The Cancer Genome Atlas (TCGA) has been one of the clearing houses of genome-wide array data for the understanding of the molecular basis of cancer from large cohorts. These analyses are intrinsically from bulk measurements of mixed cell types, derived from frozen biopsy sections that include tissues with mixed histopathology and/or microanatomies (e.g., tumor, stroma). While bulk array profiling may provide insights into molecular aberrations, it provides only an average genome-wide measurement for a biopsy and fails to reveal inherent cellular composition and heterogeneity of a tumor. On the other hand, histology sections do not provide standardized measurements, but they are rich in content and continue to be the gold standard for the assessment of tissue neoplasm.
We are developing a platform to facilitate management and analysis of data provided by the NCI’s TCGA project. The significance of this platform is its robustness and scalability on data processing, and the potential results of this initiative are: (I) An efficient and effective platform for the representation and characterization of tumor histology as well as the integrative analysis with clinical outcome; and (II) An atlas that identifies morphometric subtypes, responses to therapies, and molecular correlates. Therefore, any clinical sample can be crossreferenced against such an atlas for precision medicine and personalized therapy.
Visualization of WSIs as well as the computed nuclear architectures are available at Berkeley Cancer Morphometric Data Portal
Related Publications (Selected)
- Hang Chang*, Yin Zhou*, A Borowsky, K Barner, Paul Spellman and Bahram Parvin. “Stacked Predictive Sparse Decomposition for Classification of Histology Sections.”International Journal of Computer Vision (Special Issues on Deep Learning), (2015): 113(1) 3-18.(*Co-First Authors)
- Hang Chang, Ju Han, Alexander Borowsky, Leandro Loss, Jow W. Gray, Paul T. Spellman and Bahram Parvin. “Invariant Delineation of Nuclear Architecture in Glioblastoma Multiforme for Clinical and Molecular Association.” IEEE Trans. on Medical Imaging, 32 4 (2013): 670-682.
- Hang Chang, Ju Han, Paul T. Spellman and Bahram Parvin. “Multi-Reference Level Set for Characterization of Nuclear Morphology in Glioblastoma Multiforme.” IEEE Trans. on Biomedical Engineering, 59 12 (2012): 3460-3467.
- H. Chang, G. V. Fontenay, J. Han, G. Cong, F. L. Baehner, J. W. Gray, P. T. Spellman and B. Parvin. “Morphometic Analysis of TCGA Glioblastoma Multiforme.”BMC Bioinformatics, 12 484 (2011).
- Yin Zhou, Hang Chang, Kenneth E. Barner and Bahram Parvin. “Nuclei Segmentation via Sparsity Constrained Convoluational Regression.” IEEE International Symposium on Biomedical Imaging (ISBI 2015), Brooklyn, NY, U.S., April 2015.
- Hang Chang and Bahram Parvin. “Predictive Sparse Morphometric Context for Classification of Histology Sections.” IEEE International Symposium on Biomedical Imaging (ISBI 2015), Brooklyn, NY, U.S., April 2015.
- Yin Zhou*, Hang Chang*, Kenneth Barner, Paul Spellman, and Bahram Parvin. “Classification of Histology Sections via Multispectral Convolutional Sparse Coding.” IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2014), Columbus, Ohio, U.S, June 2014. (*Co-First Authors)
- Hang Chang*, Yin Zhou*, Paul Spellman, and Bahram Parvin. “Stacked Predictive Sparse Coding for Classification of Distinct Regions in Tumor Histopathology.” International Conference on Computer Vision (ICCV 2013), Sydney, Australia, December 2013.(*Co-First Authors)
- Hang Chang, Nandita Nayak, Paul Spellman, and Bahram Parvin. “Characterization of Tissue histopathology via Predictive Sparse Decomposition and Spatial Pyramid Matching.” International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI 2013), Nagoya, Japan, September 2013.
- Hang Chang, Alexander Borowsky, Paul Spellman, and Bahram Parvin. “Classification of Tumor Histology via Morphometric Context.” IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2013), Portland, U.S, June 2013.
All resource, including data and source code related to this project, have been released on BMIHub for public consumption.