TACOMA
TACOMA stands for tissue array co-occurrence matrix analysis, a tissue image scoring software. It is also named after Tacoma, a city in the greater Seattle area where the software was conceived and developed. Various parts of the software were crucially inspired by the stunning natural scenes in the region, including the two beautiful and awe inspiring national parks -- the Olympic and the Mount Rainier national park.

TACOMA is a new algorithm that automatically scores tissue array (TMA) images
in an objective, efficient, and reproducible manner despite some forbidding challenges
  • Images with the same score can be highly heterogeneous
  • Staining patterns are not localized in size, shape or position in the image
  • Labels can be fairly noisy
  • Sample size is often very small.
Attractive features of TACOMA include
  • The use of co-occurrence counting statistics to capture the spatial regularity inherent in a heterogeneous and irregular set of TMA images;
  • The incorporation of pathologists' input via informative training patches so that the algorithm can easily adapt to specific markers and cell types;
  • The ability to report salient pixels in an image that determine its score.

Overall architecture of TACOMA is shown as the following figure. TACOMA captures statistical regularity in TMA images by some spatial histogram which are used as image features. The pathologists' knowledge are encoded by a few representative image patches which are used to construct a feature mask to be applied to the image features. Feature masking as we do can be viewed as a nonparametric approach of feature selection (or dimension reduction). The filtered image features are then input to a classifier.



Citation

[1] D. Yan, P. Wang, B. S. Knudsen, M. Linden and T. W. Randolph. Statistical methods for tissue array images--algorithmic scoring and co-training. Annals of Applied Statistics, Vol 6(3), 1280-1305, 2012.  arXiv:1102.0059