 |
 |
|
|
|
|
|
|
Image analysis in Chromatography
- The Program - Features and functions - Noise region addition - Logarithmic scaling - Linear scaling - Wallis Filtering - Smoothing - Copyright - Please note - Instructions 
Very small differences in these chromatographic images can be distinguish by the eye yet chromatographic matrices are almost always treaded in one direction at the time after a reduction to selected peaks and spectra. 
The difference are easily seen in these images from analysis of simple plug extracts of two closely related Penicillium species: P. cyclopium and P. polonicum although there are several common features. In order to screen large number samples for either common or distinct features as seen in chromatograms, application of techniques from image analysis will greatly enhance the data analysis productivity. One of the core problems in comparing of chromatograms directly is aligning of the profiles to compensate for minor drift in retention times. This problem as well as other problems were solved through a close collaboration with the Image Analysis Group at Department of Mathematical Modelling, DTU, responsible for the development of the Correlation Optimized Warping (COW) technique (1). The implementation of this technique is the first and most important step toward automated sample identification and component analysis using non data reducing image analysis. COWTool The intention of the COWTool program is to allow chromatographers without extensive knowledge of programming (or e.g. Matlab) the possibility Align chromatograms without any peak detections by Correlation Optimised Warping (COW) (1) relying on a few easy selected parameters. COWTool can furthermore be used to build a sample identification system directly from raw chromatogram (2) using principle adapted from image analysis. A detailed discussion can be found in:
Note to (1): Line 7 in the pseudo-code on p. 34 should read F(N,LT)=0 not F(N, 0)=0. The error in the formula can also be seen by comparing with figure 2. page 21, 2nd column last paragraph "As it is seldom possible..." The expressions given for r2 is not optimal if LT<LP. A better solution is to use r2=r1. The discrepancy between r1 and the original expression LT-(m+)*N is automatically corrected by the COW algorithm by appropriately stretching of the sections. Using this approach, both the size of the remainders and the discrepancy that COW must correct are minimised. Note to (2): On page 732 1st column, line 6 from the top: should read r2=r1 not r2=LT-r1
A short description of the functions included in COWTool can be found below, more detailed information and user instruction can be found in the Help-File included with the program. COWTool is made easy to use and it is in compliance with the general rules of Windows for selecting files and executing functions: You simply select the files you wish to process, select the desired processing operation and its parameters, and click "OK". Alternatively, run scripts can be used with COWTool. These scripts allow an easily reproduction or to run multi-file processing sequences. It is also the preferred option for complicated multi-step processing sessions. COWTool is written and compiled using Borland C++ Builder for Windows 95 and Windows NT 4. Features and functions The following functions is included in COWTool. The functions is performed on the full chromatographic matrices collected using spectral detection e.g. HPLC with diode array detection without any pre-data treatment e.g. peak detection and integration. Input and output is single- or multi-signal chromatograms stored as plain ASCII text files. COWTool will simply read the spectral records stored as rows delimited by a space between each data point and each record separated by a <tnl> thus with chromatographic traces as columns. Single chromatogram functions
Adds a section containing only noise to one or both ends of the chromatogram, allowing e.g. the align function some degree of freedom to determine a proper end node positioning. Logarithmic scaling Performs a logarithmic transformation of the data, yielding a chromatogram with a more even peak height distribution. Used to enhance qualitative difference between chromatograms. Linear scaling Linear scaling scales the chromatogram to a specified mean and variance. Wallis Filtering Wallis filtering is used in image analysis to enhance edges (i.e. sudden transitions) in images. Can be used to accentuate, smooth, or otherwise transform peaks in the chromatogram, according to how the filter is applied. Smoothing Smoothing removes high frequency noise, and notches in chromatograms, such as those that may result from a median averaging operation. Multi-chromatogram functions
COW-aligning corrects the drift in peak retention times commonly encountered from run to run in chromatographic analysis which is vital for comparing chromatograms point to point. Aligning works only on the time axis but not the spectral axis which usually well calibrated and stable. Averaging Combines the selected chromatograms to form a single, average chromatogram. Only to be used on aligned chromatograms of exactly same length. Similarity Calculate a similarity matrix for the selected files using a selected coefficient. Only to be used on aligned chromatograms of exactly same length. Download
The COWTool program is Copyrighted© by Department of Biotechnology, Building 221 DTU, Technical University of Denmark, 2800 Lyngby, Denmark. The program is for individual use and may not be distributed further, included in software packages or other software compilations etc. Please respect this and make reference to our papers / this with any use of downloaded information. Please note All downloads from this site are made available as they are for individual non-commercial use to illustrate the concept of using image analysis to process and evaluate full chromatographic matrices as described in (1) and (2). Instructions |