly available. Also, with the programming skills in most laboratories today, software development for specific applications is a practical proposition provided that sufficient expertise in video and computer technology is available. Considerations necessary for effective hardware selection and software design are discussed by example of the application to light microscope densitometry. Attention is drawn to errors which can arise when employing conventional video systems for analytical purposes.

The differentiation between Spitz naevus and melanoma is at times difficult. The present study was undertaken to define means to positively identify such melanocytic tumours of doubtful malignancy. Immunohistochemical staining intensity for S100 protein and neurone specific enolase (NSE) was measured in sections of 35 Spitz naevi using a microcomputer image analysis system. The data were compared with results previously obtained from 19 cases of malignant melanoma and 16 benign compound naevi. Disaggregated cells from paraffin-embedded material were stained by the Feulgen technique for DNA estimation. The nuclear DNA content distributions were measured using the same image analysis system. Compared with the malignant cases, the Spitz naevi showed significantly lower staining intensity for both S100 protein (P less than 0.0001) and NSE (P less than 0.0001). When compared with the benign compound naevi, the staining intensity was significantly lower for S100 protein (P = 0.003). The nuclear DNA distribution in Spitz naevi proved to be a normal diploid pattern in 31 cases. Four cases showed a small proportion of hyperdiploid nuclei. The results show that Spitz naevi can be significantly distinguished from malignant melanoma by staining intensity for S100 protein and NSE. A normal diploid DNA content distribution appears to be typical for Spitz naevi. Spitz and benign compound naevi show dissimilar expression of S100 protein which may indicate different patterns of differentiation in these two types of lesion. The image analysis equipment used in this study is accurate, simple to use, produces results rapidly, and is economic. Therefore, it is clinically practicable.

No abstract

The grading of gastric epithelial dysplasia has been studied by means of computer aided morphometry. Measurements of histological features were made on segments of epithelium from 38 selected cases of dysplasia graded by subjective assessment. The measurements were statistically compared with the subjective scores using discriminant analysis. The measurements were found to provide significant discrimination between all groups. Nuclear size proved to be the main discriminating variable. Prediction of the likely group membership of individual cases was possible using classification equations derived from the discriminant analysis. Classification of the original data set revealed prediction errors which suggested a bias against diagnosis of dysplasia. As a result of this study reliable, repeatable and objective gradings of gastric epithelial dysplasia can be obtained by inexperienced persons with an accuracy approaching that of a skilled pathologist.