dontcha love it —
” In our opinion, the problem is so vast and severe that the damage caused by the users’ processing of the data can easily over-shadow any measurement imperfections, ruining instrumentation improvements and rendering them potentially worthless.” This is a quote from a paper entitled “Study of user influence in routine SPM (scanning probe microscopy) data processing” D Nečas and P Klapetek, (I recognize the names and email addresses of these individuals from Gwyddion support groups and comments) in Measurement Science and Technology. “eye roll” I appreciate the sentiment, and I join the voices that comment. I know it is disappointing to have tools not used “properly”, and from my perspective however, coming from an interest in scientific diagramming, particularly  looking at the countless diagrams about  tissues, cells, cell structures, organelles, protein cascades, signaling pathways and all things relational. While the comments might need to be said, and I will say them as well, this doesn’t make what we attempt to do as we image TEMs AFMs and other images “worthless” but what it does do is make us vigilant.

Comparisons before and after processing are all that is needed here and that goes with common good practice in any branch of science.
One quote from this aper — “Finally, just selecting the right tools for the data analysis task also seems to be a key issue. Here, better SPM education, training and documentation would probably help;” comes to the main purpose of my looking at various software applications to examine countless images of, from various researchers, of SP-D and DMBT1. I have concluded that the processing is just an aid to the eye, not the reverse.

COMMENT:  After 8 more months of trying to produce an informative diagrammatic model  — using “image processing” of SP-D images produced with AFM, and beginning the use “signal processing of the plots” (another story) and having a much better understanding of Gwyddion software as a mechanism to do this, I would probably add emphasis to the value of up front “documentation” of every possible use and misuse of processing software.  Not everyone will spend the time to do this… and while doing this with Gwyddion i did find huge difficulties in creating a “line” through molecules to get grayscale plots along that line. In fact there was a gross misrepresentation of grayscale plots because if my inability (or perhaps it wasn’t easily found in their software) to draw a freehand or segmented line through the molecules.  In addition, and more seriously, the lines i drew vertically were grossly skewed when plotted.   So  one does need to be “awake” or in the current vernacular “woke” to such errors in processing.

I was reading an article on using ImageJ for measurements on microscope images (stereology and morphometry as it was called in the late “dark ages” of the 1970s) which gave a rational for using edge detection (in ImageJ) for refining, for segmenting, for counting (in this case nuclei) objects automatically.  While I appreciate the recent nature of that publication, i marveled that I still feel the way I felt in the 1980s that it was far faster and more accurate for me to “look” at the image and manually tick off the nuclei as i counted them using a marker on my micrograph (yep, wet darkroom, printing negatives, and all the efforts that went with it – ha ha). It was of course in the day when computers were slow, programs for processing (i love that word applied to micrographs ha ha) were cumbersome and recording data in a spreadsheet meant hand entering each value.

I purposed then, and still do, that the proliferation of special image processing programs is nice, but not that useful yet, and while i get no feedback, or monetary remuneration for saying the following, in fact I don’t even have an updated version of the software, Photoshop is a wonderful, efficient, easy to use image informatics software.  It doesn’t do measuring but image manipulation is super easy, and exporting to numerous formats for measuring is convenient in programs like ImageJ where distance, numbers etc can be obtained.

The sliders for adjusting image processing algorithms are what give the visual input (from each investigator) which is likely better input at this point in AI development, than an algorithm which is applied without user input.

Case in point might be edge detection in Gwyddion, ImageJ, and Photoshop 6  (yep old version and a new version would be much better) shows me that i have to run through every Data processing>Presentation>Edge detection option in Gwyddion,  or as Process>find edges in ImageJ, and Stylize>Find edges in photoshop.  There are more options in Gwyddion, but few show results that can help determine the peaks or LUT values in AFM images, there is no slider option in ImageJ, nor in Photoshop, so whatever algorithm is in their programs is not modifiable… that said, in ImageJ, it could likely be programmed differently in this open source and editable software.  It seems highly unlikely that Photoshop would be so gracious.

Here is a selection of images processed (of original stated above (upper left) and the image imported into gwyddion next to it on the right) and rows of the options for edge detection within gwyddion noted with each outcome.  For me, this program for edge detection did not provide information on the structure of the collagen like domain (or any domain in SP-D ). Edge detection in corelPhotoPaint and Photoshop perform differently.