Alpha particle spectroscopy using FNTD and SIM super‐resolution microscopy

Structured illumination microscopy (SIM) for the imaging of alpha particle tracks in fluorescent nuclear track detectors (FNTD) was evaluated and compared to confocal laser scanning microscopy (CLSM). FNTDs were irradiated with an external alpha source and imaged using both methodologies. SIM imaging resulted in improved resolution, without increase in scan time. Alpha particle energy estimation based on the track length, direction and intensity produced results in good agreement with the expected alpha particle energy distribution. A pronounced difference was seen in the spatial scattering of alpha particles in the detectors, where SIM showed an almost 50% reduction compared to CLSM. The improved resolution of SIM allows for more detailed studies of the tracks induced by ionising particles. The combination of SIM and FNTDs for alpha radiation paves the way for affordable and fast alpha spectroscopy and dosimetry.


Introduction
The use of alpha particle emitting radionuclides for the treatment of cancer is increasing rapidly (Sofou et al., 2004;Sartor et al., 2012;Jadvar & Quinn, 2013;Bandekar et al., 2014). Alpha radiation has a high Lineal Energy Transfer (LET), a measure for the amount of energy deposited per distance travelled, which leads to high biological effectiveness (Franken et al., 2011;Tracy et al., 2015) and insensitivity to the oxygen concentration in the tumour (Barendsen et al., 1966;Sgouros et al., 2010). This can be a promising prospect for the treatment of tumours resistant to conventional radiation therapy (Elgqvist et al., 2014). Unfortunately, to this day alpha radiation spectroscopy and dosimetry remain challenging due to the short range of alpha particles (less than 100 µm in water) J.J.M. Kouwenberg (Sgouros et al., 2010). Improvements in such measurements are expected to greatly enhance the development of alpha particle mediated cancer treatment.
Fluorescent nuclear track detectors (FNTD) have shown to be an excellent tool for alpha radiation research (Sykora et al., 2009;Kouwenberg et al., 2016), offering a high spatial resolution in combination with fast and stable read-out using confocal laser scanning microscopy (CLSM). FNTDs are aluminium oxide crystals with carbon and magnesium dopants, giving them a characteristic green colour (M. S. . The dopants introduce defects in the crystal, called colour centres, that undergo radiochromic transformations when irradiated (G. M. . These transformed colour centres are very stable and are fluorescent with an excitation wavelength of ß650 nm and emission at 750 nm. The FNTD can therefore be imaged using fluorescence microscopy to yield high-resolution images of the tracks induced by passing alpha particles in the crystal (Bartz et al., 2013). The direction, length and spatial scattering of these tracks is used to calculate the initial energies and directions of alpha particles. From this the various characteristics of the alpha radiation field, like absorbed dose rate and biological effectiveness, can be deduced. The spatial and energy resolutions of alpha tracks in FNTDs are however restricted by the resolution limit of CLSM. Here, we investigated the use of an alternative fluorescence super-resolution microscopy technique, structured illumination microscopy (SIM) (Lukosz & Marchand, 1963;Gustafsson et al., 2008), for FNTD read-out, utilising both its fast read-out and high spatial resolution to improve spatial and energy resolution in alpha radiation measurement.

FNTD Irradiation
Al 2 O 3 :C,Mg Fluorescent Nuclear Track Detectors (Landauer, Inc., Glenwood, Illinois, USA), kindly donated by Landauer, Inc., were irradiated using a 1.1 cm diameter 394 kBq Am-241 source with an estimated 1.3 µm layer of americium oxide and a 2 µm protective gold layer (Czech Metrological Institute, Brno, Czech Republic), placed behind a 3D printed honeycomb collimator. The honeycomb collimator restricted outgoing alpha particles to a maximum vertical angle of ß45°, a requirement for sufficient penetration in the FNTD (Kouwenberg et al., 2016). Four 8 × 4 × 0.5 mm 3 FNTDs were irradiated at a 5.00 ± .05 mm distance from the source for 267 s each above the centre of the collimator to induce an approximate water equivalent dose of 1 Gy in the detector.

FNTD Read-out
Small arbitrarily chosen volumes near the centre of the FNTDs were read-out using both CLSM and SIM super-resolution microscopy. The measured volumes were not the same for both devices. A Leica SP5 (Leica Microsystems, Mannheim, Germany) with an Avalanche Photo Diode (APD) behind a Chroma Technology ET655lp 655 nm long pass filter (Chroma Technology Corp., Bellows Falls, Vermont, USA) together with a 633 nm excitation laser was used in the CLSM read-out. A HCX PL APO CS 63 × 1.4 NA Oil objective, 2.5 zoom factor, 40 Hz scan speed and 2 times line averaging were used to yield 96 × 96 × 490 nm 3 sized voxels in a 98 × 98 × 10 µm 3 read-out volume. Imaging of one read-out volume took 25 min and no post-processing was performed on the resulting images.
SIM imaging was performed on a Zeiss Elyra PS1 with an Andor iXon DU 885 EMCCD camera (Carl Zeiss AG, Oberkochen, Germany) behind a LP655 filter with a Plan Apochromat DIC 63 × 1.4 NA Oil objective and 642 nm laser excitation with 500 ms exposure times and EM gain set at 30. Samples were illuminated with a spatial line pattern that was shifted in five phases and rotated in five orientations. The raw images were reconstructed (Gustafsson et al., 2008) into a high-resolution 3D-dataset using the Zeiss 2012 PS1 ZEN software. Reconstruction was done using default settings. Sectioning values for filtering of zero, first and second order peaks in the Fourier transformed images were 100, 83 and 83 respectively, where 100 is maximum filtering and 0 is no filtering. Noise filtering was done using a generalised Wiener filter with a regularisation parameter of 10 −6 . No correction was made for anisotropy and the results were baseline shifted. The SIM check toolbox (Ball et al., 2015) was used to evaluate the SIM reconstruction results. SIM reconstruction yielded 39.7 × 39.7 × 144 nm 3 sized voxels in a 75 × 75 × 10 µm 3 read-out volume. Imaging of one read-out volume took 18 min.
Signal-to-noise ratios (SNR) were calculated via: where μ t and σ t are respectively the mean and standard deviation of the measured track intensities, whereas μ b and σ b are respectively given by the mean and standard deviation of the background intensity.

Resolution measurement
A line was fitted through the maximum Z projection of all tracks in a field of view in FIJI (Schindelin et al., 2015). A Gaussian model was fitted to the average intensity profile perpendicular to the track at a single pixel spacing, using the nonlinear least squares algorithm in the software package R version 3.2.2 (R Core Development Team, 2011). Track diameter was defined as the FWHM (full width half maximum) calculated via: where σ is the fitted standard deviation.

Particle tracking
The fluorescent tracks produced by the alpha particles in the FNTDs consist of series of fluorescent spots in the 2D images. Track reconstruction is required for further analysis of the tracks. The fluorescent spots were used for track reconstruction using an expanded version of in-house built software. Spots in the 2D images were connected to reconstruct the tracks based on proximity and similarity in trajectory and intensity. This procedure is described in more detail in a previous publication (Kouwenberg et al., 2016). A summary of the tracking parameters used in said publication for the CLSM and SIM sets are given in Table 1. I is the intensity threshold for spot selection, r and r p are related to the spot size and drift, d p,max and d t,max are related to the maximum distance between spots for linking, R and H control the minimum number of spots per track and the amount of spots used for track averaging, θ p,max is the maximum allowed angle and w d , w i , w v and w a are the weighting factors for respectively distance, intensity, velocity and angle. D max is the maximum allowed cost for track relinking.

Range estimation
Let l be a set of trajectories in image stack A, where each trajectory contains a set of N points p i (Fig. 1). Each point can be described using its coordinates x i , y i , z i , given in voxels from the image stack origin, and intensity I i . The intensity was taken as the average of the five centre voxels of a spot whereas Fig. 1. Schematic of an alpha particle penetrating a FNTD, producing a series of fluorescent spots (p 1 .. p n ) in the CLSM and SIM images. The depth in the FNTD of a spot p i is given by f(z ' i ), and the separation between spots in the z-direction is determined by the respective step-size of the measurement. The initial trajectory of the particle upon entering the FNTD (γ x,z , γ y,z ) is determined using the first five spots of the track. The distance travelled through the FNTD by the alpha particle before the first spot is measured, is determined using this trajectory and the surface plane of the FNTD. The lateral scattering (L l,scattering ) is given by the perpendicular deviation from the initial trajectory of the track, measured at the last fluorescent spot within the track. Fluorescence intensity decay at the end of a track is illustrated by the change in the brightness of the spots. the x and y coordinates followed from intensity weighting of the same voxels. Each coordinate is given as the far corner of a voxel, that is the first voxel is given by (1, 1, 1). Reflectance imaging was used to establish the angle of the surface plane. In cases where the location of the surface plane was later located using the lengths of the measured tracks, 25% of the measured tracks were used for surface plane localisation, whereas the remaining 75% of the tracks were used for spectroscopy purposes (see Results and Discussion). Let the surface plane of image stack A obtained from the maximum reflectance be given by A (x, y, z) ≡ ψ x x + ψ y y − z = −z 0 , so that the corrected z-coordinate is given by The length L l,i of track l at point i is then given by where s x , s y are the voxel sizes in µm in the x and y direction as given by the microscopy control software. f is a function for the depth correction required for the refractive index mismatch between the immersion oil (n = 1.518) of the objective and Al 2 O 3 (n = 1.765), given by Eq. (13) in Van Elburg et al. (Van Elburg et al., 2007). The total penetration of track l follows from the length of the track and the intersection of the initial track direction with A . Because the slope of the surface plane is often small compared to the slope of the tracks, the distance between p 1 and the point of intersect can be simplified to: where χ l is the linear slope between s x x i and s z (z 1 ) · z 1 , and γ l is the linear slope between s y y i and s z (z 1 ) · z 1 , both based on the first five points. The position of the end of the track was refined using the fluorescence intensity I . Because the colour centre concentration in the FNTDs is low (G. M. , colour centre saturation inside the track is expected to occur at a Linear Energy Transfer (LET) far below the average LET of an alpha track. For samples with a good SNR, an exponential function was fitted to the fluorescence intensity I i as function of track length L l,i in accordance with the dose profile of the alpha track (Cucinotta et al., 1998): where the parameters a and b were fitted using the implementation of the Broyden-Fletcher-Goldfarb-Shanno algorithm as found in software package R version 3.2.2. The endpoint of the track was chosen at I fit = 95 % because it is theorised that colour centre saturation happens at LETs far below those of alpha particles. The drop in intensity therefore indicates the end of the alpha particle track. Residual energy deposition is the result of high-energy delta rays travelling beyond the path of the alpha particle (Scholz & Kraft, 1996). The fitting was limited to 2.5 µm from the endpoint to avoid faulty fitting due to scattering of the alpha particle within the track. For cases where the SNR was too low for accurate fitting, the endpoint was chosen, when possible, as the last point with an intensity above the mean track intensity minus the track intensity standard deviation. When either method failed, no further refinement was attempted. The detour factor is an indicator for the deviation of the track from a straight line of a particle in a material and is defined as: where L l, projected is the track length projected onto the initial track direction vector and L l, CSDA is the Continuous Slowing Down Approach (CSDA) range which is given here by the procedural track length as described by Eq. (5). The scattering is defined here as the deviation from the projected orthogonal to the projection vector (L l,scattering ). The relative scattering can then be expressed as The code framework for track following and calculation was written in a combination of the software packages R version 3.2.2 and Java version 8.

Energy calculation
Alpha particles originate in a 1.3 µm thick Am 2 O 3 layer and pass through a 2 µm layer of gold and then a 5 mm air layer before entering the FNTD. Using the approximated initial direction vector z = χ l x + γ l y, the path length through the kth layer, denoted as τ (k) l , could be estimated using the equation below, assuming that the track passes through the whole layer.
where t k is the thickness of the kth layer. This approach ignores scattering in either layer. The path length of an alpha particle can be related to its loss of energy using translation tables, which in this case were obtained from the ASTAR application by the National Institute of Standards and Technology, USA (NIST) (Berger et al., 2005). The energy loss of an alpha particle passing through a layer is then given by where e mat is a conversion function for the residual range to the corresponding energy in a material and R mat acts as the inverse of e mat and e (k−1) l is the particle's energy the moment it leaves layer k − 1 and enters the kth layer. The total energy of an alpha particle then follows from iteration of Eq. (10) for each layer including the FNTD, where τ (FNTD) l was given by the refined track length. When the depth z 0 of the surface plane could not be found, it could be retrieved by optimisation of z 0 for |e peak − e Am241 | ≈ 0, where e peak is the peak in the measured energy distribution for image stack A and e Am241 is the mean alpha energy for Am-241.

FLUKA simulation
A simulation was built in FLUKA 2011.2c.5 (Ferrari et al., 2011) to measure the energy distribution of alpha particles entering the FNTD. The collimator was modelled as the material BLACKHOLE to improve calculation speed. The source was modelled as an infinitely thin layer halfway within the americium oxide layer, which was modelled using the same density, but with gold atoms to approximate americium atoms because higher Z atoms are not available in FLUKA. The ener-

Results and discussion
To compare SIM and CLSM imaging for FNTD-based alpha particle spectroscopy, four randomly chosen areas were scanned using both modalities. Because the range of an alpha particle in matter is directly related to its energy, the surface plane for both the SIM and CLSM scans were found by matching the calculated pathlength of alpha particle tracks through the FNTD, together with the observed angle of incidence and known pathlength through the air and source, to the energy of alpha particles emitted by the source. Am-214 emits alpha particles with 2 possible energies: 5486 keV (85% chance) and 5443 keV (15% chance). The energy difference of 43 keV would not be expected to be distinguished due to the thickness of the test material and variations in angle of incidence that would degrade the fidelity of the energies impinging on the FNTD. Indeed, even conventional alpha spectrometers can be challenged by this energy separation for thick specimens. An average alpha particle energy of 5480 keV was therefore assumed. Figure 2 shows a close-up of an area imaged with CLSM and SIM, respectively. The full field of view of the scans is shown in Fig. S1.
In single optical sections nuclear fluorescent tracks appeared as round or elongated spots (Figs. 2A, C and S1A). Round spots referred to alpha particles, which propagated nominally perpendicular to the FNTD surface (i.e. the imaging plane). Spot elongation was caused by a combination of the incident angle of the alpha particle trajectory with the FNTD and the optical slice thickness of the microscope. Shallow angle tracks were identified by more elongated spots in single sections and their apparent increase in length in a maximum projection along the z axis (Figs. 2B, D and S1B, D). Spot elongation was less for SIM, as a result of the improved optical sectioning capability. SIM imaging yielded a SNR (14.4 ± 4.8) comparable to CLSM (13.9 ± 3.6).
The radial track profile was predicted to be in the order of 3 to 100 nm (Cucinotta et al., 1998). Here, we measured the radial track profile as the FWHM perpendicular to the length of the alpha tracks. As expected, the SIM images showed smaller track diameters (FWHM 209 ± 69 nm), compared to CLSM (FWHM 322 ± 19 nm). The fact that these values are somewhat higher than the theoretical resolution limits for SIM (140 nm) and CLSM (280 nm; pinhole 1 airy unit) might be explained by the large difference in wavelength between excitation (647 nm) and emission (750 nm). In addition, the anisotropic properties of the FNTD material (see below) could have negatively affected the measured track width. However, it is noteworthy that the radial track profile found in FNTDs irradiated with 3 MeV protons (LET in Al 2 O 3 : 35 keV µm -1 ) and 1094 MeV carbon ions (LET in Al 2 O 3 : 90 keV µm -1 ), imaged by stimulated emission depletion (STED) microscopy, yielded comparable track diameters of respectively 196 ± 28 nm and 240 ± 29 nm (FWHM) (Niklas et al., 2017). Note that, 3.7 MeV alpha particles, used for this work, have a LET in Al 2 O 3 of 308 keV µm -1 (Berger et al., 2005).
For SIM imaging the orientation of the FNTD on the microscope was important, due to the optically anisotropic properties of Al 2 O 3 :C,Mg (Akselrod et al., 2003). The FNTDs were cut with the 'c' optical axis along the long axis of the chip (Akselrod, personal communication). Hence the 'c' optical axis was perpendicular to the optical axis of the microscope and more than one refractive index could be observed in the imaging plane. We expected that polarisation dependent fluctuations in the fluorescence yield might adversely affect SIM image reconstruction. To test this, FNTDs were imaged at different orientations. Shadow images were observed next to the fluorescent tracks, but appeared to be absent when FNTDs were oriented at 0° (Fig. S2). The expected variation in the separation distance between the shadow images and the main fluorescent track should be dependent on the track's angle of incidence relative to the detector surface and therefore the thickness of the FNTD through which the imaging is performed. These observations still need to be compared to the theoretical separation based on track depth and crystallographic orientation of the FNTD. Nevertheless, all subsequent SIM measurements were done under conditions that eliminated the observation. In addition, we used the SIMcheck toolbox (Ball et al., 2015) to identify putative problems with the 3D-SIM reconstructions. This analysis verified that the raw and processed SIM data were of good quality and showed no evidence for common SIM reconstruction issues (Fig. S3).
Because of the increased resolution of SIM compared to CLSM, we expected SIM imaging of FNTDs to allow for more detailed studies of the tracks induced by ionising particles. Plotting the normalised fluorescence intensity as function of distance from the track resulted in the intensity profiles shown in Figure 3. Note that the visible gaps are the result of the discrete step-size in z-direction. An exponential intensity decay Fig. 3. Scatter plot of the normalised fluorescence intensity of spots as function of distance from the last detected spot (indicated by n) in their respective track. An exponential decay of fluorescence is visible for CLSM (left) and SIM (right). Note that the distance between two spots for a track perfectly perpendicular to the z-axis is given by the step-size (Fig. 1) of the image stack in the z-direction (ß0.6 µm for CLSM, ß0.18 µm for SIM), hence the visible gaps in the figures. This distance increases when the track is more slanted (Eq. (5)). The decrease in fluorescence intensity at the end of the measured tracks indicates that alpha particles come to halt within the measured tracks. The point where the alpha particle came to a halt can therefore be estimated using Eq. (6).   5. The detour factor distribution of alpha particles penetrating a FNTD as measured using CLSM (left) and SIM (right). A detour factor of 1 indicates that the particle followed a perfectly straight track through the FNTD, whereas a value close to zero indicates a large deviation from its initial trajectory upon entering the FNTD. The median values for CLSM and SIM were respectively 0.995 and 0.997. is seen at the end of the tracks for both methodologies, but is more evident with SIM. This allowed for the intensity endpoint fitting approach given in Eq. (6).
Distributions of the calculated alpha particles energies after endpoint refinement are given in Figure 4, together with the distribution obtained from FLUKA simulations. Both the distributions from CLSM and SIM match closely with the expected energy distributions, yielding standard deviations of respectively 306 and 284 keV, compared to 208 keV in the simulation. The exact uncertainty of this method due to noise and resolution for alpha particle energy measurement is unknown, because the true energy distribution is unknown. Although the FLUKA simulation serves as an indicator for the true distribution, this simulation does not account for manufacturing errors in the collimator and degradation of the source. The true energy distribution can therefore reasonably be expected to be broader than the simulation is showing. Nevertheless, for both imaging modalities the proposed method showed a bet-ter energy resolution than would likely be achievable using track width or track intensity based methods for alpha radiation (Sykora et al., 2008;Klimpki et al., 2016). It is however important to note that the method proposed in this work is only applicable for tracks ending within the FNTD, which is only the case for (very) low-energy ions, whereas the other methods can be used for all types of tracks.
The detour factor and relative scattering distributions are given in respectively Figures 5 and 6. Given a mean entrance alpha energy of approximately 3900 keV in the FNTD, NIST reports a detour factor in Al 2 O 3 of 0.984 (Berger et al., 2005). The mean detour factors for CLSM and SIM were respectively 0.993 and 0.995. As shown in Figures 5 and 6, SIM showed an almost 50% reduction in measured scattering compared to CLSM for the same experiment conditions (Fig. 6), which was most likely the result of the improved resolution of the SIM images, enabling more accurate track reconstruction. These results might be an indication that the simulation approach  (Lewis, 1950) used by NIST to estimate the spatial scattering of alpha particles overestimates said scattering of alpha particles in Al 2 O 3 . Although uncertainty evaluation of particle ranges and scattering is difficult at these low energies, stopping power uncertainties up to 10% for 100 keV alpha particles were reported (ICRU, 1993). Further research is however required to thoroughly estimate the uncertainty in scattering estimation using FNTDs before any concrete conclusions regarding the accuracy of the scattering simulations can be made. Although these results on itself are not expected to make an impact, the overestimation suggests that this method might find its use in Monte Carlo and cross-section validation for low energy ions, a domain that is difficult to validate. Validation of scattering and stopping power data for low-energy alpha particles is important for accurate microdosimetry in alpha radionuclide therapy, a field that studies the effects of alpha radiation on a small (sub-)micron scale and thereby mainly depends on simulation data (Roeske & Stinchcomb, 1997;Kvinnsland et al., 2001;Chouin et al., 2007). The low C and Mg dopant concentrations are expected to be too low to influence the scattering significantly. Figure 7 displays tracks with weak and strong scattering to illustrate that the calculated scattering could be verified via visual inspection.

Conclusions and discussion
We have shown for the first time that SIM can be used for quantitative FNTD imaging. The maximum spatial resolution was almost doubled without increasing the scan time. The scan time for SIM could be further reduced by using three instead of five rotations of the illumination pattern, however this would lower the resolution of the reconstructed image and in addition, could lead to an increase of resolution anisotropy. Currently, SIM methods are being developed using different illumination patterns, which require fewer rotations, or no rotations (Chakrova et al., 2015;Schropp et al., 2017) at all Fig. 7. Alpha particles entering the FNTD, measured using SIM, undergoing no scattering (left) and strong scattering (right). In the absence of heavy collisions between alpha particles and atoms within the FNTD, the alpha particle follows a (close to) perfectly straight path (left). A strong collision with an atom will force to the alpha particle to be deviated from its original path (right). The images of the tracks are maximum intensity projections along the y-axis after alignment of tracks with the x-axis. The relative scattering of the tracks in the left and right were respectively 0.004 and 0.29. Note the change in track intensity when the track angle changes in the right figure due to the change in illumination by the anisotropic excitation beam/spot. and this would further reduce the scan time. Although SIM image reconstruction was computationally intensive and typically took the same amount of time as the image acquisition, the prospect of Graphics Processing Unit (GPU)-based image reconstruction (Lu-Walther et al., 2015;Müller et al., 2017) is expected to make real-time reconstruction possible in the near future. The proposed method for particle energy estimation based on track length showed an excellent energy resolution, which is required for proper alpha dosimetry. No significant improvement in energy resolution with SIM could be observed compared to CLSM. A significant difference was however visible in the relative spatial scattering of alpha track measurements between the two modalities. The difference in detour factor obtained using SIM versus the numbers obtained from NIST could be an indication that the current data from NIST might be overestimating the scattering of alpha particles in Al 2 O 3 and that revision of the current cross section (a measure for interaction probability) might be necessary, but thorough determination of the uncertainty in scattering estimation using FNTDs is required before concrete conclusions can be made regarding the current NIST data. From these results, it is expected that the high resolution of SIM allows for more complex analyses of particle interactions, which can be of great value of the validation of the Monte Carlo simulation code and interaction cross section data. Because ion radiation therapy depends heavily on the accuracy of these simulation codes, the combination of SIM and FNTD imaging might find more applications in this field. Stimulated Emission Depletion (STED) was the only other super-resolution technique applied for FNTDs, but has yet to find its way to wider applications in part due to the long scanning time and technical hardships (Greilich et al., 2013;Niklas et al., 2017). SIM overcomes these challenges by offering super-resolution imaging at scanning speeds compatible with large-scale commercial read-out. The increased resolution can also be used for better detection of secondary particles in clinical carbon beams. Due to the low fluorescence intensity and often large angle of incidence of these secondary particles, their numbers are often underestimated when using CLSM for FNTD read-out (Klimpki et al., 2016). SIM is therefore considered an excellent competitor for high-resolution FNTD read-out, especially with the possibility of fast image acquisition and real-time SIM reconstruction in combination with automatic surface plane detection (Akselrod et al., 2014). The FNTD/SIM combination paves the way for a new type of alpha dosimeter, both for research and personal dosimetry.

Supporting Information
Additional Supporting information may be found in the online version of this article at the publisher's website:  Fig. 2. The CLSM images were cropped to match the field of view of the SIM data (75 µm × 75 µm). Scale bars are 10 micrometre. Fig. S2. FNTD orientation-dependent shadow tracks. Maximum projections of SIM images obtained at the indicated FNTD sample orientation. Shadow tracks are indicated by arrows. CLSM image is shown for comparison. Scale bar is 2 µm. Graph displays the average line plot across a single track (red line) for FNTD at 260°(indicated by *). The blue and black lines are the average line plots for tracks imaged by CLSM and SIM (FNTD at 0°), respectively. The resolution of CLSM is too low to detect shadow tracks. Fig. S3. SIM check was performed on both raw and reconstructed data. (A) Cropped area of raw and reconstructed data. The full image was used for SIMcheck. (B) modulation contrast to noise ratio (MCNR) of raw and reconstructed data. MCNR is depicted on a colour scale, where orange to yellow pixels indicate moderate to good modulation. (C) Channels intensity profile of raw data, the total intensity variation measured is 30.5%. (D) Fourier transform analysis of reconstructed data of both a lateral (xy) and a radial view, no clear artefacts were observed. (E) Spherical aberration mismatch analysis minimum values (black line) and maximum values (grey line) are depicted a z-minimum variation of 0.0523 was measured. (F) Reconstructed intensity profile on a linear (black) and a logarithmic (grey) scale, a min-to-max intensity ratio of 3.30 was measured.