6th ESACP Congress, Heidelberg, April 7-11, 1999

A126
BREAST CANCER: SPATIAL HETEROGENEITY, HISTOLOGICAL GRADES AND GEOSTATISTICS
Sharifi-Salamatian V, Rigaut JP, de Roquancourt *

Laboratoire d'Analyse d'Images en Pathologie Cellulaire (AIPC), * Service de Pathologie; Hôpital St. Louis, Université Paris 7, Paris, France

Chromosomal patterns specific of a type of cancer are present even though phenotypic spatial heterogeneity is omnipresent. If any invariance paradigm representing a stochastic, chaotic, or geostatistic function could be discovered, this might help in solving the problem of the insecurity ot histopathological grading (40% for breast cancer). We demonstrated through multiimmuno-marking that breast cancers display a large heterogeneity, thus presenting a major problem for any grading. Geostatistics represent a set of methods which have not been used much outside of a mining context. Standardised microscopical slides in paraffin with labetling by MIB-1 (Ki-67) allow numbers and coordinates of labelled nuclei to be filed. The dispersion variance is then estimated and an invariant function based on ergodicity and integral range allows a bypass of the heterogeneity problem. The first results are very encouraging. The grades are well separated; the certainty of a correct classification will, however, require a confrontation with the results of survival studies.