This website is maintained by Jem Hebden.

Last update of this page: October 26, 2002.

Phantom Studies

Miscellaneous results achieved using MONSTIR since October 1998 are presented below, in roughly chronological order. These results have been selected in order to illustrate the expanding capabilities and sophistication of the instrumentation and the reconstruction software as progress is made towards full clinical implementation.

1. 2D Imaging of a cylindrical phantom
2. Multiple slice imaging of a cylindrical phantom
3. 3D Imaging of a conical breast phantom
4. 3D Imaging of a realistic head phantom

1. Two-dimensional imaging of cylindrical phantom

A solid tissue-equivalent phantom was constructed in the form of a cylindrical block, 85 mm in length and 70 mm in diameter [1]. The optical properties at a wavelength of 800 nm corresponded to a transport scatter coefficient of µ_s' = 1.0 mm^-1 and an absorption coefficient of µ_a = 0.01 mm^-1. Embedded along the length of the phantom are three regions with coefficients having the following relative values: A) 5µ_s' and µ_a, B) µ_s' and 5µ_a, and C) 2µ_s' and 2µ_a. Each region is 8 mm in diameter, and located 15 mm from the central axis of the phantom. All 32 source fibres and 32 detector fibre bundles were arranged around the circumference of the phantom and held in position by a black plastic ring, as shown in figure 1. Each source was activated in sequence, and data were recorded for 30 seconds at 22 detector positions immediately opposite the source in a fan-beam configuration. The experiment was automatically executed under computer control. Further experimental details are described elsewhere [2].

Figure 1: The 32 sources and detectors arranged around the tissue-equivalent phantom.

The data were calibrated to reduce the effects of systematic noise and the finite temporal response of the system, and a correction was performed to account for the fact that the image reconstruction is based on a 2D forward model [3]. Three data types were derived from each of the 704 acquired TPSFs: the mean flight time, the variance about the mean, and the normalised Laplace Transform (with a coefficient of 0.005 ps-1). Then, using an appropriate FEM mesh (consisting of 7392 triangular elements) describing a uniform circular cross-section of the phantom, the TOAST algorithm was employed to derive simultaneous maps of the absorbing and scattering properties. We used the non-linear conjugate gradient method with median filtering after each update. The initial guess was obtained by first globally fitting homogeneous µ_a and µ_s' values to all the data. The 2D/3D correction of the data, combined with the symmetry of the phantom and the fact that the optodes were arranged in a single plane enabled TOAST to employ a 2D forward model, which results in a considerable reduction in computation time compared to a 3D model. The results are shown below in figure 2. The absorption image is on the left, and the scatter image is on the right. Qualitative agreement with the known locations and properties of the three regions is remarkably good. The absorption map is dominated by region B, whereas the scattering map clearly indicates the highly scattering region A, and the smaller contrast of region C.

Figure 2: Absorption (left) and scatter (right) images of the phantom.

2. Multiple slice imaging of a cylindrical phantom

MONSTIR was used to generate multiple slice images of a phantom without uniform properties along the axial direction, while still using a computationally fast 2D reconstruction algorithm [4]. The cylindrical phantom, 140 mm in length and 70 mm in diameter, has optical properties corresponding to µ_s' = 0.85 mm^-1 and µ_a = 0.01 mm^-1. Inserted inside are three small cylinders, 10 mm in length and 8 mm in diameter. Each is centred 15 mm from the central axis, at different heights above the base: A) 50 mm, B) 75 mm, and C) 100 mm. The three cylinders have coefficients with relative values of A) 10µ_s' and µ_a, B) 5µ_s' and 5µ_a, and C) µ_s' and 10µ_a. The ring of fibres and detectors (as shown in figure 3) were held around the circumference of the phantom at 14 different heights, at intervals of 5 mm between 40 mm and 105 mm above the base. TPSFs were acquired automatically for 15 seconds for each source position at each height.

Figure 3: Scatter (top row) and absorption (bottom row) images at multiple heights.

After performing a 2D/3D correction of the data [3], scatter and absorption images were generated for each vertical position of the source-detector ring, using the mean, variance, and Laplace Transform of the TPSFs. The results are shown in figure 3 above and in the animation below.

Figure 4: Animated version of figure 3.

A larger version of this movie (433 kbyte) may be viewed by clicking here. The positions of the three embedded cylinders become apparent in the absorption and scattering images at the appropriate heights above the base (labelled in yellow in figure 3). As for the earlier experiment described above, the scattering images generally exhibit higher contrast and spatial resolution. Note that scatter and absorbing properties are not perfectly separated: some evidence of the scattering cylinder A (at a height of 50 mm) appears in the absorption images, and evidence of the absorbing cylinder B (at a height of 100 mm) appears in the scatter image. For more information about this experiment, see the paper by Schmidt et al [4]. The data generated by this experiment have also been used to test a three-dimensional image reconstruction scheme, described by Arridge et al [5].

3. Three-dimensional imaging of a conical breast phantom

A full 3D image reconstruction was performed using data acquired using MONSTIR and solid phantoms with a conical geometry [6]. An identical pair of phantoms were constructed with optical properties of µ_s' = 0.8 mm^-1 and µ_a = 0.007 mm^-1. Inserted inside one of the phantoms are three small cylinders, 10 mm in length and 10 mm in diameter, with relative optical properties of A) 2µ_s' and µ_a, B) µ_s' and 2µ_a, and C) 2µ_s' and 2µ_a. Three rings of sources and detectors were held around the circumference of the phantom as shown in figure 5 below. The diameters of the rings were 55 mm, 82 mm, and 109 mm. The number of source-detector pairs in each ring were 8, 8, and 16 respectively. The embedded cylinders were located within the plane of the middle ring.

Figure 5: The 32 sources and detectors arranged around the conical breast phantom.

TPSFs were acquired automatically for 30 seconds for each source position. Similar data were acquired using the homogeneous cone phantom without the embedded cylinders. TOAST employed a 3D conical FEM mesh to reconstruct 3D images from the differences between the mean flight times and log intensities recorded on the two phantoms. The absorption and scattering images are illustrated in figure 6 as a series of transverse slices, corresponding to various heights above the base of the phantom. The image at the height of 78 mm represents the slice containing the centres of the embedded cylinders.

Bright regions corresponding to all three embedded cylinders are evident, located at the expected positions. Cylinder C, representing a perturbation in both scatter and absorption, is the most dominant feature in both sets of images. The observed contrast for both absorption and scattering images is roughly ten percent of background instead of the expected 100 percent. This may be at least partly due to the finite spatial resolution, since the observed features are roughly twice the diameter of the actual cylinders, or roughly eight times the volume. An animated version of figure 6 is shown below.

Figure 7: Animated version of figure 6.

A larger version of this movie (367 kbyte) may be viewed by clicking here. For more information about this experiment, see the paper by Hebden et al [6]. The conical fibre holder is now being employed to evaluate the method as a means of imaging the breasts of suitable patients and volunteers.

4. Three-dimensional imaging of a realistic head phantom

A solid epoxy resin phantom was constructed in the shape of a premature infant head. The circumference was 24 cm, corresponding to a baby of about 26 weeks gestation. The head phantom contained a spherical hollow region with a diameter of about 55 mm. The solid outer shell was assigned optical properties of µ_a = 0.01 mm^-1 and µ_s' = 1.0 mm^-1 at 780 nm. During experiments, the hollow region was filled with epoxy resin without hardener, which therefore remained liquid. A helmet made from thermoplastic and light-absorbing foam was constructed to couple 29 connectors, each holding a source and a detector fibre, onto the phantom. An FEM mesh was generated for the phantom using surface information derived from a x-ray CT-scan (click here for details).

Figure 8: The realistic head phantom and fibre holder helmet held upside down.

Figure 9: A schematic representation of the head phantom and fibre holder helmet.

The phantom was first filled with liquid resin with the same properties as the solid outer shell. Two epoxy resin cylinders, of height and diameter 7 mm, were suspended within the upside-down phantom, as shown in figures 8 and 9. One cylinder had five times greater absorption, and the other had five times greater scatter. MONSTIR was used to record data with and without the cylinders present. Each of the 29 sources was illuminated sequentially for 10 seconds. The phantom was then refilled with liquid resin with increased absorption (µ_a = 0.015 mm^-1 and µ_s' = 1.0 mm^-1) to simulate the higher absorption of the brain compared to the scalp and skull. Another pair of image data sets were then acquired, with and without the cylinders. Images were reconstructed using TOAST from measurements of the ratios of intensities and the differences in mean photon flight times between the data acquired with and without the two cylinders. Reconstruction required about 30 minutes per iteration on a 1.4 GHz Athlon PC with 1 GB RAM. After 25 iterations, no further improvement in image quality was observed.

Figure 10: Transverse (top) and sagittal (bottom) slices through the 3D images of the solid phantom with uniform optical properties.

Figure 10 shows the reconstruction of the phantom with the liquid having the same properties as the outer shell. The top row shows transverse slices and the bottom row shows sagittal slices through the 3D image. The left hand column is a schematic representation of the phantom showing the approximate location of the cylinders. The middle and right hand columns show the scatter and absorption images, respectively. Images of the phantom with the higher absorbing liquid inside were also generated. However, it was observed that (not surprisingly) more accurate images were acquired if the reconstruction began with prior knowledge of the optical properties in the outer and inner regions. Figure 11 shows the images generated when this starting condition was used.

Figure 11: Transverse (top) and sagittal (bottom) slices through the 3D images of the solid phantom with higher absorbing liquid.

1. Hebden, JC, Schmidt, FEW, Fry, ME, Schweiger, M, Hillman, EMC, Delpy, DT, and Arridge, S (1999): Simultaneous reconstruction of absorption and scattering images using multi-channel measurement of purely temporal data, Opt. Lett. 24, 534-536.
2. Schmidt, FEW, Fry, ME, Hillman, EMC, Hebden, JC, and Delpy, DT (2000): A 32-channel time-resolved instrument for medical optical tomography, Review of Scientific Instruments 71(1), 256-265.
3. Hillman, EMC, Hebden, JC, Schmidt, FEW. Arridge, SR, Schweiger, M, Dehghani, H, and Delpy, DT (2000): Calibration techniques and datatype extraction for time-resolved optical tomography, Review of Scientific Instruments 71(9), 3415-3427.
4. Schmidt, FEW, Hebden, JC, Hillman, EMC, Fry, ME, Schweiger, M, Dehghani, H, Delpy, DT, and Arridge, SR (2000): Multiple slice imaging of a tissue-equivalent phantom using time-resolved optical tomography, Applied Optics 39, 3380-3387.
5. Arridge, SR, Hebden, JC, Schweiger, M, Schmidt, FEW, Fry, ME, Hillman, EMC, Dehghani, H, and Delpy, DT (2000): A method for 3D time-resolved optical tomography, International Journal for Imaging Systems and Technology 11, 2-11.
6. Hebden, JC, Veenstra, H, Dehghani, H, Hillman, EMC, Schweiger, M, Arridge, SR, and Delpy, DT (2001): Three dimensional time-resolved optical tomography of a conical breast phantom, Applied Optics 40, 3278-3287.
7. Gibson, A, Yusof, R, Dehghani, H, Riley, J, Everdell, N, Richards, R, Hebden, JC, Schweiger, M, Arridge, SR, and Delpy, DT (2003): Optical tomography of a realistic neonatal head phantom. Applied Optics 42, 3109-3116. Download PDF file.