AI + SEM Track Detection
Automated CR-39 Track Analysis
Reading a CR-39 nuclear track detector by hand means counting etched tracks in a handful of microscope fields of view and extrapolating to the whole chip. It is slow, subjective, and blind to variation across the detector surface. BlankSlate Innovation replaces that workflow with SEM large-area mapping combined with an AI-driven track classifier that counts and characterizes tracks across the entire chip.
Send exposed chips to BSI for full analysis, or work with us on experiment design, etching, and imaging so you capture the cleanest data. The methodology is validated in our peer-reviewed paper on AI analysis of CR-39 with SEM large-area mapping.
What you get back
From Exposed Chip to AI-Labeled Tracks
The ship-back analysis service runs your exposed CR-39 through the SEM large-area mapping pipeline and the AI-driven track classifier, returning quantitative, reproducible results across the whole detector.
Full-Chip Particle Count
Tracks counted across the entire chip area — not just sampled fields of view, so there is no extrapolation error.
Particle-Type ID
Alpha, proton recoil, and fission fragment tracks classified by the AI, not just counted as undifferentiated marks.
Energy Estimates
Derived from track diameter, length, and morphology captured in the high-resolution SEM maps.
Noise Discrimination
Genuine particle tracks separated from polishing artifacts and surface contamination.
Head to head
Manual Counting vs SEM + AI Analysis
Optical microscopy is fine for a quick look, but it does not scale to full-chip, quantitative, reproducible track analysis. Here is how the two approaches compare.
| Capability | Manual Optical Microscopy | BSI SEM Large-Area Mapping + AI |
|---|---|---|
| Area analyzed | Sampled fields of view, extrapolated to the chip | Full chip area, tiled into a single map |
| Track counting | Manual, operator-dependent | Automated and reproducible |
| Particle-type identification | Difficult; relies on operator judgement | Alpha, proton recoil, fission fragment classified |
| Energy information | Rarely quantified | Estimated from track diameter, length, morphology |
| Artifact rejection | Manual, inconsistent | Source-vs-noise discrimination built in |
| Resolution of fine track detail | Limited by optical microscope | High-resolution SEM imaging |
| Validation | Lab-specific | Peer-reviewed (doi.org/10.1016/j.net.2025.103738) |
The workflow
How the Ship-Back Service Works
BSI supports the full CR-39 workflow — run the experiment yourself with our guidance, or ship exposed chips back to us for imaging and AI analysis.
01
Experiment Design
We help set up exposures, geometries, and shielding for your particle type, energy range, and source strength.
02
Etching
Optimize NaOH concentration, temperature, and etch time for alpha tracks, fission fragments, proton recoil, or boron-coated thermal-neutron capture.
03
SEM Large-Area Mapping
High-magnification SEM images are tiled into a single map of the whole etched surface for reliable large-area coverage.
04
AI Classification
The AI classifier delivers counts, particle-type identification, energy estimates, and artifact rejection.
Prefer to run your own imaging? We also advise on optical and electron microscopy procedures, magnification, contrast, and tiling strategies to produce reliable large-area maps in your own lab. See the CR-39 best-practice guide for handling, etching, and imaging detail.
Questions researchers ask
CR-39 AI Analysis FAQ
Can AI analyze CR-39 nuclear track detectors?
Yes. BSI pairs SEM large-area mapping with an AI-driven track classifier to analyze exposed CR-39 chips. The AI locates and counts etched tracks across the full chip area, identifies particle type, and estimates energy from track morphology. The methodology is documented in BSI's peer-reviewed paper on AI analysis of CR-39 with SEM large-area mapping.
How do I automatically count tracks on CR-39?
Rather than manually counting a few sampled fields of view under an optical microscope, you can send exposed CR-39 chips to BSI. The SEM large-area mapping pipeline images the whole chip and the AI classifier returns a particle count across the full chip area, not just sampled regions. This removes the sampling extrapolation and subjective counting that limit manual optical methods.
What is SEM large-area mapping of CR-39?
SEM large-area mapping is a scanning electron microscopy workflow that tiles many high-magnification images together into a single map covering the entire etched CR-39 surface. It resolves fine track detail that optical microscopy can miss, and provides the high-resolution imagery the AI classifier uses to count and characterize tracks. You can explore an interactive example of a CR-39 chip exposed to Am-241 alpha particles here.
Can the AI tell alpha tracks from proton recoil or fission fragments?
Yes. The AI track classifier performs particle-type identification, distinguishing alpha tracks, proton recoil, and fission fragments. It also estimates energy from track diameter, length, and morphology, and applies source-versus-noise discrimination to separate genuine particle tracks from polishing artifacts and surface contamination. See an example of AI-detected U-238 fission-fragment tracks here.
Do I have to buy the detectors from BSI to use the analysis service?
BSI manufactures CR-39 detectors (standard chips at $1/cm2 and boron-coated chips at $3/cm2) and offers a ship-back SEM and AI analysis service. You can order chips, request analysis of exposed chips, or both. Describe your exposure conditions and what you are trying to detect on the CR-39 service request form, and BSI advises on experiment design, etching, and imaging.
Get started
Order Chips or Start an Analysis Project
Tell us how many chips you need, whether you want boron coating, and what you are trying to detect. For ship-back analysis, describe the exposure conditions and what you are looking for in the data.
Related: passive neutron detection · CR-39 for teaching labs · published research