Non-linear registration of histological images for 3-D middle-ear modelling

Shruti Nambiar, Mallar M. Chakravarty, W. Robert J. Funnell, D.Louis Collins


The adaptation of the non-linear warping algorithm for registration of a human middle-ear histological data set was investigated. The histological block was cut into 20-μm-thick sections and every tenth slice was stained with hamatoxylin and eosin to yield a total of 56 slices. Each slice was digitized using a slide scanner to form images of size 3644×2152 pixels. The non-linear warping algorithm used is a variant of the Automatic Nonlinear image Matching and Anatomical Labelling (ANIMAL) algorithm. The quality of the non-linear transformation can be altered by three parameters such as the similarity cost ratio, the stiffness and the weight. The application of the registration algorithm resulted in good report for some structures but poor results for other structures.


Algorithms; Data structures; Image matching; Pixels; Sensory perception; Three dimensional; Data set; Hamatoxylin; Non-linear registration; Non-linear warping algorithm

Full Text:



  • There are currently no refbacks.