Image Mining for Analysis of Expression Patterns in Tissue Microarray - Center for Biomedical Imaging & Informatics Skip to main content

The capacity to distinguish among subclasses of disease affects how patients are treated, which medications are appropriate, and what levels of risk are justified. Tissue microarray (TMA) technology makes it possible to investigate and confirm clinico-pathologic correlations which have been postulated based upon the evaluation of whole histology sections. Unfortunately, inconsistencies often arise during the evaluation process as a result of subjective impressions and inter- and intra-observer variability. Advanced imaging and computational tools make it possible to detect and track subtle changes in measurable parameters leading to insight regarding the underlying mechanisms of disease progression and the discovery of novel diagnostic and prognostic clues which are not apparent by human inspection alone.

The overarching goals of this renewal application are to build upon progress made in the first phase of the project and design, develop and evaluate new capabilities by meeting the objectives of the following specific aims:

(1) Develop and evaluate a new family of multi-stage, searching algorithms to facilitate quick, reliable interrogation of larg-scale, clinical and research, microscopy applications including whole-slide imaging and tissue microarray;

(2) Develop and evaluate a suite of high-throughput services capable of automatically detecting, archiving and indexing user-specified objects (e.g. tissues, cells) in large collections of images and implement extensions to the data models and support for optimized pipeline selection. These capabilities will enable large-scale correlative outcomes studies and support expansion of the “gold standard” image archives and correlated clinical repositories. The services will take advantage of state-of-the-art parallel CPU-GPU machines and the searching algorithms described in Aim 1;

(3) Optimize the imaging, computational and content-based image retrieval algorithms and tools using a wide range of different tissues, cancer types and biomarkers to support clinical and research experiments and studies involving patient stratification, quality-control, and outcomes assessment; and

(4) Deploy the analytical tools, data models, user-centered interfaces and reference libraries of imaged specimens to participating adopter sites to conduct open-set usability and performance studies and make these resources available to the clinical and research communities as open source software and resources to support future development and testing of new hypotheses, algorithms and methods.


Source Code

Please register to download the source code for “cell segmentation algorithm”.   You can also contact us directly to receive the Matlab source code.


Please see example retrieval results.