Development of novel methods for computational microscopy
Thanks to the affiliation of the University of Chicago with the Marine Biological Laboratory in Woods Hole, we have developed a number of collaborations that apply our expertise in inverse problems to the development of new computational microscopy approaches. One strand, in collaboration with Hari Shroff of NIH, involves developing novel approaches to modeling and fusing multi-view data in light-sheet microscopy, including a three-lens, three-view system and a mirror-based system to create orthogonal light sheets and capture four views of the sample. A second strand involves developing novel approaches to estimate the orientation of molecules that have been tagged rigidly with anisotropic fluorophores like GFP. At present, we are seeking to merge the two strands by developing novel multiview, light-sheet approaches to imaging of molecular orientation.

  1. Wu, Y., Chandris, P., Winter, P. W., Kim, E., Jaumouille, V., Kumar, A., Guo, M., Leung, J., Smith, C., Suarez, I., Upadhyaya, A., Liu, H. Waterman, C., Ramamurthi, K., La Riviere, P.,Shroff, H., “Simultaneous multi-view capture and fusion improves spatial resolution in wide-field and light-sheet microscopy,” Optica, 3: 897-910, (Featured article by OSA, MBL, and UChicago News.) PMCID: PMC5066810.
  2. Mehta, S., McQuilken, M., La Riviere P., Occhipinti, P., Verma, A., Oldenbourg A.R., Gladfelter A. S., Tani T., “Dissection of molecular assembly dynamics by tracking orientation and position of single molecules in live cells,” PNAS 2016 113 (42) E6352-E6361; doi:10.1073/pnas.1607674113 (2016). (Featured article by MBL, UChicago News, and Gizmodo.) PMCID: PMC5081662.
  3. Day K.J., La Rivière P.J.,Chandler T., Bindokas V.P., Ferrier N.J. and Glick B.S., “Improved deconvolution of very weak confocal signals,” F1000Research 2017, 6:787 (doi: 10.12688/f1000research.11773.2). PMCID: PMC5553083.
  4. Wu, Y., Kumar, A., Smith, C., Ardiel, E., Chandris, P., Christensen, R., Rey-Suarez, I., Guo, M., Vishwasrao, H.D., Chen, J., Tang, J., Upadhyaya, A., La Riviere, P. J., and Shroff, H., “Reflective imaging improves resolution, speed, and collection efficiency in light sheet microscopy,” Nature Communications, vol. 8, no. 1, p. 1452, 2017. PMCID: PMC5682293.

The images obtained by the combination of the new coverslip and computer algorithms show clearer views of small structures. [Paper 4 above. ]

In this diagram, you can see how the mirrored coverslip allows for four simultaneous views. [Paper 4 above]

Development of approaches for x-ray histology at cellular resolution
Histological studies providing cellular insights into tissue architecture have been central to biological discovery and remain clinically invaluable today. Extending histology to three dimensions would be transformational for research and diagnostics. However, three-dimensional histology is impractical using current techniques. We have customized sample preparation, synchrotron X-ray tomographic parameters, and three-dimensional image analysis to allow for complete histological phenotyping using whole larval and juvenile zebrafish. The resulting digital zebrafish can be virtually sectioned and visualized in any plane. Whole-animal reconstructions at subcellular resolution also enable computational characterization of the 50 zebrafish nervous system by region-specific cell nuclei detection and quantitative assessment of individual phenotypic variation. Three-dimensional histological phenotyping has potential use in genetic and chemical screens, and in clinical and toxicological tissue diagnostics. We are exploring the use of multi-stain, multi-energy techniques, allowing for “color” x-ray histology.

  1. Ding, Y., Daniel J. Vanselow, D.J., Yakovlev, M.A., Katz, S. R., Lin, A. Y., Clark, D. P., Vargas, P., Xin, X., Copper, J.E., Canfield, V.A., Ang, K.C., Wang, Y., Xiao, X., De Carlo, F., van Rossum, D.B., La Riviere, P. J.,and Cheng, K. C., “Three-Dimensional Histology of Whole Zebrafish by Sub-Micron Synchrotron X-ray Micro-Tomography,” submitted to Science (2018). BiorXiv.
  2. La Rivière, P. J.,Clark, D., Rojek, A., Vargas, P., Xiao, X., De Carlo, F., Kindlmann, G., and Cheng, K. “Optimizing synchrotron micro-CT for high-throughput phenotyping of zebrafish,” Proc. SPIE7804, 78040M (2010); doi:10.1117/12.860783.

Flowchart of the color x-ray histology process.

Development of algorithms and novel imaging geometries for X-ray fluorescence tomography
We have worked for several years to develop new image reconstruction algorithms and new image acquisition strategies for X-ray fluorescence computed tomography (XFCT). X-ray fluorescence computed tomography (XFCT) is an emerging imaging modality that allows for the reconstruction of the distribution of nonradioactive elements (mostly metals) within a sample from measurements of fluorescence X-rays produced by irradiation of the sample. Many endogenous metals and metal ions, such as Fe, Cu, and Zn, play critical roles in signal transduction and reaction catalysis, while others (Hg, Cd, Pb) are quite toxic even in trace quantities.

XFCT is a stimulated emission tomography modality and thus correction for attenuation of the incident and fluorescence photons is essential if accurate images are to be obtained. We have developed three different attenuation-correction algorithms for XFCT, each more general and powerful than the last [1]. We have also proposed new data acquisition schemes to minimize attenuation [2].

In recent years, in collaboration with Ling-Jian Meng at UIUC, we have begun to explore radically different ways of measuring XFCT data. Our insight was to exploit the fact that X-ray fluorescence is a stimulated emission modality to perform selective illumination coupled with detection by pixelated cameras through collimating apertures to perform direct imaging without need for tomographic image reconstruction.

  1. La Rivière, P. J.and Vargas, P. A., “Monotonic penalized-likelihood image reconstruction for X-ray fluorescence computed tomography,” IEEE Trans. Med. Imag.,25, 1117–1129, 2006.
  2. La Rivière, P. J., Vargas, P. A., Newville, M., and Sutton, S., “Reduced-scan schemes for X-ray fluorescence computed tomography,” IEEE Trans. Nucl. Sci,54, pp. 1535-1542, 2007.
  3. Meng, L. J., Li, N., and La Rivière, P. J., “X-ray Fluorescence Tomography Using Emission Tomography Systems,” IEEE Trans. Nucl. Sci., 58: pp, 3359-3369, PMCID: PMC3251222
  4. Fu, G., Meng, L.-J., Eng, P., Newville, M., Vargas, P., and La Riviere, P.J.,“Experimentaldemonstration of novel imaging geometries for x-ray fluorescence computed tomography” Medical Physics, 40, pp. 061903 (11 pages), 2013. PMCID: PMC3663849.
  5. Groll, A., George, J., Vargas, P., La Rivière, P.J.and Meng, L.-J., “Element Mapping in Organic Samples Utilizing a Benchtop X-Ray Fluorescence Emission Tomography (XFET) System,” IEEE Trans. Nucl. Sci., 62 (5), Oct 2015. PMCID:PMC4686274

XFCT images of an osmium-stained zebrafish, with and without proper attenuation correction. [Paper 4 above]

Development of sinogram restoration strategies for computed tomography
We have a long-standing research program in sinogram restoration for computed tomography.My group has developed a novel strategy for CT data processing that entails application of a statistically principled penalized-likelihood estimation of the ideal, low-noise line integrals needed for image reconstruction from the noisy degraded transmission measurements [1]. The goal was to achieve some of the benefit of fully iterative reconstruction at a fraction of the computational cost. We collaborated with Philips Research and Development on this work [2] and some ideas were incorporated into Philips’ clinical iDose dose-reduction scheme. We showed that for quadratic penalties, we could achieve very similar performance to fully iterative reconstruction [3] and that edge-preserving penalties provided an improvement over quadratic penalties but could not match the performance of fully iterative reconstruction with edge preservation [4]. While computational power now supports the use of fully iterative in routine CT scans, there is still a role for the ideas of sinogram restoration in dynamic and spectral CT.

  1. La Rivière, P. J., and Bian, J., and Vargas, P. A., “Penalized-likelihood sinogram restoration for computed tomography,” IEEE Trans. Med. Imag.,25, 1022–1036, 2006.
  2. Forthmann, P., Kohler, T., Defrise, M., andLa Rivière, P. J.,“Comparing implementations of penalized weighted least squares sinogram restoration,” Med. Phys. 37, pp. 5929-5938, 2010. PMCID: PMC2988831
  3. Vargas, P. A., andLa Rivière, P. J., “Comparison of image-domain and sinogram-domain penalized likelihood image reconstruction estimators, ” Medical Physics, 38, pp. 4811-4823, 2011. PMCID: PMC3172866.
  4. Little K.J. and La Rivière P. J., “Sinogram restoration in computed tomography with an edge-preserving penalty,” Med Phys. 2015 Mar;42(3):1307. doi: 10.1118/1.4907968. PubMed PMCID: PMC4344471.

Results of applying sinogram restoration to data corrupted with beam hardening alone (labeled BH), to data corrupted with off-focal radiation alone (labeled OF), and to data corrupted with both of these effects plus compound Poisson and electronic noise (labeled BH, OF, Noise). [PAPER 1 above]

Development of acquisition and reconstruction strategies for cardiac perfusion CT
Dynamic contrast-enhanced computed tomography (CT) could provide an accurate and widely available technique for myocardial blood flow (MBF) estimation to aid in the diagnosis and treatment of coronary artery disease. However, one of its primary limitations is the radiation dose imparted to the patient. In collaboration with Adam Alessio at the University of Washington, we are exploring techniques to reduce the patient dose by either reducing the tube current or by reducing the number of temporal frames in the dynamic CT sequence and dealing with the resulting lower-quality data through a combination of sinogram restoration, fully iterative reconstruction, or both. We have developed a sophisticated simulation environment and used it to explore quantitative blood flow estimation models [1]. We have also extended sinogram restoration into 4D and compared it with Karhunen-Loeve smoothing [2], with clear benefit seen for sinogram restoration.

  1. Bindschadler, M., Modgil, D., Branch, K. R., La Rivière, P. J., Alessio, A., “Comparison of blood flow models and acquisitions for quantitative myocardial perfusion estimation from dynamic CT,” Med. Biol., 59, pp. 1533-56, 2014. PMCID: PMC4057043.
  2. Modgil, D., Alessio, A., Bindschadler, M., and La Rivière, P. J., “Sinogram smoothing techniques for myocardial blood flow estimation from dose-reduced dynamic CT,” Med. Imag. 1 (3), 034004 (2014);  doi: 10.1117/1.JMI.1.3.034004. PMCID: PMC4307866.
  3. Bindschadler, M., Modgil, D., Branch, K.R., La Rivière, P. J., andAlessio, A., Evaluation of static and dynamic perfusion cardiac computed tomography for quantitation and classification tasks,” Med. Imag.3(2), 024001 (2016). doi:10.1117/1.JMI.3.2.024001. PMCID: PMC4852211.
  4. Modgil, D., Bindschadler, M.; Alessio, A., La Riviere, P.,“Variable temporal sampling and tube current modulation for myocardial blood flow estimation from dose-reduced dynamic CT,” J. Med. Imag. 4 (2), 026002 (May 13, 2017); doi: 10.1117/1.JMI.4.2.026002. PMCID: PMC5429861.

Example CT images of simulated dynamic acquisitions. Images at (a) 4, (b) 14, and (c) 24 s postinjection of 100 ml of contrast for the 70 kg patient with a true flow of 3.0  ml/min/g are presented. [FROM 3 ABOVE]

Development of novel approaches to spectral CT
We have recently developed a number of algorithms for reconstructing images in spectral CT, with applications to both existing dual-energy systems and emerging photon-counting systems. In general, the algorithms seek to exploit the common structure expected in the raw data (among the images/sinograms acquired at different energies) or in the reconstructed basis-material images.

  1. Rigie D.S. and La Rivière P. J.,“Joint reconstruction of multi-channel, spectral CT data via constrained total nuclear variation minimization,” Phys Med Biol. 2015 Feb 21;60(5):1741-62. doi: 10.1088/0031-9155/60/5/1741. Epub 2015 Feb 6. PMCID: PMC4669200.
  2. Rigie, D. and La Riviere, P.J.,“Optimizing spectral CT parameters for material classification tasks,” Med. Biol. 61(2016) 4599–4622 ; doi: 10.1088/0031-9155/61/12/4599. PMCID: PMC5444336.
  3. Rigie D, Sanchez A. A., and La Rivière, P. J., “Assessment of Vectorial Total Variation Penalties on Realistic Dual-Energy CT Data,” Phys. Med. Biol. 62 (2017) 3284–3298. PMC: PMC5575889.
  4. Rigie, D. S., La Rivière P. J.,and Petschke, A., “Image domain pansharpening method and system for spectral CT with large pixel energy discriminating detectors,” Patent Application20150043795, Filed: Aug 7, 2013. Published Feb 12, 2015.

CT Images of a frozen turkey after denoising by a variety of total variation based methods, including a novel multi-channel TV algorithm we developed for multi-energy CT data. [FROM PAPER 3 ABOVE]

Development of contrast agents and algorithms for optoacoustic tomography
Proteases represent a particularly appealing target for molecular imaging, as they are overexpressed in a number of pathologies. While there are fluorescently labeled probes for protease imaging, these suffer the limitations of all optical molecular imaging strategies. Reflecting on these limitations, we conceived of the idea of an optoacoustic imaging molecular probe sensitive to proteases, in which the molecule’s absorption peak would shift upon being cleaved by proteases [2].

In parallel with these chemical and experimental developments, our group has developed novel image reconstruction strategies aimed at accounting for and correcting acoustic inhomogeneities. For example, we developed and published the first method for modeling and correcting for acoustic attenuation in optoacoustic imaging [1]. Dimple Modgil has developed methods for correcting speed of sound variations, has evaluated novel linear geometry reconstruction algorithms, and optimizing wavelength choice in optoacoustic imaging [3,4].

  1. La Rivière, P. J., Zhang, J., and Anastasio, M., “Image reconstruction in optoacoustic tomography for dispersive acoustic media,” Optics Letters,31,pp. 781–783, 2006.
  2. Green, A., Norris, J., Wang, J., Xie, Z., Zhang, H.F., andLa Rivière, P. J“In vitro testing of a protease-sensitive contrast agent for optoacoustic imaging,” Journal Biomedical Optics,15, 021315(8 pages), 2010.
  3. Modgil, D., Anastasio, M. A., and La Rivière, P. J., “Image reconstruction in photoacoustic tomography with variable speed of sound using a higher order geometrical acoustics approximation,” Journal Biomedical Optics,15, 021308(9 pages), 2010.
  4. Modgil, D. and La Rivière, P. J., “Wavelength optimization in optoacoustic imaging using the Cramer-Rao Lower Bound,” Phys. Med. Biol.,55, pp. 7231-7251, 2010.

Images reconstructed from noiseless (top row) and noisy (bottom row) optoacoustic signals without and with compensation of dispersion effects are shown in the left and right columns, respectively. [FROM PAPER 1 ABOVE]