E-mail Alert
RSS
| Citation: |
Wu ST, Yang ZC, Ma CG, Zhang X, Mi C et al. Deep learning enhanced NIR-II volumetric imaging of whole mice vasculature. Opto-Electron Adv 6, 220105 (2023). doi: 10.29026/oea.2023.220105
|
Deep learning enhanced NIR-II volumetric imaging of whole mice vasculature
-
Abstract
Fluorescence imaging through the second near-infrared window (NIR-II,1000–1700 nm) allows in-depth imaging. However, current imaging systems use wide-field illumination and can only provide low-contrast 2D information, without depth resolution. Here, we systematically apply a light-sheet illumination, a time-gated detection, and a deep-learning algorithm to yield high-contrast high-resolution volumetric images. To achieve a large FoV (field of view) and minimize the scattering effect, we generate a light sheet as thin as 100.5 μm with a Rayleigh length of 8 mm to yield an axial resolution of 220 µm. To further suppress the background, we time-gate to only detect long lifetime luminescence achieving a high contrast of up to 0.45 Ιcontrast. To enhance the resolution, we develop an algorithm based on profile protrusions detection and a deep neural network and distinguish vasculature from a low-contrast area of 0.07 Ιcontrast to resolve the 100 μm small vessels. The system can rapidly scan a volume of view of 75 × 55 × 20 mm3 and collect 750 images within 6 mins. By adding a scattering-based modality to acquire the 3D surface profile of the mice skin, we reveal the whole volumetric vasculature network with clear depth resolution within more than 1 mm from the skin. High-contrast large-scale 3D animal imaging helps us expand a new dimension in NIR-II imaging.-
Keywords:
- NIR-II fluorescence /
- time-gated /
- light sheet illumination /
- deep learning /
- vessel enhancement /
- 3D imaging
-
-
References
[1] Ntziachristos V, Bremer C, Weissleder R. Fluorescence imaging with near-infrared light: new technological advances that enable in vivo molecular imaging. Eur Radiol 13, 195–208 (2003). doi: 10.1007/s00330-002-1524-x [2] Li BH, Zhao MY, Feng LS, Dou CR, Ding SW et al. Organic NIR-II molecule with long blood half-life for in vivo dynamic vascular imaging. Nat Commun 11, 3102 (2020). doi: 10.1038/s41467-020-16924-z [3] Zhu XY, Liu X, Zhang HX, Zhao MY, Pei P et al. High‐fidelity NIR‐II multiplexed lifetime bioimaging with bright double interfaced lanthanide nanoparticles. Angew Chem Int Ed 60, 23545–23551 (2021). doi: 10.1002/anie.202108124 [4] Du HT, Wan H, Dai HJ. Recent advances in development of NIR-II fluorescent agents. In Benayas A, Hemmer E, Hong GS, Jaque D. Near Infrared-Emitting Nanoparticles for Biomedical Applications 83–101 (Springer, 2020);http://doi.org/10.1007/978-3-030-32036-2_5. [5] Welsher K, Sherlock SP, Dai HJ. Deep-tissue anatomical imaging of mice using carbon nanotube fluorophores in the second near-infrared window. Proc Natl Acad Sci USA 108, 8943–8948 (2011). doi: 10.1073/pnas.1014501108 [6] Hong GS, Lee JC, Robinson JT, Raaz U, Xie LM et al. Multifunctional in vivo vascular imaging using near-infrared II fluorescence. Nat Med 18, 1841–1846 (2012). doi: 10.1038/nm.2995 [7] Hong GS, Diao S, Chang JL, Antaris AL, Chen CX et al. Through-skull fluorescence imaging of the brain in a new near-infrared window. Nat Photonics 8, 723–730 (2014). doi: 10.1038/nphoton.2014.166 [8] Yu WB, Guo B, Zhang HQ, Zhou J, Yu XM et al. NIR-II fluorescence in vivo confocal microscopy with aggregation-induced emission dots. Sci Bull 64, 410–416 (2019). doi: 10.1016/j.scib.2019.02.019 [9] Liu YF, Gou HL, Huang X, Zhang GY, Xi K et al. Rational synthesis of highly efficient ultra-narrow red-emitting carbon quantum dots for NIR-II two-photon bioimaging. Nanoscale 12, 1589–1601 (2020). doi: 10.1039/C9NR09524A [10] Wang FF, Ma ZR, Zhong YT, Salazar F, Xu C et al. In vivo NIR-II structured-illumination light-sheet microscopy. Proc Natl Acad Sci USA 118, e2023888118 (2021). doi: 10.1073/pnas.2023888118 [11] Lu LF, Li BH, Ding SW, Fan Y, Wang SF et al. NIR-II bioluminescence for in vivo high contrast imaging and in situ ATP-mediated metastases tracing. Nat Commun 11, 4192 (2020). doi: 10.1038/s41467-020-18051-1 [12] Liao JL, Yin YX, Yu J, Zhang RL, Wu T et al. Depth-resolved NIR-II fluorescence mesoscope. Biomed Opt Express 11, 2366–2372 (2020). doi: 10.1364/BOE.386692 [13] Wang FF, Wan H, Ma ZR, Zhong YT, Sun QC et al. Light-sheet microscopy in the near-infrared II window. Nat Methods 16, 545–552 (2019). doi: 10.1038/s41592-019-0398-7 [14] Girkin JM, Carvalho MT. The light-sheet microscopy revolution. J Opt 20, 053002 (2018). doi: 10.1088/2040-8986/aab58a [15] Pampaloni F, Ansari N, Girard P, Stelzer EHK. Light sheet-based fluorescence microscopy (LSFM) reduces phototoxic effects and provides new means for the modern life sciences. Proc SPIE 8086, 80860Y (2011). doi: 10.1117/12.889443 [16] Jin DY, Piper JA. Time-gated luminescence microscopy allowing direct visual inspection of lanthanide-stained microorganisms in background-free condition. Anal Chem 83, 2294–2300 (2011). doi: 10.1021/ac103207r [17] Zheng XL, Zhu XJ, Lu YQ, Zhao JB, Feng W et al. High-contrast visualization of upconversion luminescence in mice using time-gating approach. Anal Chem 88, 3449–3454 (2016). doi: 10.1021/acs.analchem.5b04626 [18] Ortgies DH, Tan ML, Ximendes EC, del Rosal B, Hu J et al. Lifetime-encoded infrared-emitting nanoparticles for in vivo multiplexed imaging. ACS Nano 12, 4362–4368 (2018). doi: 10.1021/acsnano.7b09189 [19] Nimmegeers B, Cosaert E, Carbonati T, Meroni D, Poelman D. Synthesis and characterization of GdVO4: Nd near-infrared phosphors for optical time-gated in vivo imaging. Materials 13, 3564 (2020). doi: 10.3390/ma13163564 [20] Morimoto K, Ardelean A, Wu ML, Ulku AC, Antolovic IMM et al. Megapixel time-gated SPAD image sensor for scientific imaging applications. Proc SPIE 11654, 116540U (2021). doi: 10.1117/12.2582444 [21] Connally RE, Piper JA. Time‐gated luminescence microscopy. Ann N Y Acad Sci 1130, 106–116 (2008). doi: 10.1196/annals.1430.032 [22] Cubeddu R, Comelli D, D'Andrea C, Taroni P, Valentini G. Time-resolved fluorescence imaging in biology and medicine. J Phys D Appl Phys 35, R61–R76 (2002). doi: 10.1088/0022-3727/35/9/201 [23] Marriott G, Clegg RM, Arndt-Jovin DJ, Jovin TM. Time resolved imaging microscopy. Phosphorescence and delayed fluorescence imaging. Biophys J 60, 1374–1387 (1991). doi: 10.1016/S0006-3495(91)82175-0 [24] Cheng SM, Shen B, Yuan W, Zhou XB, Liu QY et al. Time-gated ratiometric detection with the same working wavelength to minimize the interferences from photon attenuation for accurate in vivo detection. ACS Cent Sci 5, 299–307 (2019). doi: 10.1021/acscentsci.8b00763 [25] Yin XX, Ng BWH, He J, Zhang YC, Abbott D. Accurate image analysis of the retina using hessian matrix and binarisation of thresholded entropy with application of texture mapping. PLoS One 9, e95943 (2014). doi: 10.1371/journal.pone.0095943 [26] Christensen KA, Morris MD. Hyperspectral Raman microscopic imaging using Powell lens line illumination. Appl Spectrosc 52, 1145–1147 (1998). doi: 10.1366/0003702981945138 [27] Iandola F, Moskewicz M, Karayev S, Girshick R, Darrell T et al. DenseNet: implementing efficient convnet descriptor pyramids. arXiv: 1404.1869 (2014). [28] Vogt N. Hybrid volumetric calcium imaging. Nat Methods 16, 461 (2019). doi: 10.1038/s41592-019-0442-7 [29] Yamauchi K, Yang M, Jiang P, Xu MX, Yamamoto N et al. Development of real-time subcellular dynamic multicolor imaging of cancer-cell trafficking in live mice with a variable-magnification whole-mouse imaging system. Cancer Res 66, 4208–4214 (2006). doi: 10.1158/0008-5472.CAN-05-3927 [30] Ulyanov SS, Ganilova YA. Speckle-microscopy for blood flow measurements in the smallest microvessel. Proc SPIE 5139, 206–215 (2003). doi: 10.1117/12.500259 [31] Chao M, Ming G, Xun Z, Liu Y, Sitong W et al. High Spatial and Temporal Resolution NIR-IIb Gastrointestinal Imaging in Mic. Nano Lett 22, 2793–2800 (2022). doi: 10.1021/acs.nanolett.1c04909 [32] Feng Z, Tang T, Wu TX, Yu XM, Zhang YH et al. Perfecting and extending the near-infrared imaging window. Light Sci Appl 10, 197 (2021). doi: 10.1038/s41377-021-00628-0 -
Supplementary Information
Supplementary information for Deep learning enhanced NIR-II volumetric imaging of whole mice vasculature 
-
Access History
Figures(6)
Article Metrics
Export File
Citation
Wu ST, Yang ZC, Ma CG, Zhang X, Mi C et al. Deep learning enhanced NIR-II volumetric imaging of whole mice vasculature. Opto-Electron Adv 6, 220105 (2023). doi: 10.29026/oea.2023.220105
Format
Content
DownLoad:
-
Figure 1.
Trade-offs among lateral resolution, depth resolution, speed, and volume of view by using the three light illumination strategies, wide-field11, spot scanning12, and light-sheet scanning13, in NIR-II fluorescent volumetric imaging (log scale).
-
Figure 2.
Adjustable large-FoV light-sheet illumination system for high-contrast whole mice imaging. (a) light-sheet illumination system with adjustable thickness from 183.1 to 100.5 μm and Rayleigh length from 8 to 26 mm, formed by a 30 mm focus lens (Thorlab-1700-B), a 16 mm focus 0.79 NA aspherical lens (Thorlab-ACL254U-B), and a 10 × 0.3 NA objective lens. (b) The illumination of light sheet through the whole mice. (c) Comparison of a wide-field imaging result and a top view of light-sheet scanning result at a distance of 100 μm from the mice. (d) Line profile analysis of (c).
-
Figure 3.
Phantom experiment to evaluate the scattering effect under the large-FoV light-sheet illumination. (a) Schematic illustration of light-sheet scanning across the phantom matrix. (b) Scattering simulation using silicone phantom containing lanthanide doped NIR-II fluorescent nanoparticles and TiO2 nanoparticles. (c) Letter patterns stained by lanthanide-doped NIR-II fluorescent nanoparticles and embedded in silicone phantom. (d) Reconstruction of the letter patterns by light-sheet scanning. The diameter of the container is 5.5 cm. The step of light-sheet scanning between each layer is 1 mm.
-
Figure 4.
Time-gated light-sheet imaging system. (a) TG imaging system. (b) Comparison results of TG light-sheet imaging and CW filter-based light-sheet imaging with different average laser power. (c) Icontrast of TG mode and CW mode with different average laser power. (d) Range of intensity of TG mode and CW mode with different average laser power.
-
Figure 5.
Comparison schematics and results of vessel enhancement processes by using Hessian matrix v.s. DenseNet deep learning algorithm based on protuberance detection. (a) Raw image. (b) Hessian matrix enhancement. (c) DenseNet enhancement result. (d) Intensity discriminator. (e–h) are the profiles of white dotted lines in (a–d). (i) Training process of Hessian algorithm and our algorithm.
-
Figure 6.
The depth information of the high contrast images of the whole mice skin and blood vessel revealed by the time-gated light-sheet NIR-II volumetric imaging system enhanced by deep learning algorithm. (a) The 3D reconstruction image of mice skin and vessel networks. (b) Luminescent intensity map. (c) The height map of the mice’s blood vessels. (d) The height map of the mice’s skin. (e) Vessels with colors that represent their depth under the skin. (f, g) Local details of the same position in (a) and (c). (h, i) Intensity profiles of vessels in different depths. (h) A magnified part of (b). Intensity profiles of lines in (i) with depths of 500 μm, 750 μm, and 1000 μm.
