Nearest-neighbor analysis of spatial point patterns: application to biomedical image interpretation

Analysis of the spatial distributions of objects is fundamental to biomedical image interpretation. Nearest-neighbor (NN) methods are generally used to assess whether objects are arranged at random or in a deterministic manner. Simple standard NN techniques, however, may fail to identify complex spatial organizations. To overcome this problem the present study proposes a NN iterative algorithm that enables deterministic spatial patterns to be detected by identifying the distances between objects for which there is the greatest deviation from randomness and hence the amplitude of the areas of maximum reciprocal influence between objects. The performance of the algorithm is evaluated by applying it to both manufactured and experimental data. The manufactured date example showed that the proposed procedure produced neither false positives or negatives. The method proved to be extremely sensitive, detecting even small deviations from randomness. The experimental analysis was applied to the study of the spatial distribution of apopototic structures in malignant neoplastic tissue. It showed that the apopototic cells and bodies are characterized by a complex spatial pattern, and aggregate closely.