Time Resolved Diffuse Optical Tomography using a Digital Light Processor Vivek Venugopal1, Jin Chen1, Frederic Lesage2, Xavier Intes1 1
Department of Biomedical Engineering, Rensselaer Polytechnic Institute, Troy New York 12180 Département de génie électrique and Institut de génie biomédical, École Polytechnique de Montréal, Québec, H3C 3A7, Canada
2
Author e-mail address:
[email protected]
Abstract: We experimentally demonstrate that time-resolved measurements of broad spatial patterns transmitted through an optically heterogeneous medium retain their temporal characteristics providing robust signals for efficient decoupling of optical properties and allowing fast quantitative high-resolution reconstruction. ©2010 Optical Society of America OCIS codes: (170.6920) Time-resolved imaging; (110.0113) Imaging through turbid media
1. Introduction Diffuse Optical Tomography (DOT) has undergone significant development over the past decade with developments in novel laser and detector technology. The development of DOT systems operating in the time-domain and frequency-domain has improved the sensitivity and quantitative accuracy of the system while reducing the inherent ill-posedness of the problem. Time resolved measurements provide more quantitatively accurate reconstructions with higher resolution in a point sourcedetector arrangement when compared to Frequency domain or steady state measurements. However due to the diffused nature of light propagation in tissues, the resolution of these reconstructions obtained using point source-detectors are limited to the millimeter range. The issue can be resolved by increasing the number of source-detector pairs employed by the system [1], but this requires a lengthy acquisition time, making the solution inviable for clinical applications. An alternate approach involves the use of spatially-modulated wide-field patterns which provide information equivalent to dense source-detector measurements while employing significantly fewer source patterns. With the fundamental idea being the projection of significant modes present in the dense point measurements, much of the research has focused on the use of varying spatial frequencies/Fourier modes and their measurement in steady-state or Frequency domain [2-3]. Recent work has also demonstrated the propagation of the said spatial frequencies in time-domain [4] allowing for a potentially fast and accurate high resolution DOT system. We propose that spatially varying wide (low frequency) patterns generated using ultra-short pulses propagating through a scattering medium and detected at discrete point detectors will retain their temporal characteristics and will provide significant resolution and sufficient contrast when measuring perturbations in an optically heterogeneous medium. Further, we will demonstrate that the use of time-gates data type will allow an efficient separation of absorption and scattering contrasts when employing wide-field patterns. This will lead to the development of a simple, fast and accurate DOT system. 2. Methods A time-resolved instrumentation was developed using tunable Ti-Sapphire laser (Mai Tai HP) with 100fs pulses as the source operated at 700nm. The source is then conditioned using a closed loop power control system (providing up to 400 mW at the output) before being injected into a 50μm multi-mode fiber (NA 0.22). The fiber is connected to a collimating lens which
Fig. 1. (left)A: MaiTai HP, B:Power Control, C: Fiber coupler, D: Collimator-Expander assembly, E: Discovery DMD board, F: 75mm reimaging lens, G: Imaging chamber, H: ICCD camera, I: Optical trigger generator, J: HRI, K: Trigger delay unit. (right) DLP board with the DMD showing a user controlled pattern.
projects a 1mm source on a beam expander. The output from the beam expander is a 1.5cm diameter source which projected on a DLP board having a 1024x768 resolution digital micro-mirror device ( DMD Discovery 1100). The patterns are generated using the DLP software by loading user-generated indexed images. The reflected pattern is refocused on the imaging stage using a 75mm bi-convex lens to obtain a 4.5cmx3cm area of illumination. Time-resolved measurements are made using a gated Intensified CCD camera (Picostar HR, Lavision) and this facilitates the use of a dense arrangement of several discrete point detectors. The high-rate intensifier controlling the shutter can be operated at a minimum of 200ps gate width in sync with the laser pulses. The trigger conditioning unit can provide 1ps temporal resolution. The system was found to have a jitter of 5ps and drift of 5ps/hr. Based on prior characterization, the system is operated at 300ps gate width and 20ps gate interval with measurements recorded over a 2.2ns time window providing 110 time-gate measurements. The instrument response function (IRF) measured by projecting the pattern on a white paper, had a full-width at half maximum (FWHM) of 230ps. This is comparable to the FWHM obtained using point source (220ps) indicating minimal effect of the pattern generation optics. Moreover, the entire field of the pattern was found to have uniform t0 (photon arrival time) at each detector employed signifying a temporally uniform pattern. The experiments were performed on a 2cm thick homogeneous liquid phantom (μa = 0.05 cm-1; μs' = 9.0 cm-1) constructed using Intralipid and India Ink. In the second part of the experiment, two 9mm diameter tubes were suspended in the phantom with an Intralipid-Ink mixture providing contrast in absorption and scattering in each of the inclusions. 36 patterns used in prior in-silico studies were used to illuminate the phantom [5]. It should be noted that speckles resulting from the multi-mode fiber were magnified upon projection leading to a non-uniform intensity distribution. This issue is however obviated by recording the homogeneous and heterogeneous measurements canceling out systemic errors. Another issue with the implementation of a DLP chip is the reduced photon density on the source plane necessitating the use of higher power, which is still within the maximum exposure limits. Moreover, this problem is alleviated due to lossy free space coupling on the DLP chip and presence of higher order diffraction patterns at the chip. The above experiments were performed at 60mW power incident on the imaging chamber. 3. Results Figure 2(b) shows the steady state measurement of the test pattern shown in Figure2(a) after propagating through the homogeneous slab phantom. The smearing of the pattern is observed due to the diffusive propagation through the phantom. The temporal point spread functions (TPSF) obtained at discrete detectors are shown in Figures 2(c)-(d). The TPSF are recorded at detectors 1mm apart along the x-axis and y-axis, as shown in 2(a) in blue and green respectively. It may be noted that variation in signal intensity is observed due to the speckled pattern is propagated in CW as well as time-domain. Further, Figure 2(d) shows that the spatio-temporal characteristics of the pattern are preserved after diffusion through the phantom. Also, the temporal behavior is consistent with the TPSF measured using point source-detector pairs.
Fig. 2. a) Test pattern b) Steady-state measurement of the pattern after diffusion c) TPSF measured along the axis (blue detectors, d) TPSF measured along the y-axis (green detectors).
Figure 3 shows the contrast observed using the same pattern with an heterogeneous phantom. The phantom shown in Figure 3(a) has an 8-fold contrast in absorption in the left inclusion and two-fold contrast in scattering in the right inclusion. Contrast determined using the Rytov approximation is shown in Fig 3(b)-(e) for the steady-state and three time-gates (rising half-max, peak and falling half-max respectively). The early gates show higher sensitivity to scattering contrasts and have a higher resolution , similar to the temporal behavior observed in the point source detector pairs,. The later gates only show a
contrast in absorption.
Fig. 3. Contrast measures for phantom shown in (a) using b) CW c) rising half-max gate d) peak gate e) falling half-max gate. Pattern shown in Fig. 2(a) is applied to the region marked by the red box.
4. Discussion The results of the preliminary investigations presented above establish the following two facts. First, the temporal behavior of wide-field patterns measured at discrete detectors is similar to the TPSF measured using point source-detector pairs. Owing to the low frequency content in the patterns, they can be transmitted farther through thick tissues. Second, the TPSF retains the temporal characteristics critical in the use of time-gates as a data type. The lack of resolution provided by low frequency patterns necessitates the use of higher spatial frequencies. However, owing to the temporal fidelity of the signals measured using wide-field patterns, the higher resolution afforded by the use of time-gates can be used to impart higher resolution to the reconstruction. Another advantage of this technique is in the efficient separation of the scattering and absorption contrasts. Figure 3 shows that the early gates are more sensitive to scattering perturbations and the late gates only show the absorptive perturbations. It may also be noted that the entire protocol of scanning the 4.5cmx3cm and measuring signals at high temporal resolution was completed in under 9 minutes. An equivalent dense point source-detector arrangement (2mm separation) would require ~45 minutes of acquisition time. In conclusion, the advantages offered by the use of low frequency patterns in the time domain exist in their ease of implementation as well as the rich information content provided by the temporal characteristics of the measurements. This coupled with the short acquisition time is indicative of a fast and accurate DOT system. 4. References [1] S. D. Konecky, G. Y. Panasyuk, K. Lee, V. Markel, A. G. Yodh and J. C. Schotland, "Imaging complex structures with diffuse light", Optics Express 16: 5048-5060 (2008). [2] S. Belanger, M. Abran, X. Intes, C. Casanova and F. Lesage, "Real time Diffuse Optical Tomography based on Structured Illumination", Journal of Biomedical Optics , Volume in press. [3] D. J. Cuccia, F. Bevilacqua, A. J. Durkin, F. R. Ayers and B. J. Tromberg,"Quantitative optical tomography of sub-surface heterogeneities using spatially modulated structured light",Journal of Biomedical Optics 14, 02012 (2009). [4] A. Bassi, C. D'Andrea, G. Valentini, R. Cubeddu, and S. Arridge, "Temporal propagation of spatial information in turbid media", Optics Letters 33: 28368 (2008). [5] J. Chen, V. Venugopal, F. Lesage and X. Intes, "Time gated diffuse optical tomography with wide-field strategies", Journal of Biomedical Optics , Volume submitted.
