The strengths and weaknesses of the optoacoustic method: is combining light and sound always beneficial?

Transcription

1 The strengths and weaknesses of the optoacoustic method: is combining light and sound always beneficial? DANIEL RAZANSKY Director, Laboratory of Optoacoustics and Molecular Imaging Problèmes Inverses et Imagerie Institut Henri Poincaré, Paris Feb. 12, 2014 FACULTY OF MEDICINE Intravascular Imaging Fluorescence Molecular Tomography Institute for Biological and Medical Imaging Image processing Optoacoustic Imaging IBMI Chemistry, Molecular Probes Intraoperative Imaging FACULTY OF MEDICINE TECHNICAL UNIVERSITY OF MUNICH 1

2 Imaging depth 0 0.1mm 1mm 1cm 10cm MRI Plate reader Planar TIRF Confocal Multi photon Mesoscopic imaging Ultrasound Resolution 0.1mm 1mm 100mm 1mm PET/CT What is so good about imaging with light? (+) Rich intrinsic molecular and functional contrast (+) Versatility of contrast approaches nanoparticles, fluorescent proteins, reporter genes, cell, receptor targeting, enzyme activation (+) Low entry and maintenance cost (+) Simple to use (+) Versatility of excitation mechanisms multispectral, life time, FRET, MSOT (+) Safe, non ionizing 2

3 Optical Imaging Ranges Microscopy range (0 1 mm) Mesoscopic range (1 5 mm) Macroscopic range (> 5mm) Fully diffusive Optoacoustic Imaging Light Sensor r Tissue Nd:YAG + OPO Optoacoustic imaging utilizes absorption of ultrashort pulses of light in order to generate ultrasonic responses from biological tissues. Therefore, imaging contrast is optical but spatial resolution is determined by ultrasonic diffraction. 3

4 Optoacoustic imaging is a complex inverse problem An accurate optoacoustic forward model has to accommodate for a number of parameters, e.g. multiple chromophores in tissue, light transport at multiple wavelengths, acoustic wave propagation effects, detectors properties DETECTED PRESSURE VARIATIONS MAP OF ABSORBED ENERGY MAP OF OPTICAL ABSORPTION COEFFICIENT MAPS OF CHROMOPHORE CONCENTRATIONS Headache #1: Acoustic inversion 4

5 Optoacoustic Image formation Tomography (1994) (OAT) Tissue phantoms Detected pressure variation Source (laser absorption) Optoacoustic source equation : Spherical Radon (spherical mean) transform RA Kruger et al, Med Phys, 1994, 1995 Y Xu et al, IEEE Trans Med Imag, 2002 In vivo mouse images (2002) In vivo mouse brain X Wang et al, Nature Biotech,

6 Tomographic detection using Fabry Perot resonances 9 mm depth Laufer et al. Proc. SPIE (2012) 0 mm Zhang et al. Phys. Med. Biol. (2009) Scanning optical resolution microscopy The excitation light is focused HbT Hu et al. Opt. Lett. (2011) Hu et al. Opt. Express., (2009) 6

7 Ultrasound linear array detection Kim et al., Biomed. Opt. Exp., 1, (2010) Niederhauser et al., IEEE Trans. Med. Imag. 24 (2005) Optoacoustic endoscope Yang et al., Nat. Med. 2012, 18,

8 Three dimensional tomography Brecht et al., JBO (2009) Ultrasound imaging in 1950 s The "pan scanner of Douglas Howry in University of Colorado First US images of human leg,

9 Tomographic detection approaches Lutzweiler & Razansky, Sensors, (2013) Filtered back projection Reconstructed source Inversion in 2D Detected pressure variation Model based inversion Forward model (Poisson type integral) Discretized ed model Inversion (LSQR) Inversion (SVD) Rosenthal et al., IEEE Trans. Med. Imag. 29(6), Dean Ben et al., IEEE Trans. Med. Imag. 31(10), Rosenthal et al., IEEE Trans. Med. Imag. 31(7),

10 Headache #2: Light transport in tissues Light attenuation in deep tissues Schematics of turbid tissue mimicking phantom Raw optoacoustic image H=µ a (r)u(r) Razansky et al, Med. Phys.,

11 Correction using numerical FEM solution Initial optoacoustic image Light distribution Normalized image Iteration 1 Light diffusion equation Iteration 3 Iterative normalization Iteration 10 Jetzfellner et al., Appl. Phys. Lett. 2009, Blind separation approach Our function H=µ a U is a product of two distinct functions Taking the log of the signal, we get a sum: log[µ a ]+log[u] Can one separate the signal into its components? Rosenthal et al., IEEE Trans. Med. Imag. 28(12),

12 Sparse representation Fluence is a smooth and slow function Absorption coefficient is quickly varying with strong local properties Jean Baptiste Joseph Fourier Alfred Haar Can be fitted dby a small number of Will be represented by 2D discrete smooth functions therefore will be Haar wavelet basis that can represented by polinomials or 2D successfully represent discrete Fourier basis discontinuities Rosenthal et al., IEEE Trans. Med. Imag. 28(12), Finding the coefficients The problem now becomes a problem of finding a minimal amount of coefficients c n and d m that can sparsely represent the optoacoustic image : Rosenthal et al., IEEE Trans. Med. Imag. 28(12),

13 Experimental results Decomposed light fluence Photograph of cylindrical tissue mimicking phantom µ 1 1 a =0.2cm 1 µ s =10cm 1. Two insertions have µ a =0.6cm 1 Initial optoacoustic image Decomposed optical absorption Rosenthal et al., IEEE Trans. Med. Imag. 28(12), Headache #3: Extracting chromophore distribution 13

14 Multi Spectral Optoacoustic Tomography (MSOT) 860nm. 700nm c a M b M b F M F c Hb HbO 2 IRdye Inversion linear regression, PCA, ICA Glatz et al., Opt. Exp., 19 (2011) Linear regression based inversion (Moore Penrose pseudoinverse) Headache #4: Modeling imperfections and image artifacts 14

15 Statistical correction for strong acoustic heterogeneities Standard discretized back projection Probability that the generated signal arrived via reflection Probability for detection of direct propagation Corrected version of the back projection algorithm Dean Ben et al., IEEE Trans. Med. Imag. 30(2), Dean Ben et al., Applied Physics Letters, 98(17), Out-of-plane artifacts Cylindrically-focused detector Microsphere target Uncorrected Deconvolved Rosenthal et al., Med. Phys., 2011 Buehler et al., in review,

16 Optoacoustics with linear (array) scanning Niederhauser et al., IEEE Trans. Med. Imag., 2005 Poor tomographic coverage US image of a mouse superimposed with distribution of Hb (in color) resolved by optoacoustics US image of a mouse tumor superimposed with distribution of carbon nanopraticles (in color) resolved by optoacoustics 40 o Courtesy of Visualsonics Inc. De la Zerda et al., Nature Nanotech Limited view tomographic problem 240 o 170 o 120 o Model-based unregularized Model-based PLSQR Model-based TGSVD Back-projection Buehler et al., Med. Phys

17 Motion clustering for deblurring Taruttis et al., JBO, 2012 k-means clustering algorithms seek to minimize the sum of the distances from each data point to the mean of the cluster it is assigned to. Examples of in vivo imaging results 17

18 Whole body video rate MSOT scanner Razansky et al., Nature Prot., 6(8), 2011 Buehler et al., Optics Letters, 35(14), 2010 Blood oxygenation in tumors day 6 day 10 day 13 cryo Herzog et al., Radiology,

19 Whole body Kidney perfusion video rate imaging MSOT with imaging ICG Razansky et al., Nature Prot., 6(8), 2011 Buehler et al., Optics Letters, 2010 Buehler et al., Optics Letters, 35(14), 2010 Pharmacokynetics and metabolism CW800 NIR dye (Li Cor Biosciences) 20nmol tail vein injection peak absorption at 774nm signal level (norm malized) 100% 80% 60% 40% 20% 0% injection time after injection [minutes] left renal pelvis left renal cortex Taruttis et al., PLoS ONE,

20 Handheld real time tomographic imaging Buehler et al., Opt. Lett Hand held real time 3D scanner fibre bundle transducer elements 512 simultanesouly acquired channels Dean Ben et al.,., IEEE Trans. Med. Imag., 2013 Dean Ben et al.,., Opt. Exp.,

21 Real time tracking of deep human vasculature Dean Ben et al, IEEE Trans. Med. Imag., 2013 Dean Ben & Razansky, Opt. Exp., 2013 Real-time rendering of 3D images is enabled with graphics processing unit (GPU)-based reconstruction 5D optoacoustic imaging (1) Volumetric (3D) real time (4D) spectrally enriched (5D) tomography Noninvasive real-time 3D tracking of probe biodistribution in mouse brain following tail vein injection of ICG Dean Ben & Razansky, Light Science and Applications, 3, e137 (2014) 21

22 Volumetric real time spectrally enriched tomography 5D optoacoustic imaging (2) Dean Ben & Razansky, Light Science and Applications, 3, e137 (2014) Editor in Chief Vasilis Ntziachristos (Munich, Germany) Section Editors Stanislav Emelianov (US) Sanjiv Sam Gambhir (US) Daniel Razansky (Germany) Advances in Technology Nanoparticles and Probes Imaging Applications ISSN , 2013, Volume Editorial Board Mark A. Anastasio, Bertrand Audoin, Paul C. Beard, Rinat O. Esenaliev, Martin Frenz, Christ Glorieux, Pai Chi Li, Matthew O'Donnell, Alexander A. Oraevsky, Wiendelt Steenbergen, Xueding Wang, Roger James Zemp, Vladimir P. Zharov, Quing Zhu, US France UK US Switzerland Belgium Taiwan US US The Netherlands US Canada US US 22