OBJECTIVE NUCLEAR GRADING FOR NODE-NEGATIVE BREAST-CANCER PATIENTS - COMPARISON OF QUASI-3D AND 2D IMAGE-ANALYSIS BASED ON LIGHT-MICROSCOPIC IMAGES

In a retrospective investigation for a new image-analytical nuclear gr ading method, we used 145 routine hematoxylin and eosin-stained, paraf fin-embedded tissue sections from node-negative breast carcinomas. Cel l fields of primary tumors were scanned in a light microscope in succe ssive focus levels in 1-mu m steps for thick sections (greater than or equal to 5 mu m: quasi-3D analysis) and in one focus position for thi n sections (< 5 mu m: 2D analysis). After image-segmentation, nuclear features for texture and chromatin distribution were calculated. A bin ary classification tree was constructed for determination of two mathe matically defined classes of high- and low-risk tumor cell nuclei. Aft er fixing a cut-point for the portion of high-risk tumor cell nuclei p er patient, it was possible to distinguish two different groups with s ignificantly different relapse rates of 4.2% and 74.5% in quasi-3D ana lysis and 0.0% and 52.0% in 2D analysis, respectively. Large differenc es between quasi-3D and 2D analysis were only present in the classific ation of nonrelapse patients, whereas nearly all patients with relapse had more than 50% high-risk tumor cell nuclei. The results show that the information in thicker tissue sections contains important additive components in the third dimension, with respect to the detection of c hromatin structure and distribution. This advantage should be exploite d for the development of an objective image-analytical nuclear grading system as a highly significant prognostic marker.