University of Utah              The JOHN A. MORAN EYE CENTER

HOME     TOOLS     SCIENCE     PEOPLE     PUBLICATIONS     METABOLOMICS     IMAGERY

CYTOSOMICS, METABOLOMICS and COMPUTATIONAL MOLECULAR PHENOTYPING

Micromolecules: definitions & links	Cellular domains: Genome, preoteome & metabolome
Metabolic diversity: Scale, dynamics, & phyletics	Phenotyping strategies: proteomics vs metabolomics
CMP Platforms: Platforms and workflow overview	CMP Probes: The probe library
CMP Substrates: Molecular trapping & detection	CMP Datasets: Data arrays for multichannel imaging
CMP Analysis: pattern recognition theory and tools	CMP Exploration: N-space visualization tools
CMP Annotation: browsing & annotating data

IMAGE REGISTRATION TEST SETS

Raw, unregistered 8-bit tiff series

Details: Serial 250 nm resin tissue sections. Image capture resolution = 476 nm/pixel. Post-cap operations: INVERT Images with B, TT and E signals are easiest to align. G is the most difficult in this set Capture Series (Win: right-click > save as; Mac: Ctrl-click, save as) 2366_D01.tif 2366_G01.tif 2366_B01.tif skipped 2366_E01.tif 2366_Q01.tif 2366_B02.tif 2366_TT01.tif 2366_E02.tif 2366_YY01.tif 2366_B03.tif 2366_G02.tif

Registered 8-bit tiff series

2366REG_D01.tif 2366REG_G01.tif 2366REG_B01.tif skipped 2366REG_E01.tif 2366REG_Q01.tif 2366REG_B02.tif 2366REG_TT01.tif 2366REG_E02.tif 2366REG_YY01.tif 2366REG_B03.tif 2366REG_G02.tif

IMAGE CLASSIFICATION

ARCHIVED DATA SETS

IMAGE CLASSIFICATION JEWEL SETS

Marc_2000_jewels_512_R.tif Marc_2000_jewels_512_G.tif Marc_2000_jewels_512_B.tif Marc_2000_jewels_512_RGB.tif

IMAGE CLASSIFICATION WITH MORPHIAS 1 BETA

Comparison of Morphias 1.1 and PCI Geomatica® Files: full sized files 1350 x 964, 8-bits grey, background-masked, contrast-stretched 60-220 2366invmask_y_60220.tif 2366invmask_g1_60220.tif 2366invmask_e2_60220.tif 2366invmask_d2_60220.tif 2366invmask_b2_60220.tif 2366invmask_q_60220.tif 2366_tmask.tif

Classification parameters PCI: Isodata, Clus(max 20, est 16, min 12), convergence 0.01, sample 20%, seed diagonal Morphias: Kmeans, 12 clus, sample 10%, seed random Theme Maps harmonized in GraphicConverter 5.6.2X for Mac OS 10.4 www.lemkesoft.com

↑ PCI = 11 classes Full size indexed tiff: Marc_2366_original_PCI_map.tif ↑ Morphias = 9 classes Full size indexed tiff: Marc_2366_morphias_map.tif ↑ Morphias-PCI: The class match is excellent overall and argues that Morphias is already a good substitute for PCI. This is a difference map between Morphias and PCI (after median filtering to correct for some kerf errors). The purple class is a separate PCI class contained mostly within the orange class in Morphias. The blue class is a separate PCI class is contained mostly within the lite green class in Morphias. We compared some (not all) signals in CellKit as 2D histograms for these classes with the conclusion that PCI pushed the envelope a bit, while Morphias was conservative. Of course we restricted clusters to 12 and haven't given Morphias a fair test yet. We also haven't used mindis to assess separations. Maybe by the end of next week. Full size indexed tiff: Marc_2366_morphias-PCI_map.tif

2D histogram of AGB (x axis red channel) vs GABA (y axis green channel) signals for the Morphias lite green class 2D histogram of AGB (x axis red channel) vs GABA (y axis green channel) signals for the PCI lite green and blue classes with Morphias envelope. It seems that PCI was able to pick up the trough in signals between the modes 2D histogram of AGB (x axis red channel) vs GABA (y axis green channel) signals for the Morphias orange class 2D histogram of AGB (x axis red channel) vs GABA (y axis green channel) signals for the Morphias orange and purple classes with Morphias envelope. It seems that the Morphias outcome lumping all together is more defensible, although multiple modes are evident in the cluster.

HOME     TOOLS     SCIENCE     PEOPLE     PUBLICATIONS     METABOLOMICS     IMAGERY