A comparison of image registration techniques for the correlation of radiolabelled antibody distribution with tumour morphology

Phys Med Biol. 1999 Jul;44(7):N151-9. doi: 10.1088/0031-9155/44/7/402.

Abstract

Image registration is a powerful tool for correlating functional images with images of anatomical structure. This facilitates more accurate quantitation of regional radiopharmaceutical uptake. Similarly, registration of images of radiolabelled antibody distribution, in tissue sections, with the equivalent histological images allows the comparison and measurement of radiopharmaceutical distribution with morphological structure. The images used were obtained by storage phosphor plate technology, for the radiopharmaceutical distribution, and by digitization of the stained histological sections. Here we compare four fully automatic registration techniques and one manual technique in terms of their spatial accuracy. We have found that there was no difference in accuracy between cross-correlation, minimization of variance and mutual information. These techniques were more accurate than principal axes and the manual technique. However, minimization of variance and mutual information were more time-consuming than the other methods. Consequently, cross-correlation is the method of choice for automatic registration of large numbers of these image pairs.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adenocarcinoma / diagnostic imaging*
  • Adenocarcinoma / pathology
  • Animals
  • Antibodies, Monoclonal
  • Carcinoembryonic Antigen / immunology
  • Colorectal Neoplasms / diagnostic imaging*
  • Colorectal Neoplasms / pathology
  • Humans
  • Image Processing, Computer-Assisted
  • Iodine Radioisotopes* / pharmacokinetics
  • Mice
  • Mice, Nude
  • Radioimmunodetection / methods*
  • Software
  • Tissue Distribution
  • Transplantation, Heterologous
  • Tumor Cells, Cultured

Substances

  • Antibodies, Monoclonal
  • Carcinoembryonic Antigen
  • Iodine Radioisotopes