2.1.

Instrumentation and Data Acquisition

Details of the measurement and reconstruction workflow are described in the Supplementary Material. The system currently utilizes three wavelengths: 670, 750, and 808 nm corresponding, respectively, to the main absorption peak of lipid-porphyrin nanoparticles under development for thermal dose enhancement during PTT,³⁰ an absorption peak of deoxygenated hemoglobin (Hb), and the $\mathrm{Hb}\text{-}{\mathrm{HbO}}_2$ isosbestic point. The MRI-compatible transrectal probe is cylindrical (diameter 25 mm) with $400\mu \mathrm{m}$ optical fibers arranged within an area (20 mm longitudinally × 16 mm transversally) centred 5 cm from the probe tip. The eight light delivery and eight collection fibers exit the probe perpendicular to and flush with the surface, as seen in Fig. 1. Powers of 20, 9, and 5.5 mW were used for the 670-, 750-, and 808-nm lasers, respectively. Grooves in the longitudinal and the transversal planes enable MRI positioning and orientation. Phantom measurements were made at 670, 750, and 808 nm, whereas ex vivo and in vivo measurements were made only at 750 nm to minimize the sampling time in following the dynamic coagulation changes: 28 s per scan using 8 sources when using 8 sources. Measurements in a reference phantom of known optical properties provide calibration factors for each source–detector pair,²⁸ repeated for each measurement session.

2.2.

Phantom Study

Phantom mixtures were prepared using laboratory-grade agar (BioShop Canada Inc., Burlington, Ontario, Canada) at 2% weight/volume as the bulk matrix, Intralipid^® 20% stock solution (Baxter, Mississauga, Ontario, Canada) for optical scattering, India Ink for optical absorber, and 5% w/v OMNIPAQUE 350 iohexol injection USP (GE Healthcare, Mississauga, Ontario, Canada) as a CT contrast agent and 0.8% NaCl as an antibacterial agent. A previously validated spatially resolved flat-top diffuse reflectance (DR) probe was used to measure the optical properties³¹ in a 6-cm diameter, 2.5-cm height phantom block. Separately, two cylindrical wells were fabricated in a 1000-ml container, one of 20-mm diameter representing the coagulated zone with altered optical properties and one of 25-mm diameter for the probe. Each well was 100-mm deep with an edge-to-edge separation of 2.7 mm. The sizes and separation of the wells were confirmed by cone-beam CT imaging: Fig. 2(a). The 20-mm well was filled to a height of $\sim 90\mathrm{mm}$ using six different concentrations of Intralipid and India Ink to represent zones of altered, i.e., coagulated, optical properties. The optical properties of each liquid inclusion and of the agar matrix were measured at 670, 750, and 808 nm using the DR probe that was independently calibrated in a commercial solid reference phantom (Biomimic™, INO, Quebec, Canada).²⁸

Fig. 2

Agar phantom with inclusion simulating the photocoagulation zone and holding the probe. Cone-beam CT images in (a) coronal and (b) axial views ($0.39\times 0.39\times 0.39{\mathrm{mm}}^3\text{voxels}$). (c) Top-view photograph.

The probe well was water filled for optical coupling. All measurements were repeated 5 times with the phantom covered with blackout fabric and were corrected for dark counts [Fig. 2(b)]. Reconstructions were performed using different initial estimates of the inclusion radius (6, 8, 10, and 12 mm) and length (10 and 45 mm) combined with different regularization values. Estimated errors in the reconstructed optical properties and shape parameters were calculated using the standard methods for nonlinear regressions based on calculating the Hessian matrix. The optical properties of the simulated coagulation zone and the agar phantom as measured by the DR probe were used as initial values in reconstruction and we also examined the effect of varying the initial ${\mu }_a$ and ${\mu }_s^{\prime }$ values, both individually and in combination, using the methodology shown in detail in Fig. S2 in the Supplementary Material. The reconstructions were based on NIRFAST³² meshing and were performed in MATLAB (The MathWorks, Inc., Natick, Massachusetts, USA). CT images were analyzed using ImageJ (National Institutes of Health, Bethesda, Maryland, USA).

2.3.

Ex Vivo Study

TRDOT monitoring of photocoagulation was performed in ex vivo bovine and porcine muscle using a setup similar to the prior preliminary experiments.²⁸ Briefly, two 8- to 12-mm-thick tissue slabs of either bovine or porcine muscle were used. The probe was fixed with the optical window facing upward on the table of a 1.5-T MRI system (Aera: Siemens, Toronto, Ontario, Canada). A first tissue slab was placed on top in close contact. A 1.7-mm diameter, 15-mm long diffusing-tip PTT treatment fiber delivering 5 W at 980 nm was placed on the tissue parallel to the long axis of the probe and centered on the optical window. The fiber and probe positions were confirmed by a fast MRI localization scan. ${\mathrm{T}}_2\mathrm{w}$ Turbo Spin Echo scans then were acquired pre- and post-PTT to image the photocoagulation zone in the axial plane (voxel size: $1\times 1\times 3{\mathrm{mm}}^3$; TR: 3600 ms; and TE: 57 ms). Repeated MRT scans (GRE; TR: 46 ms; TE: 9 ms; $1\times 1\times 3{\mathrm{mm}}^3\text{voxels}$) and 750 nm DOT scans were acquired simultaneously for $\sim 140\mathrm{s}$ immediately prior to PTT, $\sim 700\mathrm{s}$ during treatment delivery, and $\sim 300$ to 400 s during tissue cool down. The tissue slabs were then separated and the final photocoagulation diameter was measured using a ruler in the longitudinal and cross-sectional directions.

Rather than a reference phantom, the native tissue optical properties, either measured using the DR probe or from the literature were used for TRDOT calibration, to minimize any effect of tissue–probe contact. These properties were input to a forward model of the TRDOT signals, which were compared to the measured signals to calculate individual source–detector calibration factors. The transverse and longitudinal radii and the optical properties of the coagulation zone were reconstructed using the SBDOT algorithm. For comparison, following $3\times 3$ Gaussian smoothing, the 55°C contour from the MRT images was taken as the photocoagulation boundary at each treatment time point, noting in particular the distance between the coagulation boundary and the probe. The PTT fiber position and final photocoagulation boundary were also estimated from the posttreatment T2w MR images. As detailed in the Supplementary Material S2, a range of initial optical properties of the coagulation tissue were used in the SBDOT reconstructions to test the sensitivity to these starting values.

2.4.

In Vivo Study

In vivo PTT of the canine prostate, which is similar in size to the adult human prostate, was performed in 6 male beagle dogs (12- to 14-month old, 12 to 14 kg weight: Marshall Bioresources Inc., NY, USA). All procedures were performed under general anesthesia following 12 h fasting and 3 warm-water enemas. Laparotomy, as required for a different concurrent study, allowed direct access to the prostate. For the laparotomy, an incision was made from the umbilicus to the pubis, exposing the bladder, prostate, and kidney. The bladder was retracted toward the umbilicus and sutured to the abdominal wall to minimize free-organ motion, which also exposed and stabilized the prostate. The PTT treatment fiber was inserted through a 14-G cannula/introducer to a depth of $\sim 1\mathrm{cm}$ [see Fig. 3(a)].

Fig. 3

In vivo setup showing (a) insertion of the treatment fiber at laparotomy; (b) the TRDOT probe placed in the rectum with the optical window facing the center of prostate; and (c) the flat-top DR probe held rigidly on the ventral surface of the exposed prostate.

The TRDOT probe was inserted into the rectum in a sterile condom with lubricant gel and aligned with the prostate [Fig. 3(b)] by palpation. The optical properties of the untreated prostate at 750 nm were measured using the DR probe in direct contact with the anterior prostate surface via the laparotomy [Fig. 3(c)]. All TRDOT scans were performed at 750 nm, starting immediately prior to PTT. The first half of the treatment was given using a laser power of 3 W, which was increased to 5 W at the midpoint. This protocol minimized blood coagulation at the fiber tip while maximizing the coagulation zone. It also tested the ability of TRDOT to track the growth rate of the coagulation zone. Two different TRDOT monitoring protocols were evaluated: (i) PTT over 744 s, with TRDOT monitoring using 8 light sources (28 s per complete scan) and (ii) PTT for 210 s and TRDOT using only 4 sources to reduce the acquisition time to 14 s. In both cases, TRDOT monitoring continued for $\sim 100\mathrm{s}$ following the end of treatment.

The SBDOT algorithm was used to reconstruct the expanding photocoagulation zone and to measure the pre- and post-PTT optical properties. This utilized a NIRFAST mesh comprising four regions: photocoagulation zone, untreated prostate, rectum, and surrounding tissues. TRDOT data were calibrated as per the ex vivo study.

3.1.

Phantom Study

The ${\mu }_a$ and ${\mu }_s^{\prime }$ values of the agar phantom were 0.009 and $1.08{\mathrm{mm}}^{-1}$, 0.006 and $1.07{\mathrm{mm}}^{-1}$, and 0.004 and $1.01{\mathrm{mm}}^{-1}$ at 670, 750, and 808 nm, respectively, with coefficients of variation of 9.8% for ${\mu }_a$ and 4.0% for ${\mu }_s^{\prime }$ ($n=10$). The optical properties of the liquid inclusions simulating the coagulation zone as calculated from the TRDOT measurements are plotted against the reference DR probe measurements in Fig. 4. The SBDOT reconstructed radius was $r=10.8\pm 0.2\mathrm{mm}$ (where the error represents the average across measurements at the three wavelengths), compared with the true value of 10 mm. Table 1 summarizes the average reconstructed values at each wavelength.

Fig. 4

Comparison of (a) ${\mu }_a$ and (b) ${\mu }_s^{\prime }$ of the liquid inclusions measured by the TRDOT and flat-top DR probes. The signal was below the measurement limit at 670 nm with the highest absorption. Lines of equality are indicated.

Table 1

Inclusion optical properties showing the target range of μa and μs′, the measured range over 670, 750, and 808 nm using the DR probe and the TRDOT probe.

TRDOT recovered the scattering coefficients within $\pm 3\%$ but systematically underestimated the absorption by about 20%. The impact of errors in reconstructed optical properties on the reconstructed size was assessed by fixing the optical properties over a range from $-100\%$ to $+300\%$ of the true values, whereas the scattering coefficient was varied from $-100\%$ to 200%. Variations in ${\mu }_a$ of up to $\pm 30\%$ and in ${\mu }_s^{\prime }$ of up to $\pm 40\%$ caused $<0.7\mathrm{mm}$ variations in the reconstructed inclusion radius. The reconstructed longitudinal size converged to the maximum dimension (20 mm) of the optical window, which was shorter than the actual value of 45 mm. However, this was as expected and the window is long enough to encompass most clinical cases, in which the diffusing optical fibers for treatment light delivery rarely exceed 30 mm.

Generally, the reconstructions were insensitive to varying the initial input values for the optical properties and lesion size. For example, varying the initial estimate of the transverse lesion radius between 2 and 12 mm gave a coefficient of variation of only $\sim 0.2\mathrm{mm}$ in radius across 21 samples. When the initial optical properties were varied by $-40\%$ to $+80\%$ of the true values, the reconstructed radius was $10.5\pm 0.6\mathrm{mm}$ averaged over all reconstructions ($n=99$) from the 670-, 750-, and 808-nm data. Only when these initial estimates were varied by more than $+100\%$ or -40% did the reconstructed transverse radius deviate from the true values by more than 2 mm (Fig. S3 in the Supplementary Material). Published optical properties of human in vivo prostate range widely,^33–36 from 0.01 to $0.16{\mathrm{mm}}^{-1}$ for ${\mu }_a$ and from 0.1 to $4{\mathrm{mm}}^{-1}$ for ${\mu }_s^{\prime }$ between 730 and 790 nm, depending on measurement technique. The effective attenuation coefficient ${\mu }_{\mathrm{eff}}$ is more consistent at between 0.20 and $0.33{\mathrm{mm}}^{-1}$, and within a patient group has a lower standard deviation than ${\mu }_a$ and ${\mu }_s^{\prime }$.³⁴ If the primary source of these variations is technical rather than biological, then the confidence interval for the starting optical properties should converge as the clinical data using TRDOT during PTT increases, thereby further improving reconstruction of the coagulation zone boundary.

3.2.

Ex Vivo Study

Table 2 summarizes the initial (“native”) optical properties of each tissue used for the reconstructions as well as the coagulated tissues. Due to low absorption and scattering of porcine muscle, the native optical properties were not accurately recovered using the DR probe, so that the initial estimates were also extrapolated from literature values at 600 nm³⁷ and 750 nm.^38–43

Table 2

Optical properties of native and coagulated tissues at 750 nm measured with the TRDOT and fat-top DR probes, and compared to the literature values.37–43

For the coagulated tissues, the measured scattering coefficients for bovine muscle using either the DR or TRDOT probes are consistent with literature values, while those for porcine tissue are $\sim 50\%$ of the literature value. The absorption coefficients vary markedly between the DR probe, TRDOT probe, and literature values. Most likely this reflects differences in blood content due to tissue collection and handling, as could be resolved by measuring at wavelengths above and below the Hb isosbestic point.

Figure 5 shows photographs of the coagulation zone, each at 15 min post-PTT, MRI ${\mathrm{T}}_2\mathrm{w}$ images overlaid with an estimate of coagulation zone boundary based on contrast within each image, as well as MRT images taken immediately at the end of treatment, with the 55°C contour shown as a black dashed line and the SBDOT lesion shown as a white dashed line. The photocoagulation zones are clearly visible on the photographys and T2w images.

Fig. 5

Examples of post-PTT coagulation zones in ex vivo tissues: (top) porcine and (bottom) bovine. (a), (b) Photographs at 15 min post-PTT in the coronal plane with TRDOT boundary shown as dashed black lines. (c), (d) Transverse MR ${\mathrm{T}}_2\mathrm{w}$ images with the lesion boundary outlined in green. (e), (f) False-color MRT images showing the 55°C temperature contours (black dashed line) immediately at the end of PTT. White dashed circles represent the reconstructed TRDOT boundaries centered on the treatment fiber.

In Fig. 6, the temporal evolution of the coagulation zone size in the two tissues from the SBDOT reconstruction is compared with the 55°C MRI contours, whereas Table 3 summarizes the final lesion sizes. Note that the temperature starts to fall as soon as the treatment laser is turned off, whereas the reconstructed coagulation zone remains constant, as expected. In general, SBDOT systematically overestimates the lesion size compared with MRT in the earlier stages of treatment when the coagulation zones were small, albeit by $<1\mathrm{mm}$. However, the two monitoring methods came into close agreement as the coagulation zone expanded and as the coagulation boundary progressed toward the transrectal probe. This was confirmed also by the direct photographic measurements at the end of treatment.

The accuracy in the coagulation size derived from MRT is limited by the axial pixel resolution ($1\times 1{\mathrm{mm}}^2$) and uncertainty in the temperature assumed for coagulation: the transverse radius increases by up to 0.6 mm as coagulation thresholds were increased from 53°C to 57°C. The accuracy of the TRDOT reconstructions was estimated by the following bootstrap procedure. For each measurement set, i.e., at each treatment time point, the reconstruction was repeated 50 times using random sets of the 64 source–detector pairs⁴⁴ and the standard deviation in the coagulation size was calculated. The overall uncertainty in the reconstructed coagulation zone radius was $\pm 0.6\mathrm{mm}$ transversely and $\pm 3\mathrm{mm}$ longitudinally. This could not be attributed to instrument noise, which was $<1\%$ for all source–detector pairs, both at baseline and during PTT monitoring.

Fig. 6

Comparison in ex vivo (a) porcine and (b) bovine muscle of the coagulation radius given by the 55°C MRT contour and from the SBDOT reconstruction. PTT started at $t=0\mathrm{s}$ and stopped $t=700\mathrm{s}$ (dashed green line).

Table 3

PTT lesion radius (mm) at the end of treatment from T2w MRI and MR thermometry (55°C threshold), TRDOT, and direct measurement from photographs.

Since the TRDOT probe operates in reflectance geometry, it is less sensitive to optical changes occurring deep in the tissue. Consequently, early in the treatment, the changes in detected optical signal are small, resulting in an underestimate of the size of the coagulation zone. However, this is not of concern in this clinical application. If required, it could be addressed by placing interstitial source/detector fibers within the treatment zone but at the cost of added complexity and invasiveness. It is also possible that our light modeling method is suboptimal, especially at the boundary of the native and coagulated tissues. We are investigating the mesh resolution and assumption of diffusion theory by comparisons to Monte Carlo simulations.

Reconstruction of the final coagulation lesion size was generally insensitive to the initial estimates of the optical properties during reconstruction: varying the initial ${\mu }_a$ and ${\mu }_s^{\prime }$ values by a factor of 2 typically resulted in $<10\%$ error in the reconstructed radius.

3.3.

In Vivo Study

Table 4 summarizes the optical properties of native prostate tissue in vivo at 750 nm measured by the DR and the TRDOT probes, those of coagulated prostate measured by the TRDOT probe and, for comparison, literature values of ex vivo samples measured using a double integrating sphere method.⁴⁵ The DR values were used as the initial parameters in the reconstruction algorithm. The ${\mu }_a$ values of native prostate measured by the two probes were lower than the published ex vivo canine prostate values at 1064 and 633 nm.⁴⁵ The absorption decreased by about 40% upon photocoagulation. For ${\mu }_s^{\prime }$ in native prostate, the DR measurement was almost twice the TRDOT value, likely due to the former measuring only anterior prostate and the latter having contributions from 2 to 3 mm of rectal wall and connective tissue in addition to the posterior prostate. ${\mu }_s^{\prime }$ measured by the TRDOT probe showed nearly twofold increase upon coagulation, which is comparable to the 1.4-fold and 1.6-fold increases reported in ex vivo prostate samples at 1064 and 633 nm, respectively.⁴⁵

Table 4

Reconstructed optical properties at 750 nm (mean ± 1 standard deviation) for native and coagulated canine prostate tissue in vivo (n=6 dogs). Measured values are compared to those measured using double integrating sphere measurement in Ref. 45 at 1064 and 633 nm.

The interpretation of the tissue response as “binary,” namely coagulated versus unaltered native tissue, is clearly an oversimplification in vivo, since temperature-induced homeostatic changes in blood perfusion in the prostate and/or rectum^46,47 will affect the optical measurements. Blood accumulation around the PTT fiber (Fig. 7) is a further potential confounding factor that would impact primarily the reconstructed absorption value. The SBDOT reconstructions gave coagulation zone radii between 6 and 12 mm across the 6 animals and converged consistently post-treatment. During PTT, however, the reconstruction is affected markedly also by heat-induced physiological changes within the DOT-interrogated region. In our previous preliminary in vivo canine studies, we reported large cyclical fluctuations in TRDOT signals during PTT²⁷ and these are apparent also in the examples shown in Fig. S4 in the Supplementary Material. However, the photocoagulation zone expands monotonically and smoothly, suggesting that this behavior is an artefact due to physiological responses to either the heating and/or mechanical stresses from probe placement: it remains to be seen if these fluctuations occur also in human patients during PTT. As shown in Fig. 8, applying five-point smoothing to the reconstructed axial radius largely eliminated the problem, allowing the true coagulation boundary growth to be monitored. This may also be improved further using multiple wavelengths in order to decouple the contribution of changes in blood flow from the true coagulative changes. An additional factor that adversely affects the accuracy of the coagulation monitoring is 28 s TRDOT scan time, during which the photocoagulation radius grows typically by $\sim 0.5\mathrm{mm}$, so that the system is not sampling the same coagulation zone between the start and end of each scan.

Fig. 7

Representative photographs of the photocoagulation zone (indicated by arrows) in four canine prostates excised shortly after completion of the treatment located (a) near the urethra, (b) upper left quandrant, (c) lower left quadrant, and (d) right side. The transrectal probe is located near the dorsal side.

Fig. 8

Examples of SBDOT reconstructions during dual-power PTT in vivo with (a) four or (b) eight light sources activated with (red) and without (black) 5-point smoothing. PTT starts at $t=0\mathrm{s}$ and stops at the green vertical line.

Figure 8 presents two specific cases of in vivo PTT on canine prostate where the treatment power was increased during PTT. For the case shown in Fig. 8(a), TRDOT used four light sources and all eight detection channels. The post-smoothing reconstruction resulted in a final reconstructed radius of $6.8\pm 0.2\mathrm{mm}$ compared to 7.0 mm by direct visualization after resection: here the treatment fiber and coagulation front were located at 10 and 3 mm, respectively, from the TRDOT probe surface. Figure 8(b) shows an example using 8 light sources and 8 detection channels that yielded a final reconstructed radius of $7.7\pm 0.2\mathrm{mm}$ compared to 8.0 mm directly with the PTT fiber and coagulation front at 11 and 3 mm from the TRDOT probe surface.

In both cases, when the treatment laser power was increased at the treatment midpoint, there was a corresponding increased rate of coagulation growth as measured by TRDOT, showing that TRDOT is properly tracking the dynamic changes, as required. We note that, even for closely matched dogs and using normal prostate, there is considerable animal-to-animal variation in the thermal response, which highlights the need for such monitoring in individual cases. Experiments to obtain point-by-point quantitative confirmation versus MR thermometry are planned and are expected to be successful since the final reconstructed size of the coagulation zone is consistent with direct measurements post mortem.