GitHub

Concepts: Overview

SMLMAnalysis is an integration layer: it does little numerical work itself, and instead composes the JuliaSMLM ecosystem's specialized packages into one reproducible pipeline. These pages explain the few ideas you need to use it well and to reason about what a pipeline is doing. Read them top to bottom, or jump to a topic:

  • The Pipeline Model — how analyze() dispatch turns a list of step configs into an analysis, the (result, info) tuple pattern, and state threading. Start here.
  • The JuliaSMLM Ecosystem — what each upstream package does, which steps SMLMAnalysis owns itself, and when to reach for a package directly.
  • Data Model & Provenance — the BasicSMLD localization container and emitter types that flow between steps, the camera and micron coordinate system, and the StepInfo / AnalysisInfo records that capture what happened.
  • The Multi-Target Model — how the single-channel pipeline generalizes to multi-color experiments: a Vector{BasicSMLD} state and cross-channel steps.

For day-to-day tasks (installing, running, multi-dataset and multi-channel acquisitions, I/O, adding a step) see the Workflows section; for the per-step reference see Pipeline Steps.

The mental model in one paragraph

You hand analyze() some data and a description of the analysis. The description is an AnalysisConfig: a camera, an output directory, and an ordered steps vector of typed configs. The pipeline applies each step in turn, passing the localization set from one to the next, and returns the final localizations (AnalysisResult) together with a complete provenance record (AnalysisInfo). Each step is backed by one ecosystem package (or by SMLMAnalysis itself), selected purely by the config's type.

Glossary

  • Localization — a single fitted blinking event: a position with a fitted uncertainty (and photons, background, …). Stored as an Emitter2DFit / Emitter3DFit. Positions are in microns throughout.
  • SMLD — Single-Molecule Localization Data: the SMLMData.BasicSMLD container holding a vector of emitters plus the camera, frame count, and dataset count. This is the value that flows between steps.
  • Camera — an AbstractCamera (IdealCamera or SCMOSCamera) that defines the pixel grid in microns and so the coordinate system every step works in. Set once on the AnalysisConfig; see The camera & coordinate system.
  • Emitter — after Bayesian grouping the word also means a true fluorophore reconstructed from many localizations; in raw data "emitter" and "localization" coincide.
  • Step — one transformation in the pipeline, named by its config type (DetectFitConfig, FilterConfig, …). Each step has an analyze(state, cfg) method.
  • Config — a typed struct describing a step's parameters (<: SMLMData.AbstractSMLMConfig). Configs carry parameters, not logic.
  • Info — the second element of every analyze() return tuple: a StepInfo per step (wrapping the upstream package's own info struct), aggregated into an AnalysisInfo for the whole run.
  • Dataset — an independent acquisition block within one SMLD (e.g. a cell, an ROI, or a registered segment). Frames are numbered per dataset. See Multi-Dataset.
  • Channel / target — one color in a multi-color experiment. Several SMLDs are analyzed together by MultiTargetConfig. See The Multi-Target Model for the concept and Multi-Channel for the workflow.
  • Tuple pattern — the JuliaSMLM convention that analysis functions return (result, info) rather than mutating arguments or hiding state in globals.
  • CRLB — the Cramér–Rao lower bound, the theoretical minimum localization uncertainty; the per-localization σ reported by the fitter is a CRLB estimate, and several steps use it as a weight.