2. Competing comprehensive descriptions of area fraction profiling methods: who was first?
Even though exact determination of priorities seems useless, and even impossible, it is interesting to notice that first attempts to calculate number of cells in biological tissue in microscopic images could be found since 1940’s (see, for example: Abercrombie, M.: Estimation of nuclear population from microtomic sections. Anat. Rec. 94, 239–247 (1941)). Unsatisfied by limited reliability of counting methods based on microphotography, E. Agduhr published original approach of estimation of the number of neurons in microscopic images of serial sections of nervous tissue, corrected for diameter of the cell as well as for the thickness of the section (Beitrag zur Technik für die Bestimmung der Anzahl Nervenzellen je Volumeneinheit Gewebe. Anat. Anz. 91, 70–81 (1941)). These and similar methods, however, were based on “manual” counting.
Apparently, the first computer-based technology of evaluation of layered structure of human cortex was implemented and published by Herbert Haug in 1976 (“Die verschiedenen Verfahren zur Werteerfassung in der biologischen Morpfometrie ans Stereologie”. Microsc. Acta 78, 197-220, (1976), see fig below). We can notice that H.Haug developed three approaches to evaluate cortical structure. The first was based on systematic horizontal scanning of the section, and the profile of the cortex was plotted as densities of neurons per mm^3 with confidence limit at each profile point for p=0.05. Practically the same approach was implemented and published by Schleicher and Zilles two years later.
After 1978 numerous papers were published essentially by the same group (Schleicher, Zilles and co-authors), describing the application of GLI profile methodology to different species (Galago demidovii, gray mouse lemur, some other primates, albino rat, guinea pig, albino mouse, domestic pigeon, and, finally, human).
However, the explanation of the technology of measurement of GLI value seems to be published four years after the first publication, in 1982. It was described rather meticulously in the paper “Estimation of volume fractions in nervous tissue with an image analyzer” by Andreas Wree, Axel Schleicher and Karl Zilles (Journal of Neuroscience Methods, 6, 1982, 29″43 29). Here we can find the comparison of GCC (grey cell coefficient), based on manual “point-counting” stereological method, and GLI (grey level index), obtained automatically with image analyzer. The paper also carefully analyzed the correlation between uncorrected GLI value and the corresponding volume fraction measurements under various conditions for the sections of different thicknesses. One of many important results of this study was a practical conclusion, that in the sections up to 4.6 micrometers the value of GLI is statistically indistinguishable from volume fraction (see below, Section 4A).
Interestingly, the description of the software and hardware used for GLI measurements appeared in two surprisingly similar papers by different authors. The earlier paper was published in 1983 by Bernward Sauer from the Department of Neuroanatomy of Medical School of Hunnover: “Semi-automatic analysis of microscopic images of the human cerebral cortex using the grey level index,” Journal of Microscopy, Vol. 129, Pt 1, January 1983, 75-87 (text PDF). The second paper was published three years later by A. Schleicher, K. Zilles and A. Wree, at that time – from the Institute of Anatomy of the University of Cologne) was entitled “A quantitative approach to cytoarchitectonics: software and hardware aspects of a system for the evaluation and analysis of structural inhomogeneities in nervous tissue” (Journal of Neuroscience Methods, 18 (1986) 221-235, text PDF). Despite striking similarities to Sauer’s paper (identical image processing device – Micro-Videomat-2, the same approach to the discrimination threshold selection, very similar image measurement approach and practically identical calculation method) there is no reference to his publication of GLI technology. Sauer’s paper, however, has multiple references to GLI-related papers by A. Schleicher and K. Zilles.