Limits...
Smart-aggregation imaging for single molecule localisation with SPAD cameras

View Article: PubMed Central - PubMed

ABSTRACT

Single molecule localisation microscopy (SMLM) has become an essential part of the super-resolution toolbox for probing cellular structure and function. The rapid evolution of these techniques has outstripped detector development and faster, more sensitive cameras are required to further improve localisation certainty. Single-photon avalanche photodiode (SPAD) array cameras offer single-photon sensitivity, very high frame rates and zero readout noise, making them a potentially ideal detector for ultra-fast imaging and SMLM experiments. However, performance traditionally falls behind that of emCCD and sCMOS devices due to lower photon detection efficiency. Here we demonstrate, both experimentally and through simulations, that the sensitivity of a binary SPAD camera in SMLM experiments can be improved significantly by aggregating only frames containing signal, and that this leads to smaller datasets and competitive performance with that of existing detectors. The simulations also indicate that with predicted future advances in SPAD camera technology, SPAD devices will outperform existing scientific cameras when capturing fast temporal dynamics.

No MeSH data available.


Projections from model simulations.(a) Localisation performance of current SPAD versus a modelled future SPAD (b) Future SPAD as compared with sCMOS and emCCD. (c) The effect of 60x objective on the localisation of current and future SPADs and (d) the effect of hot pixels on the localisation error of current SPAD. All solid lines are guides to the eye. Dashed lines represent blink durations with <90% detection sensitivity.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC5120354&req=5

f5: Projections from model simulations.(a) Localisation performance of current SPAD versus a modelled future SPAD (b) Future SPAD as compared with sCMOS and emCCD. (c) The effect of 60x objective on the localisation of current and future SPADs and (d) the effect of hot pixels on the localisation error of current SPAD. All solid lines are guides to the eye. Dashed lines represent blink durations with <90% detection sensitivity.

Mentions: Figure 5a compares the localisation error resulting from the current SPCImager, with the projected, F-SPAD. Both fixed and smart frame aggregation are considered. The error is seen to (approximately) half with the future SPAD device. Figure 5b shows the F-SPAD (with smart aggregation) as compared with existing commercial emCCD and sCMOS cameras. As previously, we consider the effect of merging consecutive emCCD/sCMOS localisations if within 40 nm of each other. The solid lines are guides to the eye, with the dashed section showing blink duration events with <90% detection sensitivity. The results show the future SPAD camera outperforming the emCCD for almost all blink durations. The F-SPAD matches the sCMOS in all but below the 20–30 ms times, in which case sCMOS gives better localisation results, but lower detection sensitivity, dropping below the 90% threshold. The SPAD camera remains above 90% for all blink durations. Figure 5c shows the effects of switching to the 60x objective, for the cases of the current and future SPAD camera and smart aggregation is assumed throughout. It is noted that the 60x objective reduces the localisation error of the current SPAD by a moderate extent (by around 20% for longer blink durations), but has negligible effect on the F-SPAD. The reason is that with the future camera, the DCR is no longer significant; the error is mainly caused by the background photon count, which simply gets re-distributed across the pixel array as the magnification is changed (in a similar way, the simulated error curve for the sCMOS was found to be largely unaffected by the change in assumed objective).


Smart-aggregation imaging for single molecule localisation with SPAD cameras
Projections from model simulations.(a) Localisation performance of current SPAD versus a modelled future SPAD (b) Future SPAD as compared with sCMOS and emCCD. (c) The effect of 60x objective on the localisation of current and future SPADs and (d) the effect of hot pixels on the localisation error of current SPAD. All solid lines are guides to the eye. Dashed lines represent blink durations with <90% detection sensitivity.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC5120354&req=5

f5: Projections from model simulations.(a) Localisation performance of current SPAD versus a modelled future SPAD (b) Future SPAD as compared with sCMOS and emCCD. (c) The effect of 60x objective on the localisation of current and future SPADs and (d) the effect of hot pixels on the localisation error of current SPAD. All solid lines are guides to the eye. Dashed lines represent blink durations with <90% detection sensitivity.
Mentions: Figure 5a compares the localisation error resulting from the current SPCImager, with the projected, F-SPAD. Both fixed and smart frame aggregation are considered. The error is seen to (approximately) half with the future SPAD device. Figure 5b shows the F-SPAD (with smart aggregation) as compared with existing commercial emCCD and sCMOS cameras. As previously, we consider the effect of merging consecutive emCCD/sCMOS localisations if within 40 nm of each other. The solid lines are guides to the eye, with the dashed section showing blink duration events with <90% detection sensitivity. The results show the future SPAD camera outperforming the emCCD for almost all blink durations. The F-SPAD matches the sCMOS in all but below the 20–30 ms times, in which case sCMOS gives better localisation results, but lower detection sensitivity, dropping below the 90% threshold. The SPAD camera remains above 90% for all blink durations. Figure 5c shows the effects of switching to the 60x objective, for the cases of the current and future SPAD camera and smart aggregation is assumed throughout. It is noted that the 60x objective reduces the localisation error of the current SPAD by a moderate extent (by around 20% for longer blink durations), but has negligible effect on the F-SPAD. The reason is that with the future camera, the DCR is no longer significant; the error is mainly caused by the background photon count, which simply gets re-distributed across the pixel array as the magnification is changed (in a similar way, the simulated error curve for the sCMOS was found to be largely unaffected by the change in assumed objective).

View Article: PubMed Central - PubMed

ABSTRACT

Single molecule localisation microscopy (SMLM) has become an essential part of the super-resolution toolbox for probing cellular structure and function. The rapid evolution of these techniques has outstripped detector development and faster, more sensitive cameras are required to further improve localisation certainty. Single-photon avalanche photodiode (SPAD) array cameras offer single-photon sensitivity, very high frame rates and zero readout noise, making them a potentially ideal detector for ultra-fast imaging and SMLM experiments. However, performance traditionally falls behind that of emCCD and sCMOS devices due to lower photon detection efficiency. Here we demonstrate, both experimentally and through simulations, that the sensitivity of a binary SPAD camera in SMLM experiments can be improved significantly by aggregating only frames containing signal, and that this leads to smaller datasets and competitive performance with that of existing detectors. The simulations also indicate that with predicted future advances in SPAD camera technology, SPAD devices will outperform existing scientific cameras when capturing fast temporal dynamics.

No MeSH data available.