@palebluedot , welcome to the forum
your image has not much to offer for conversion. Tried it manually in colour and B&W
Even if we’d squeeze more detail out of the image, it would mostly be grain. Look at the histogram. wide U-shapes don’t convert well in many cases.
The one thing to try would be to scan with exposure set to high- and lowlights respectively and combine both into an HDR file. This could stretch the extremes in the histogram a little.
To convert, I’d crop both top and bottom heavily (like in the screen above)
In general, NLP can only work what it’s getting fed as described here. The less info there is, the more erratic conversions will be.