The story of Grey Level Index: forty years of misleading consistency?

For whatever reason (probably, to prove the priority by consistency), the same approach was described again in 1990, now without Andreas Wree, in the paper “A quantitative approach to cytoarchitectonics: analysis of structural inhomogeneities in nervous tissue using an image analyzer” (Journal of Microscopy, Vol. 157, Pt 3, March 1990, pp. 367-381). Embarrassingly similar to 1986’s publication this paper includes very few new results: (1) a table with statistical variability data for GLI of primary visual cortex, (2) a graph with comparison of GLI profiles of different areas of rat’s visual cortex (Oc2ML and Oc1M were compared in 1986 paper vs. OC1M and OC1B compared in 1990 paper). Also, in 1990 paper the processing and smoothing of the profiles looked improved, so they reflected the laminar pattern much better. In all other aspects 1986 and 1990 papers are practically identical: the structure and titles of the sections are the same, 10 out 12 of pictures are identical. The text of the second paper is slightly inflated compared to the first one, but essentially the same: some sections show up to 85% syntactic similarity, and 82% semantic similarity, as reported by Dandelion demo tools(

Despite the controversy, one important detail should be noticed from these papers: the choice of term “Grey level Index” is explained by the limitations of image processing functionality of the Micro-Videomat system.

Micro-Videomat was a hardware-based device, were the signal from TV-camera was processed by a series of specialized circuit boards. Each of them was dealing with a certain stage of signal processing: image thresholding, measuring mask selection, calculation of the pixel area of discriminated image, etc. The system did not have digital memory, so the choice of image processing functions was limited to a simple combination of binary operations: signal selection “above”, “below” or “between” two thresholds, basic logical functions (AND, OR, XOR), masking, pixels counting etc. Due to the system design and rather limited capacity of RAM, the attached computer (Wang-720 or Wang-2200) could not be used for image processing either; it was used merely for accumulation and post-processing of measurement results. Thus, the measured value of image area above selected threshold incorporated all stained elements, included in the “measuring mask”: neuronal perikaria, nuclei of the glial cells, endothelial cells of the micro-vessels and capillars, as well as some staining artifacts. The GLI was defined as “the ratio of the area covered by image elements which are darker than a given grey value threshold, to the entire area of the measuring field”. In other words, the value was not corrected by image processing. Additionally, for various reasons, the GLI measurements were not corrected for the thickness of the section (20 microns). Therefore, instead of usually accepted term “neuronal area fraction”, or “area fraction of stained elements“, authors employed rather vague term “grey level index“.

As a side note I would like to add that Micro-Videomat system for its time was top-of-the- line image processing device. I was lucky to be able to use this system in Moscow, USSR, in late 1970s and early 1980s in the Central Research Laboratory of the Second Moscow Medical School and in Moscow Institute of Psychiatry. We used it in several automatic microscopy projects for  quantitative analysis of histological sections, including serial sections of Nissl-stained human neocortex. Our first paper describing the application of Micro-Videomat for quantitative analysis of secretory cycle of EC cells population was published in 1981 (Bulletin of Experimental Biology and Medicine, XCI, 5, 628-630).  At the same time we replaced Micro-Videomat by more powerful system – Leitz-TAS, which had advanced image processing functionality due to binary and grey-level image memory and comprehensive implementation of Mathematical Morphology. Our first description of profiles of volume fraction, numerical density and surface area of areas 10, 17 and  44 was published in 1984, and more comprehensive description of the technology – in 1985 and 1987. I will describe our cortical profile registration  algorithms, which included elimination of vessels, glia cells and staining artifacts,  in later post (see “The story of MCG: lost opportunity” article). Unfortunately, these papers, as well as many others, published “behind the iron wall” were completely ignored by West- European scientists.

Leave a Reply

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