Fig 1.
Bone structure, THG microscopy setup and principle.
(a) simplified view of the osteocyte network in cortical bone (left) and the corresponding cavities in the bone matrix forming the lacuno-canalicular network (LCN). (b) Microscopy setup: a pulsed infrared 1180 nm beam is scanned and focussed on the sample, and the THG signal is collected in transmission onto a photomultiplier tube (PMT) using a 1.4NA condenser. (c) simplified Jablonski diagrams for single-photon excited fluorescence used in our confocal microscopy experiments (1PEF, left), and third-harmonic generation (THG, right). The THG signal is created at exactly one third of the excitation wavelength. Dotted line, virtual energy level. (d) the coherent buildup of the THG signal and the excitation phase shift of the excitation beam at the focal spot result in a non-zero signal only in the presence of optical heterogeneities within the focal volume, as created by canaliculi in the mineralized collagen bone matrix. This signal is strongly dependent on the canaliculi diameter and orientation.
Fig 2.
Comparison between confocal fluorescence microscopy (a,b,c) and THG imaging (d,e,f) of two different samples. (a,d) 2D composite images of full transverse cross-sections of a mouse femur. Regions shown in (b,e) are indicated by a green box in (a,d). Regions in (c,f) are indicated by a yellow box in (b,e). Red stars in (d) indicate the presence of air bubbles due to localized heating by absorption of the excitation light in the presence of polishing residues. Scale bar 500 μm (a,d), 20 μm (b,e) and 5 μm (c,f).
Fig 3.
(a) Zoom of a region of interest of a THG image at 3.6 μm from the surface. (b) Corresponding image after vesselness filtering. (c) Intensity projection of the full image stack along Z. (d) Maximum intensity projection of the result of the skeletonization of the filtered data. Scale bar 20 μm.
Table 1.
Uncertainty estimation for the proposed image processing methods.
Two parameters are used to characterize the canalicular network: the density of canalicular porosity, and the volumetric density of connections. Skeletons obtained with different processing methods are compared with a reference skeleton that was manually corrected to match the network visible on the original THG image. In each case, the full volume is used, or a part corresponding to the position of the lacuna is removed from the analysis. % uncertainty indicated in parenthesis.
Fig 4.
Anatomical variability of the LCN.
THG images of the LCN of a sample from the control (left), synchro (center) and flight (right) groups in the M, L, A and P regions (from top to bottom). The inset shows the module of the Fourier-transformed images after vesseleness filtering.Scale bar 30 μm.
Table 2.
LCN characteristics.
Fig 5.
Bone volume fraction of canalicular porosity (Ca.V/TV), density of connections and density of lacunae for Control, Flight and Synchro groups, for all four quadrants (light squares: lateral;dark squares: medial;diamonds: anterior/posterior). Intra-individual inter-individual variations within one group are large compared to measurement uncertainties and to the differences between groups.
Table 3.
Theoretical group size required to measure porosity network variations in the present study.
Averages over the four quadrants were used for this calculation but little variations are observed when considering one given anatomical region.