A “Tale of Two Cities” – new version, neuroscience – style.

4. Quantitative staining control, image processing and cortical profiles measurements

Quantitative staining control procedure (first paper, page 558-559) was published for the first time in 1985, and was based on original procedure of mean gradient distribution measurement. This All steps of image processing, including sections “Scanning procedure”, “Segmentation”, “Autofocusing”, “Elimination of artifacts and blood vessels”, “Separation of overlapping cells”, described in the first paper (pages 559-561), were developed in the Laboratory of Clinical Neuroanatomy of the Institute of Psychiatry in Moscow approximately 10 years before the discussed publication became available. For description of our technology an interested reader can look at this article: “The story of MCG: lost opportunity“.

As it was stated in the discussed paper, the hardware used was texture analysis system TAS of Ernst Leitz, Wetzlar. GmbH. However, it is important to notice, that this system was custom-made, or modified for our laboratory in Moscow, and had no analogues anywhere in the world. Consequently, any developed software could be used only with our system. All measurements and image processing results using described methods could be obtained  only in our Moscow laboratory and nowhere else. Moreover, the Institute in Dusseldorf never had TAS system (also, see section 6 below), so there were no physical possibility to develop any software for this device in Dusseldorf. Even though A. Schleicher and K. Zilles could not have anything to do with it, published description of the technology with A. Schleicher and K. Zilles as “co-authors” implies their participation, or even supervision of the development of the technology, what was completely not true.

5. Re-use of pictures and of scientific data

Measurement_Steps_from_1995_fig2

 

The picture on top, above, published as Fig. 2, p. 560, is a slightly modified image, originally published in 1992 [7], as Fig.2, page 19 (bottom, above). K. Amunts contributed to manuscript preparation for this publication, and was listed as co-author. However, K. Zilles or A. Schleicher did not contribute to this work or its publication.

Equation1_from_1995_and_1985
Mean Gradient definition

The definition of the mean gradient distribution and the equation (1), page 559,  (left) developed to calculate the value of mean gradient, and explanation of its application for segmentation of human brain cortex were originally published in Russia, in “Korsakov’s Journal of Neuropathology and Psychiatry”, 1985 [10], page 1028 (right). K. Amunts, K. Zilles or A. Schleicher did not contribute to this work or its publication. The initial source of the algorithm is my dissertation.

Distributions published as Fig. 3, comparison of two mean gradient distributions,  and fig. 5, mean gradient (1 and 2, left), are slightly modified pictures, originally published as Fig. 4 (a), page 1030, and fig. 2g, page 1027 in publication [10] (3,4 right). Same here: K. Amunts, K.Zilles or A.Schleicher did not contribute to this publication.

Fig. 4, above left, with no substantial modifications, was taken from Fig.3,b (above right) originally published in [8], page 146. This paper was prepared and published before K. Amunts started working under supervision of K. Zilles. K.Zilles or A.Schleicher did not contribute to this paper or its publication. The initial source of the mean gradient algorithm is my dissertation ([13], page 63), see below.

Dissertation_page_63
Explanation of the rationale for application of the mean gradient distribution for automatic segmentation. From [13], 1987, Moscow, USSR

Fig_8_1995_and_fig_1_1992

Fig.8, page 562 (left) is practically the exact copy of a figure, which was originally published as fig. 1 in 1992, [7], page 18 (right). This paper was also prepared and submitted for publication before K. Amunts started working under supervision of K. Zilles. K.Zilles or A.Schleicher did not contribute to this paper or its publication.

6. How to throw out a baby with the water without even noticing it?

In this paper, without proper attribution, co-authors report result of original technology, which I gave a name “Automatic Morphocorticography”, or MCG.

Fig.10, a-c illustrate typical examples of automatic morphocorticography results presentation (left). One should notice (a) vertical bars, dividing characteristic patterns of the profile according to layers of the cortical plate; (b) small vertical bars that indicate standard error of mean; (c) characteristic set of three reported parameters: Va, Na, Sa. This style of cortical profile presentation was originally presented before (right), for example, in [11] (1987, fig 2, a-c, page 346), and [12] (1988, fig.2, a-c, p.963). K.Amunts, K.Zilles or A.Schleicher did not contribute to these papers or their publication.

Moreover, on page 563, left column, bottom paragraph, we find: “In addition, local maxima were found on each profile curve (Figs. 7-9, 10A-C). Their number was, on the average, 16 (SE 0.3). This number was larger than the number of cortical layers (and sublayers) in area 4 as described in
classical cytoarchitectonic reports.” And further: “The confidence interval was narrow (Fig. 10A-C), indicating significance for most of these maxima. Interestingly enough, this constant finding of local maxima was achieved despite the large interval of averaging along the cortical surface (approximately 10 mm).” These two statements summarize my original finding, published before only in my dissertation. These findings constitute essential result, which allowed to give the special name to this technology “MCG, or automatic Morphocorticography”. Unfortunately, in this publication, Amunts, Zilles and Schleicher not only did not give a reference to the source of this finding. They also demonstrated complete lack of understanding with respect to what technological principles were developed in order to achieve this result. They write: “constant finding of local local maxima was achieved despite the large interval of averaging “. It’s a typical example of throwing away the baby with the water: the main technological requirement for finding stable and reproducible maxima of the cortical profile is ergodicity [17] of the measurement interval. In other words, “large” interval of averaging was not an obstacle, but an important requirement  of getting stable and reproducible profile.

In my work ([13], page 92) we can find: “Thus, as a result of application of developed automated technology we obtained a set of characteristic profiles , which quantitatively describe the individuality of cytoarchitectonic of a particular area”. And more: “We feel that it is proper to call such a set of stable and reproducible profiles describing the most general and reproducible features of laminar cortical structure a “morphocorticogram”, and the method itself – “Morphocorticography” (translated from Russian). As I said already, curious reader might look at more or less comprehensive description of this technology in my earlier post (The story of MCG: lost opportunity).

As a recent PhD graduate (which was the case in 1995), who used my programs and algorithms for all imaging data collection and analysis, Katrin should remember this principle as a matter of course, due to constant discussions of this idea. Satisfaction of the measurement interval to ergodicity condition was one of cornerstone of developed technology. By the way, ergodicity requirement was a novelty in 1980s and  1990s, but today it is very well accepted. It even became a requirement of stereological measurement protocol ([18]), named “consistent estimator”. Of course, 30 to 80 mm. sq. of averaged sample required assurance that all selected measurement fragments were localized within boundaries of a particular cortical area.

So,  reporting of my original findings was abbreviated and vague,  and also published without proper attribution and reference. Regrettably, it also failed to explain the major condition of how these findings were achieved. Not only my co-authors threw a baby out with the bathwater, they clearly had no idea that the baby was there!

7. Contribution of the co-authors

I personally spent at least  six years (from 1980 to 1986) developing described technology in Moscow Institute of Clinical Psychiatry. Working in both scientific and engineering capacity, I developed all image processing, focusing, segmentation, image measurement and data analysis algorithms, as well as software tools for planing and verification of all measurement and analysis procedures. It was truly pioneering work, which became the core of my dissertation. Of course, I knew papers published by H. Haug, as well as by K.Zilles and A. Schleicher in Germany between 1976 and 1985. However, their approaches, mainly due to hardware problems, were significantly less powerful and much less sophisticated (please, read my analysis of GLI method here).

As I explained before, due to political situation in the Soviet Union and Europe, for all these years any “unofficial” cooperation with people from West Germany was practically impossible, not to mention – very dangerous. Moreover, for political reasons as well, any official cooperation program between Moscow Institute of Psychiatry and O. & C. Vogt Brain Research Institute could not possibly exist: Moscow Institute Of Psychiatry was headed at that time by prof. A.V. Snejnevsky. He was notoriously infamous and accused many times in European media for “using of psychiatry for political persecution of dissidents”.

Working in our Laboratory, Katrin Amunts used the described technology at least for four years collecting data for her dissertation, which was defended in 1989. Between 1980 and 1992 we published together about 10 papers, some of them are listed below. In 1992 she brought her data to Düsseldorf. It would be reasonable to assume that she wrote the entire manuscript of the discussed paper.

Guessing from published results, the contribution that A.Schleicher could make was the analysis of the peaks periodicity (fig. 6 and 7), calculation of the thicknesses of the layers (fig. 13 and 14), correlation analysis (fig. 11 and 12), and possibly – discussion of the results. All this work was based on re-processing of profiles originally obtained in Moscow, and it could be done in a matter of weeks, maybe – in a couple of month.

The specific contribution of K.Zilles to this paper is unknown to me. However, I have to add one more fact to this story. To the best of my recollections, it was September of 1994, when I personally visited O. & C. Vogt Brain Research Institute in Dusseldorf and had the honor of meeting prof. K. Zilles. Latest results from my work in the Department of Psychiatry of the Mount Sinai Hospital in New York were included in my presentation. Working there I re-designed MCG technology, adapted it to a new device (Quantimet-570), and expanded its capabilities to automatic measurements of immunohistochemical specimens. Also, I saw with my own eyes microscopes, computers, image processing system and other equipment available at this time in O. & C. Vogt Brain Research Institute. I had to conclude that with available equipment the reproduction of my technology in Düsseldorf was impossible. Moreover, we made a trip from Dusseldorf to Wetzlar, to the headquarters of Ernst Leitz, GmbH, to discuss the possibility to purchase or to rent Leitz-TAS Plus, which was necessary to re-implement MCG technology. Unfortunately, our effort did not succeed, which was one of two reasons why I did not accept the offer of K. Zilles to take a position in his Laboratory. Despite numerous personal conversation during this visit, we never discussed anything even remotely related to this publication.

8. Absent References

Data published in this paper was initially the part of Ph.D. defended by me in 1987, and Katrin Fiedler (Amunts) in 1989 ([13], [14]). However, references to these dissertations in the discussed paper are absent. Moreover, as a requirement for Ph.D. consideration in USSR, main scientific results of the thesis had to be published before the defense. That was indeed the case, see [10], [12], [15], [16]. These papers were not referenced either.

Moreover, careful analysis of all papers of prof. Amunts demonstrates a surprising fact: to the best of my knowledge, none of the papers, published by K. Amunts (or K. Fiedler) in Moscow, Russia or in East Germany were never referenced or mentioned anywhere since 1997. Even the most complete list of publications of K. Amunts (list of publications) starts from this 1995’s paper; traces of all prior publications vanished into the thin air.

Actually, there is one exception, which only underlined the rule: in 2015 a paper was published [19], where  one reference to this publication [8] was provided.

9. Funding and acknowledgements section

Small acknowledgements section mentions the source of funding: “The work was supported by the Deutsche Forschungunsgemeinsshaft (SFB 194/A6)”. However, this work could be supported by this source only partially. Once more: all results published from page 558, the last paragraph, to page 564, and page 566 (methodical aspects), 7.25 pages total, were obtained in Moscow, between 1980 and 1989.

Eight pictures,  figs. 2 – 9 were published between 1985 and 1992, what means – before K. Amunts started working in Dusseldorf, and fig. 10, a-c was obtained as a direct printout of measurements results, which could be obtained only using hardware and software available in Moscow laboratory.

Only 5 out of 14 published pictures (fig. 1 – micro-photograph of area 4, and figures 11-14, based on correlation analysis and measurements of parameters of the layers, performed as “secondary” data processing) could be obtained in Düsseldorf and contributed by A. Schleicher and/or K.Zilles. So, a lion share of published results, at least 57% of the text  material (7.25 of 12.75 pages, not counting 2.25 pages of references) and 43% of picture material (6 figures out of 14), were not supported by above-mentioned funding source. This ratio becomes even bigger if we counted the time needed for contribution to this paper (see section 6 above). In other words, the main part of this work was supported between 1978 and 1989 by the budget of the All Union Center for Mental Health and Institute of Psychiatry of the Academy of Sciences Of USSR. As the reader might guess, “Deutsche money” had very little to do with it.

Authors thanked Vladimir Vinogradov without explaining what was his contribution to this work. However, to the best of my knowledge, Dr. V. Vinogradov is a pathologist from Morosov’s Children’s Hospital in Moscow, who personally performed all postmortem procedures on child subjects, and contributed all children tissue material used in the study of K. Amunts. Under normal circumstances such level of participation should warrant co-authorship of the publication, however, K. Amunts, A. Schleicher and K. Zilles decided otherwise. Additionally, all postmortem studies of adult control were performed and collected by me, as described above (section 3). My contribution in collecting adult brains was not acknowledged at all.

The second paper mentioned above [6] reports the second part of the results of Katrin Amunts’s dissertation [14]. This paper also has no information about the source of brain tissue. The same technology, developed in Moscow, was used for profiles measurement, but there is no proper attribution to the place where it was done.  The same obscure “thanks to Vladimir Vinogradov” hints to Moscow Children Hospital as to the origin of children brains (cases of cerebral palsy, and young cases of “normal” control).

10. So many questions remain…

Working together with K. Amunts we published a lot of results together, including imaging algorithms (see, for example [7], [8] and [9]). Obviously, existence of papers published together in Moscow in the past does not give any rights to publish the same data again, adding new “co-authors” from Düsseldorf. Moreover, in my opinion it was very improper to mix together our original technology and data with additional statistical analysis of cortical profiles performed 9 years later, and to publish the whole bulk of results in the same paper, without proper explanation of original sources. This paper made an impression that the entire work was done in Düsseldorf under the supervision of K. Zilles. The questions that puzzles me a lot even today is: what was the reason for publishing this paper the way it was done?

All technology aspects of our work were published already in Moscow between 1980 and 1989. If additional results were obtained after 1992, when K. Amunts started working in Düsseldorf with professor Zilles, a new original paper could be published with all necessary references to prior developed methods and programs. Why was it necessary to re-publish “from Düsseldorf” our pioneering technology, called “Automatic Morphocorticography” (MCG) created quite independently 10 years before in Moscow? Why to publish the same again, as if it was developed in Düsseldorf with participation of Zilles and Schleicher? It looks even more peculiar, considering that these authors never mentioned MCG technology or Leitz-TAS neither before nor after these two publications.

Quite frankly, when I saw these papers for the first time they surprised me a lot. In 1995, Katrin was a young and relatively inexperienced. Probably she could not influence the decision of professor Zilles, even if she wanted. But why did a renowned professor with 148 papers published at 1995, need two more, which his Institute had very little to do with?

What was the need of hiding the fact that the lion share of published work was based on tissue and results collected in Moscow, not in Düsseldorf?

Any reasonable explanation regarding this subject is welcome…

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.