The 250 nm inflation: Dobrucki explains that even the smallest light-emitting object — a single 3 nm fluorescent protein — produces a diffraction-limited image approximately 250 nm wide. This is an 80-fold inflation. A 25 nm microtubule appears 250 nm wide (10× inflation). An endosome that is 250 nm appears roughly its true size. The critical implication: below ~250 nm, all objects appear the same size regardless of their true dimensions. Their apparent size tells you nothing about their physical size — only the optics.
Why LoG filtering works: The Laplacian-of-Gaussian filter is perfectly matched to blob detection. The Gaussian component smooths the image to suppress noise at frequencies above the expected dot size. The Laplacian component (second derivative) responds maximally to blob-like structures at the Gaussian's scale. The combined LoG produces strong positive responses at the centers of blobs and strong negative responses at their edges — with zero crossings at the boundary. By tuning the Gaussian sigma to match the PSF width, the LoG filter becomes a matched detector for diffraction-limited spots.
The Rose criterion for dots: A dot is reliably detectable only when its peak intensity exceeds 5× the local noise (SNR > 5). For Poisson-limited fluorescence, this means a dot must produce at least 25 detected photons at its peak pixel. With a typical PSF spreading the signal across ~50 pixels, a detectable dot needs at least ~1,250 total photons — a real constraint for weakly labeled targets.
Counting below the diffraction limit: When two FISH signals are separated by less than 250 nm, they appear as a single brighter spot. If each single dot produces approximately I photons, a cluster of n dots produces approximately n×I photons. Dividing the total intensity by the single-dot calibration intensity gives an estimate of the number of overlapping dots. This is inherently approximate — Poisson noise means a 2-dot cluster has √(2I) noise, giving roughly 7% uncertainty in the count for typical signal levels.
Any object smaller than ~250 nm looks the same size under the microscope — the optics inflate it. A single fluorescent molecule and a 100 nm vesicle both appear as ~250 nm spots. Dots Detection uses a filter matched to this spot size to find them, and estimates how many overlapping dots contribute to bright clusters by dividing total brightness by the expected single-dot brightness.
