AT500963B1 - METHOD FOR ANALYZING BAND PICTURES - Google Patents

Description

2 AT 500 963 B1

The invention relates to a method according to the preamble of claim 1. The method according to the invention can be used in particular for the evaluation of samples in anti-doping test laboratories. The invention relates above all to a method for the correction of non-homogeneous background in digital EPO (erythropoietin) images and for the calculation of a cut-off line " (Reference line), which is used as a doping-positivity criterion. The underlying digital gel images are obtained in particular with the aid of isoelectric electro-focusing (electrophoresis) and created with image recording and / or image processing equipment. The annealed rectangular band structures are located in preferably perpendicular tracks of the image and are predominantly horizontal (a typical example of a protein gel image is shown in Figure 1). The picture illustrates the accumulated deposits of isoelectric electro-focusing glycoprotein forms and "double blotting". 15

Peptide hormones such as EPO are widely used in endurance sports such as cross-country skiing or cycling because of their performance enhancing effect and heavy detectability. The main goal of rEPO (recombinant EPO) doping is to increase red blood cell production, which results in increased oxygen transport capacity of the blood. It can achieve a performance increase of up to 10%. While human EPO - a glycoprotein - is produced in the kidneys, rEPO is e.g. produced in ovarian cells of Chinese hamsters.

Conventional doping substances such as anabolic steroids can be detected by gas chromatography 25 (GC) and mass spectrometry (MS). However, with these techniques, it is not possible to detect the application of rEPO since the molecular weight of glycoproteins is in the range well beyond the range of GC-MS. It was found that for the detection of rEPO isoelectric focusing (IEF) in gels and the " double blotting " Method, which transfers the proteins contained in gels in a more adequate medium 30, are suitable. By applying chemiluminescent reaction a final EPO image is generated (Fig.1). The purpose of the EPO Doping Control is to define and calculate a quantitative characteristic based on comparing the positions of reference standard REPO bands and the bands in the tracks of urine samples from athletes in EPO digital gel images. 35

In Fig. 2, a conventional method for calculating the reference " cut-off line " outlined. It is based on the use of standard graphics tools and a manual analysis. On the right side a profile of the mean line intensities of a rEPO standard track is displayed. The user searches for the local maximum P ,, corresponding to the 40 worst-case band of the rEPO standard track. Thereafter, the valley V, between P, and the next local maximum in the basic direction is searched and the distance P, V, = δ is measured. Finally, the position of the cut-off line is defined as the position defined by the distance δ from Pf in the opposite direction to the valley V? given is. It is above all an object of the invention to be able to achieve an exact creation of a cut-off line with little computational effort and to be able to unambiguously separate bands of different substances.

These objects are achieved in a method of the type mentioned above with the features indicated in the characterizing part of claim 1.

The accuracy of the procedure is increased or the computing time is reduced when proceeding according to the features of claim 4. In order to bring the track images in a defined starting position for the evaluation, the features of claim 6 55 are advantageous. To delineate Spurbildem differentre substances is advantageous according to the 3 AT 500 963 B1

Characteristics of claim 7 proceeded.

In the following the invention will be explained in more detail with reference to the drawings: FIG. 3 shows schematically the evaluation of a track image. 4 shows schematically the formation of the intensity variability and the absolute differences. 5 shows schematically the separation of bands contained in a track image of two different substances. 6 shows schematically a flowchart of the method. The following section explains how to calculate a cut-off line for a rEPO track image.

A single intensity value Pmed (i) is determined for each line of the track image present in digital form. In order to suppress image noise, it is advantageous to calculate a 1D median profile (column vector) Pmed (i) of the given standard 2D digital lane image L (ij) for all lines i-1,2,..., M using the following formula:

Pmed (i) = median {L (i, j): j = 1,2, ..., n}, where j represents the column indices. This means that the median of a given line / n intensity values of this line of the track image is found. Instead of a median formation, another averaging or weighting of the intensity values of the pixel of a row could be undertaken. 25

In order to reduce irrelevant local maxima of the median profile, it can be smoothed by a mean filter. In convolution notation, the filtered median profile Pai (i) is passed through

Pwi (i) = Pmeö (i) ff (i) 30, where f (i) represents the convolution kernel of the mean filter.

In the following, a set {(Ak Bk)} of all possible partitions of the track image L (i, j) is constructed into adjacent rectangular blocks Afe Bto k = 1,2, ..... r (FIG. 3) 35

Each image block Ak is characterized in continuous notation (for real variable x instead of integer indices /) by an overall measure m / A *) of the intensity variability within the block: 40 m, (Ak) = 1 a *

dPm (χ) dx | dx

Each image block Bk is characterized in continuous notation by a total measure mb (B ^ of the intensity variability within the block: mb (Bk) ~

dPw i (x) dx \ dx 50

As shown in FIG. 4, it is possible that, based on the vector of the intensity values Pfmfi) for the respective adjacent blocks, the respective integral intensity variability m / A *), mb (Bd, in particular for the upper block, the intensity variability m / A * ) and for the lower block the intensity variability mb (B *) is determined. In particular, for all these blocks, the sum of the absolute values of the differences of the intensity values Pwi0) of each of two consecutive lines can be determined over each line. For all possible positions k of the dividing line between two adjacent blocks Ak, Bk two functions are defined: namely a function topdown given by the values of the intensity variable m ^ A *) of the blocks Ah Ah ..., Ar, and furthermore, a function bottomup given by the values of the intensity variability mb (BiJ of the blocks Bh Sa Br.

It is advantageous if the integral intensity variability determined for the respective blocks is normalized, preferably by dividing the sum obtained by the number of rows of the respective block which may have been reduced by 1.

Finally, for the discrete version of these two functions, a difference function is defined as follows:

Dif (k) = \ Qi (k) -Q2 (k) \.

For this purpose, it may be provided that, in order, the quotients Qi (k) determined for the blocks located closer to the one track boundary, in particular the upper blocks, are determined from the blocks adjacent to the other track boundary, in particular the lower blocks Quotient Q2 (k) subtracted and the absolute values of the differences are formed.

A search is made for the global maximum of this function Dif (k). The 25 line index of this maximum represents a pair of adjacent image blocks for which the difference in their intensity variability is greatest. This row index is considered the row index of the cut-off line.

The determination of the difference in the intensity variability is illustrated by means of a number example with reference to FIG. The profile Pai (i), which is shown in Fig. 4 above, is determined for the individual lines by averaging the intensities of the digital pixels, in particular, a median formation is made. For the individual lines, the filtered median profile Pfm (i) has, for example, the numbers given in the form of a vector. 35 This column vector is divided into the blocks corresponding to the blocks Ak and β * of the respective track image (L (i, j)); a division is shown in which block Ak has four rows (block A3) and the associated block Bk has 16 rows (block β3).

In the next step, the absolute values of the differences of the filtered median values of the individual lines are formed for the individual blocks.

There is a summation of the absolute values of the differences for the two blocks A3 and B3 and a quotient formation, in which the sum determined for the individual blocks is divided by the number of lines reduced by 1. The quotients Qi (k) and Q2 (k) 45 or Qi (3) and Q2 (3) are obtained.

In consequence, the absolute difference Dif (3) of these two quotients is determined; this determination of the quotients takes place for each partition of the column vector Pnn (i), so that a difference vector Dif (k) is obtained. The maximum of this difference vector determines the position of the cut-off line. 50

The general scheme of calculating the intensity variability within the pairs of adjacent image blocks of a track image L (ij) is advantageously modified since not all blocks are equally relevant. Part of the blocks is expediently excluded due to insufficient information (too few number of lines) or due to the advance knowledge of the band width in EPO images. The average band width Bwav (empirical value 5 AT 500 963 B1 in number of pixels or lines) at the level of a predetermined or calculated average threshold value Inav is determined therefor. In the filtered median profile PmO), the first local maximum (imaxref) from above or its row index is determined, which is greater than the threshold value lnav. The first relevant image block A is then defined as the following sub-image of the 5-track image:

Ai (i, j) = {L (i, j): i = 1,2, ..., (imaxrel-1ABwav) lj = 1,2, ..., n}

In a similar procedure, the last image block Br is defined using information about the average band width in EPO images as follows:

Bt (i, j) = {L (i, j): / = m - Bwav, ..., m, j = 1,2, ... n}.

It is thus provided that for determining the inner boundary of the first block Ai with 15 of the lowest number of lines, the maximum of the course of the intensity values Pmi (i, j) is determined, which is the smallest distance from one of the two, in particular the upper, track boundary Spt and a predetermined threshold Inav, in particular the average or. Mean value of the intensity values PmedÖ,) exceeds. Further, from the line index (imaxrel) of this maximum, a predetermined value, in particular half the line number of the average bandwidth Bwav, is subtracted and the resulting line index is regarded as the inner boundary of the first block Ai. In order to determine the inner boundary of the last block Ar having the highest number of lines r, the existing track boundary is shifted toward the inside of the track image by a line number corresponding to the average bandwidth Bwm and is regarded as the inner boundary of the last block Ar. 25

These determined limits for the blocks with minimum number of lines are based on the pitch of the track image L (i, j). The number of lines of the blocks B conjugated to the blocks Ar results because the sum of the lines of the blocks Ar and the blocks Br corresponds to the total number of lines of a track image having the boundaries Spt and Spb. 30

The evaluation of a standard NESP track in which the bands are at the bottom, i. In contrast to the evaluation of a rEPO standard track, two additional steps are used to concentrate the area of a rEPO image opposite to the area of a rEPO image: First, the track image must be aligned over one line or horizontal axis before applying a rEPO evaluation method be mirrored. This gives a temporary track image in which the bands are concentrated at the top, which is characteristic and common for EPO standard tracks. Subsequently, the procedure described above is applied to this geometrically transformed track image. Finally, the resulting line index of the cut-off line is recalculated into the original coordinate system of the input image. In principle, the method according to the invention is equally applicable to rEPO and NESP track images, e.g. could a NESP track image in its image plane rotated by 180 ° based on the evaluation.

If there is a combined rEPO + NESP standard track in a track image, calculate two Cut-45 off lines; a rEPO cut-off line for the part of the track where rEPO proteins are concentrated, and a NESP cut-off line for the part where NESP proteins are concentrated.

In the presence of track images containing band regions of different substances or substances, in particular rEPO and NESP, an intensity value Pmed (i) is determined for separating the band regions of the different substances or substances for each line of the track image, e.g. by averaging, weighted message, summation or median formation of the intensity values of the individual digital pixels of this line. For a combined rEPO + NESP 2D track image L (i, j), a 1D median profile Pmed (i) is thus first calculated, as is the case with the procedure described above.

Claims (9)

6 AT 500 963 B1 The median profile Pmed (i) can then be filtered with a mean filter with a large window size win (about 10% of the track image height, in conventional EPO track images about 30 pixels). It is advantageous if, for filtering the profile Pmed (0) of the intensity values of the individual lines, averaging of the intensity values over a 6 to 14%, preferably 8 to 12%, of the 5 line number of the entire track image amounts to, in particular symmetrically, on both sides the row in question is displayed and this value is assigned to the line under consideration as an intensity value. The resulting filtered median profile Ptn2 (i) is searched for the global maximum / Waxt io and the second highest maximum Max2. Between Max «and Max2 the system searches for the global minimum min. At the line index (coordinate) d of this minimum, the input track image is divided into two temporary fields Lt (i, j) and Lb (i, j) (Figure 5). The filtered partial median profiles PTai (i) (for upper field) and PBm (i) (for lower field) corresponding to these two fields are subsets of the original filtered median profile Pmi (i). Further, the average intensity value (threshold value) Inav of the entire track image is calculated. The partial filtered median profile ΡΤαι (ΐ) searches for the first local maximum from below, the value of which is higher than the threshold value lnav. The index of this maximum Tnsp serves as upper bound on the NESP field Lnsp (i, j), which is defined as follows: LnspfiJ) = {L (i, j): ie {Tnsp + Ά Bwav, ..., m}} , The partial filtered median profile PBai (i) is searched for the first local maximum from above 25 whose value is higher than the threshold lnm. The index of this maximum Bepo serves as the lower bound of the rEPO field Lepo (i, j), which is defined as follows: Lepo (i, j) = {L (U): i e {1. Bepo-'Λ Bwav}}. 30 The rEPO field Lapo (ij) uses the procedure previously described for a rEPO image with the aim of determining the rEPO cut-off line for the combined track image. The NESP sub-image L "sp (i, j) is applied to the procedure previously described for a NESP image with the aim of determining the cut-off line for the NESP sub-image. This cut-off line 35 is recalculated into the original coordinate system of the combined track image and thus gives the final NESP cut-off line of the combined track image. In general, it is useful to apply a noise filter, preferably a geometrically controlled diffusion (GDD) based noise filter, prior to determining the cut-off line for noise removal. It is also useful if, prior to the determination of the cut-off line, image correction, e.g. by equalizing, shifting and / or resizing, the intensity image or the image matrix is made. 5 shows schematically in the form of a flowchart the procedure for the evaluation 45 of rEPO-track images, NESP-track images and combined rEPO + NESP-trace images. 1. Method for analyzing and / or evaluating band images of gels, especially organic gels, of at least one (m), preferably organic, substance and / or substance, preferably natural and / or synthetic proteins, in particular EPO and / or or rEPO and / or NESP, which band images are obtained by electrophoresis, preferably isoelectric electrofocusing, and evaluated with image recording and image processing equipment, wherein the bands of the (s) respective substance or substance are used to delineate the band In a track image containing a track, a cut-off line or boundary line is determined and / or calculated and then the evaluation of the track image or the band images is continued, characterized in that for the creation of the cut-off line for each line (FIG. i-1, 2, ..., m) of the digital track image (L (iJ)), an intensity value (Pme <i (i)) is calculated, eg by averaging, weighted averaging, summation or median formation of the intensity values of the individual digital pixels in the columns {j = 1, 2, ..., n) of this row, - that each track image (L (i, j)) within its optionally selected or predetermined track boundaries (Spb Spb) is divided by a parallel to a line division into two immediately consecutive, extending in the column direction blocks (Ak, Bk), where all possible partitions (k = 1, 2, ..., ή of the divided track image L (i, j) into two adjacent blocks (Ak, Bk), in particular into an upper block (Ak) and a lower block (£ *), 15 - that based on the vector of the intensity values (PmedO)) for the adjacent blocks, the respective integral intensity variability (m / A *), mb (Bk)), in particular for the upper block, the intensity variability (m ^ Ak)) and preferably for the lower block the intensity variability 0mb (B <J), it is determined, in particular for these blocks, over all lines the sum of the absolute values of the differences of the intensity values (PmedO)) of every 20 two consecutive lines is determined, - that normalizes the determined for the respective blocks integral intensity variability is, preferably by dividing the sum obtained by the possibly reduced by 1 row number of the respective block, - that in turn or for all partitions for the one lane boundary closer laid blocks 25, in particular the upper blocks, determined quotients Qi (k) are subtracted from the quotients Q2 (k) determined for the adjacent neighboring blocks, in particular the lower blocks, in particular, that among all the resulting absolute difference values (Dif (k)) the global maximum ( M) is determined, and that the line having this maximum (M) is of division is considered as a cut-off line or boundary line of the band area of the (s) to be evaluated in this band track substance or substance. 2. The method according to claim 1, characterized in that after determining the intensity values and before dividing the track image into blocks, the vector of the intensity values Pmed0 determined for the individual successive lines is subjected to a convolution, in particular that for the individual lines optionally weighted average is determined by the intensity values of a predetermined number of successive, in particular symmetrical to this line located rows, weighted averaged 40 and the obtained value Pm0) is set instead of the intensity value of this line and the further evaluation is based. 3. The method according to claim 1 or 2, characterized in that are used for calculating the intensity variability or the sum only blocks with a predetermined Mindestzeilen- 45 number. 4. The method according to any one of claims 1 to 3, characterized in that - to determine the inner boundary of the first block (Ar) of the partition with the lowest number of lines, the maximum of the profile or profile (PmO)) of the given 50 if convoluted intensity values which has the smallest distance from one of the two outer lane boundaries (Spt, Spb), in particular the upper lane boundary (Spt), and a predetermined threshold value (Inav), in particular the average or mean value of the intensity values (PmedO). ) &lt; exceeds, - that from the line index f / maxre / j of this maximum, a predetermined value, in particular half of the mean width {Bwav) of all maxima on the level of 55 8 AT 500 963 B1 threshold (lnav) corresponding number of lines is subtracted, and the resulting line index determines the inner boundary of the first block (A,), that for determining the inner boundary of the last block (Ar) with the highest number of lines, the line index of the given, in particular lower, 5 track limit (Spb) a row number corresponding to the mean width (Bwav) is reduced or the line of the track boundary (Spb) is shifted toward the inside of the track image, and the obtained row is regarded as the inner boundary of the last block (Ar), and that these determined limits to determine the minimum number of lines of the division of the track image (/.(7,/)) based blocks are used. 10 A method according to any one of claims 1 to 4, characterized in that the digital image is determined before determining the intensity values of an image correction, e.g. by equalizing, shifting and / or resizing the image matrix, if necessary accompanied by noise suppression. 15 6. The method according to claim 1, wherein prior to determining the intensity values, the track image, in particular a NESP track image, is mirrored geometrically inverted or parallel to the line progression, whereupon the cut-off line according to claim 1 to 5 and for further evaluation, the cut-off line in the same way as the mirrored NESP track image is inverted back to the starting position or reflected back. 7. The method according to any one of claims 1 to 6, characterized in that - in the presence of track images, the band areas of different substances 25 or substances, in particular of rEPO and NESP, contain, for the separation of Bandenbe rich of different substances or substances for each line of the track image, an intensity value (PmedO)) is determined, eg by averaging, weighted averaging, summation or median formation of the intensity values of the individual digital pixels of this row,. The profile of the intensity values (Pmed ('')) of the lines of the track image is subjected to smoothing, filtering or convolution and a profile (PmO)) is determined and in this profile (P02Ü)) the global maximum (Max, ) and the next highest maximum (Max2) and the global minimum (min) lying between the two maxima, - that the track image in relation to the line (d) containing the minimum (min) is divided into two partial images (Lt (iJ), Lb (i, j)), - that the average intensity value (lnav) of the profile (PmedO)) of the intensity values is determined, - that for the two partial images the partial filtered profiles or median profiles (PTm (i ), PBfjn (i)) of the respective filtered profile or median profile (PmO)), 40 - that in the one, in particular upper, partial image (Lt (i, j)), the outer track boundary (Spb) removed local maximum (Tnsp) is determined whose value exceeds the threshold value (lnav) - the line index of this maximum is increased by the number of lines corresponding to half the average bandwidth (Bwav), and the resulting line index as an inner, especially upper, track boundary of the other, in particular lower, field (Lnsp0> i) i, in particular NESP -Spurbildes is considered, - that in the other, in particular lower field Lb0, j)) the maximum of the outer track boundary (Spb) lying maximum (Bep0) is determined, whose value exceeds the threshold value (Inav), then - that of line index this maximum subtracts the number of lines corresponding to half the average bandwidth (Bwav) and the resulting line index is regarded as inner, in particular lower, track boundary of the first field (Lepo (i, j)), and - that the two fields (Lepo0, j), LnspOJ)) separated as separately present and separately evaluated track images each of a single substance or a single substance 55 according to the characteristics of the claims 1 to 6 starting with the division 9 AT 500 963 B1 of the track image are evaluated in blocks. 8. The method according to any one of claims 1 to 7, characterized in that for filtering the profile (PmedO)) of the intensity values of the individual lines to separate the Bandenbe- rich 5 of the different substances averaging the intensity values over a 6 to 14%, preferably 8 to 12%, the number of lines of the entire track image amount of, in particular symmetrical, located on both sides of each line considered lines and this value of the line considered as the intensity value of the profile (PmO)) is assigned. 10 9. Use of the method according to any one of claims 1 to 9 for the evaluation of Ban-denbildem in view of the presence or existence of replaceable for human doping preparations, such. rEPO and / or NESP. 15 For this 4 sheets of drawings 20 25 30 35 40 45 50 55