Channels¶

A channel is one optical configuration of the microscope: the light source and the filters it uses to excite the sample and to select the emitted light. Simulation.channels is a list of OpticalConfig objects, and the output has one image per channel along the c dimension.

The default is a single FITC (GFP) channel.

from microsim.schema.optical_config import lib

sim = ms.Simulation(..., channels=[lib.FITC, lib.TRITC])

Built-in filter sets¶

microsim.schema.optical_config.lib has idealized versions of common filter sets, built from smooth bandpass and longpass edges, with the nominal band edges of Chroma's ET filter sets. They are not measured spectra, and are meant for quick simulations; for real filters, load them from FPbase (below). The interactive charts in this guide use the measured spectra of the Chroma sets.

Bandpass filters are written as center/width in nm. The last column is the string that loads the measured set (Filter sets from FPbase).

For point-scanning confocals, lib also has laser lines: a laser, a dichroic, and an emission bandpass, with generic filter values. Their power is 1e5 W/cm², the peak irradiance at the focus (about 50 µW at the sample with a 1.4 NA objective).

Filter sets from FPbase¶

FPbase lets you enter the optical configurations of your own microscope, with measured spectra of its light sources, filters, and detectors, and has many microscopes entered by others. These configurations can be loaded in microsim with a string of the form "<microscope id>::<config name>":

sim = ms.Simulation(..., channels=["wKqWbg::Widefield Green"])
# equivalent:
ms.OpticalConfig.from_fpbase("wKqWbg", "Widefield Green")

The microscope ID is in the URL of its FPbase page, e.g. fpbase.org/microscope/wKqWbgApvguSNDSRZNSfpN/ (the first few characters are enough). Configuration names are not case-sensitive. Some FPbase microscopes with filter catalogs:

• Chroma filter sets
• Semrock filter sets
• Nikon filter sets
• Zeiss filter sets

Building your own¶

ms.OpticalConfig(
    name="GFP, 488 laser",
    lights=[ms.LightSource.laser(488)],
    filters=[
        ms.Longpass(cuton=495, placement="BS"),
        ms.Bandpass(bandcenter=525, bandwidth=50, placement="EM"),
    ],
    power=2000,  # W/cm² at the sample
)

Filters¶

placement says where a filter sits in the light path:

The defaults are "EX" for Shortpass, "EM" for Longpass, and "ALL" for Bandpass and SpectrumFilter, so set placement explicitly for bandpass filters. Filters in the same path are multiplied together.

Light sources¶

lights is a list of LightSource objects, each with an emission spectrum. LightSource.laser(488) makes a narrow line at 488 nm. With no light source, the channel is illuminated by flat white light, so the excitation filters alone set the spectrum.

Illumination power¶

OpticalConfig.power is the irradiance at the sample in W/cm², summed over all wavelengths after the excitation filters. If it is not set, the irradiance is 1 W/cm².

Rough orders of magnitude for the irradiance at the sample:

For point-scanning modalities, power is the peak irradiance at the center of the focused spot. See Modalities. To change the power of a library channel, use e.g. lib.LASER_488.model_copy(update={"power": 1e4}).

Per-channel exposure¶

OpticalConfig.exposure_ms overrides Simulation.exposure_ms for one channel:

channels = [
    lib.FITC.model_copy(update={"exposure_ms": 50}),
    lib.TRITC.model_copy(update={"exposure_ms": 200}),
]

Bleedthrough and cross-excitation¶

microsim computes every fluorophore in every channel, so crosstalk needs no extra setup. A fluorophore excited by another channel's light (cross-excitation), or whose emission passes another channel's filter (bleedthrough), appears in that channel in proportion to the overlap of the spectra. Which light sources are on while a channel is recorded depends on whether channels are acquired one after another or at the same time.

To see where the signal in each channel comes from in an image, use optical_image_per_fluor(), which keeps the fluorophore dimension:

sim = ms.Simulation(
    truth_space=ms.ShapeScaleSpace(shape=(64, 512, 512), scale=0.064),
    sample=[
        ms.FluorophoreDistribution(
            distribution=ms.CosemLabel(dataset="jrc_hela-3", label="er-mem_pred"),
            fluorophore="EGFP",
        ),
        ms.FluorophoreDistribution(
            distribution=ms.CosemLabel(dataset="jrc_hela-3", label="mito-mem_pred"),
            fluorophore="tdTomato",
        ),
    ],
    channels=[lib.FITC, fitc_longpass],  # (1)!
)
per_fluor = sim.optical_image_per_fluor()  # (c, f, z, y, x)
per_fluor.sum(["z", "y", "x"])  # total photons/s from each fluorophore in each channel

1. fitc_longpass is lib.FITC with its emission filter replaced by ms.Longpass(cuton=500, placement="EM").

Sequential and simultaneous acquisition¶

By default, channels are acquired one after another, so each channel is excited only by its own light sources. Simulation.tracks groups channels that are acquired at the same time, as in the tracks of a point scanner. All light sources of all channels in a track are on together, and each channel's detector records at the same time, so a fluorophore can be excited by any light source of its track:

from microsim.schema.optical_config.lib import (
    LASER_405,
    LASER_488,
    LASER_561,
    LASER_640,
)

sim = ms.Simulation(
    ...,
    channels=[LASER_405, LASER_488, LASER_561, LASER_640],
    tracks=[[0, 2], [1, 3]],  # or by name: [["405", "561"], ["488", "640"]]
)

Each list is shorthand for a Track, e.g. Track(channels=[0, 2]), and every channel must be in exactly one track.

• A light source listed in several channels of a track (e.g. two detection windows behind one laser) is one light source, on once, so those channels must have the same power and excitation-side filters (including the dichroic). Channels in a track must have either the same light sources or none in common.
• On a point scanner, the channels of a track have one pixel dwell time, so they must have the same exposure_ms.
• Each channel detects through its own filters, including its own dichroic. On a real scanner, the channels of a track share one (multiband) dichroic; give them matching dichroics if that matters for your simulation.
• With a Confocal modality, the channels of a track are detected through one pinhole (unless shared_pinhole=False).

Plotting spectra¶

These need the view extra. For one channel:

oc = ms.OpticalConfig.from_fpbase("wKqWbg", "Widefield Green")
oc.plot()  # all filter spectra

fig, axes = plt.subplots(1, 2, figsize=(12, 3.2))
oc.plot_excitation(
    ax=axes[0]
)  # light source and excitation filters, and the photon flux
oc.plot_emission(ax=axes[1])  # emission filters, and their combined transmission

For a whole simulation, Simulation.plot() shows, for each channel (columns): the absorption cross sections, the excitation path, the absorption rates (with the emission spectra, dashed), the emission path, and the photons collected from each fluorophore. Here EGFP and mCherry in the FITC and TRITC channels:

sim.plot()

Spectral detection¶

lib.spectral_detector returns a list of channels that split an emission range into equal bins, as on a spectral confocal:

channels = lib.spectral_detector(bins=8, min_wave=480, max_wave=600, lasers=[488, 561])

This is a simplified model: each bin is an ideal rectangular filter, and a dichroic reflects a 10 nm band around each laser line.