SEM/TEM Particle Size & Nanomaterial Screening

Turn scanning or transmission electron microscope images into a particle size distribution report in about ten seconds per image.

Try it in one click: leave the image list empty and press Start. The Actor generates three synthetic SEM images with known particle sizes, analyses them and shows how close the measurement is to the truth. Demo runs are free.

Give it the images of one sample. Original TIFF files from Zeiss or Thermo Fisher (FEI) microscopes are calibrated automatically from their metadata; for other images, give the length of the scale bar or the pixel size. It returns D10, D50 and D90 with 95 % confidence intervals, Feret diameters, shape factors, a quality check, a screening against the EU definition of nanomaterial and a PDF report you can file or send.

EU nanomaterial screening

Is your powder a nanomaterial under the EU definition (Commission Recommendation 2022/C 229/01)? Criterion (a) is met when 50 % or more of the particles, counted by number, have at least one external dimension between 1 and 100 nm, and electron microscopy is the reference way to count them. Every run now answers with numbers, not just a yes or no:

• the share of particles whose smallest dimension (minimum Feret diameter) is 100 nm or less, with its 95 % confidence interval;
• an interval that also allows for calibration error and edge blur;
• the median minimum Feret diameter with its confidence interval;
• a verdict: meets, not met, or borderline when the interval includes 50 %, so you know when more images or a complementary method are needed;
• warnings when fewer than 500 particles were measured, when the magnification is too low near 100 nm, when many agglomerates were flagged or when the smallest particles may be missed.

This is a screening for R&D, product stewardship and pre-registration checks. It does not assess the volume specific surface area (materials below 6 m²/cm³ are not nanomaterials), dimensions below 1 nm, or the thickness of platelets lying flat, and it does not replace a regulatory determination.

Who it is for

• R&D teams working on powders, pigments, catalysts, nanoparticles, ceramics or battery materials
• Labs that already have SEM/TEM images and spend hours measuring particles by hand in ImageJ
• AI agents that need particle statistics from microscope images (callable through the Apify MCP server)

What you get

Output	Content
Dataset item	particles counted, D10/D50/D90, mean and standard deviation (nm), 95 % confidence interval on D50, precision reached, particles needed for ±2 %, quality status and checks, EU nanomaterial verdict, share of particles ≤ 100 nm and its interval, median minimum Feret
REPORT.pdf	results, confidence intervals, nanomaterial screening, distribution and cumulative curve, per-image results, quality checks, method, traceability
particles.csv	one row per detected object: status, equivalent diameter, min/max Feret, aspect ratio, circularity, solidity, position
control-sheet.jpg	every image with the detected outlines, to check the segmentation at a glance
summary.json	all statistics in machine-readable form

How it works

1. The pixel size is read from the microscope metadata (Zeiss SmartSEM, Thermo Fisher/FEI TIFF), or measured on the scale bar of the image data band, or given by you. When both metadata and scale bar are available they are cross-checked.
2. Particles are segmented (Gaussian smoothing, Otsu threshold, morphological cleaning) and touching particles are split with a distance-transform watershed.
3. Objects cut by the image border are excluded. Likely agglomerates (low solidity or aspect ratio above 2) are flagged and excluded from the statistics.
4. Number-based statistics are computed. Confidence intervals come from 10,000 bootstrap resamples with a fixed seed, so the same images always give the same numbers.
5. Nine automatic checks rate the run green, amber or red: scale consistency (between images, metadata and scale bar), particle count, statistical precision, resolution, agglomerates, border losses, image-to-image consistency, contrast and saturation.
6. The nanomaterial screening counts the kept particles whose minimum Feret diameter is 100 nm or less. A Wilson interval covers sampling; a systematic band covers calibration (at least ± 2 %) and edge blur (± 2 pixels). The verdict is "meets" or "not met" only when the whole interval is on one side of 50 %, and it is set to "borderline" when counting the flagged objects would change the conclusion.

Accuracy

• Synthetic SEM images with known particle sizes (the free demo): 580 particles measured for 578 present; D50 within +3.7 % of the truth.
• Nanomaterial screening on synthetic samples with a known share of particles at or below 100 nm: measured within 1 to 6 points of the truth (true 96 %, 70 %, 62 % and 10 %), always inside the interval with uncertainties; the 62 % sample, close to the threshold, is correctly reported as borderline.
• Limit: strongly agglomerated powders. When touching particles cannot be told apart, some are measured as one, and D50 reads 10 to 18 % high on such samples. The agglomerate check then turns amber or red: disperse the sample better, or ask for the verified service.

Input example

{
  "imageUrls": [
    "https://example.com/sample-A-field1.tif",
    "https://example.com/sample-A-field2.tif",
    "https://example.com/sample-A-field3.tif"
  ],
  "scaleBarNm": 200,
  "sampleName": "SAMPLE-A"
}

Set "reportLanguage": "fr" for a report in French.

Use 3 to 10 images of the same sample, taken at the same magnification, so that 300 or more particles are measured (500 or more for the nanomaterial screening). Images must be direct download links (TIFF, PNG or JPG, up to 60 MB each). Send the original TIFF files when you can: they carry the calibration. scaleBarNm applies to all images of the run, so do not mix images whose scale bars show different values unless they are original TIFF files.

Pricing

You pay per image analysed. There is no charge for images that cannot be downloaded, and demo runs (no image URLs) are free.

Good to know

• Works best on well-separated, compact particles on a contrasting background. Elongated fibres, porous aggregates and overlapping layers are flagged rather than measured.
• Measurements are 2D projections. Thickness is not measured.
• One sample per run, with a single particle population. Mixed populations at very different scales need separate runs at suitable magnifications.
• This is automated image analysis, not an accredited (ISO/IEC 17025) laboratory measurement.
• Need a report reviewed and signed by a materials imaging scientist? A verified analysis service is available on request (contact through the Actor's issues tab).